Synthflow’s Voice AI With Memory: The Contact Center Breakthrough
- Jörn Menninger
- 13 minutes ago
- 51 min read
Management Summary
Voice AI has matured faster than most of us expected. Not in the flashy “AGI is here” sense, but in the quiet, structural ways that matter to contact centers: latency, memory, workflow execution, and reliability at scale.
So when I sat down with Hakob Astabatsyan, Co-Founder and CEO of Synthflow AI, the Berlin-based voice AI platform now powering human-like phone conversations for more than a thousand customers, we didn’t talk about hype.
We talked about the unglamorous foundations that finally make voice AI work: sub-second latency, deterministic guardrails, memory across calls, HIPAA/GDPR compliance, and infrastructure that doesn’t cry for help at a million concurrent sessions.
This episode — and this article — decode the moment voice AI became enterprise infrastructure rather than a demo with good lighting.
Our Sponsor
Quick break for something every founder should hear. One leak on the dark web can mean account takeovers, impersonation, or a board-level crisis. That’s why we partnered with NordStellar — a business-grade threat-exposure platform from the team behind NordVPN. It gives you early signals before attackers escalate — with data-breach and dark-web monitoring, attack-surface discovery, and cybersquatting detection. You’ll spot exposed credentials, shadow IT, and fake domains fast.
Startuprad listeners get an exclusive 20% Black Friday discount — go to nordstellar.com/startupradio and use code blackfriday20 before December 10, 2025. Don’t wait until your data shows up for sale — visit nordstellar.com/startupradio, code blackfriday20.
Table of Contents
The End of IVR Trees
Why Memory Became the Missing Layer
The Reality of Contact Centers Before AI
The Infrastructure Behind Modern Voice AI
Synthflow’s Memory Framework (VAL + BELL)
Deployment Blueprint: Pilot → Production in 30 Days
KPIs That Actually Matter
Why This Technology Finally Works in Regulated Industries
What Breaks at a Million Calls — and How to Avoid It
The Future of Voice-to-Voice LLMs
Advice for Founders and CX Leaders
FAQ (12–15 questions)
The End of IVR Trees
If you ever called an insurance hotline in Germany in the early 2010s, you know the drill:
“Press 1 for service. Press 2 for something unclear. Press 3 for an existential crisis.”
And if you dared to say “YES” to stay in line, the system replied:“I understood no. Goodbye.”
IVR wasn’t built for humans. It was built to cut cost until humans stopped complaining.
Voice AI changes that — not because it sounds pretty, but because it can finally follow context, understand your intent, and respond in real time without freezing for two seconds like a junior employee staring at a spreadsheet for the first time.
This is the shift Synthflow is designed for:From menus → to memory.
Why Memory Became the Missing Layer
Early LLM voice wrappers failed for one simple reason: no memory.
A caller could say:“My name is Anna from Berlin, we spoke last week about my claim.”
The AI forgot everything the moment the call ended.
Synthflow’s new memory layer — built into its VAL framework (Voice-Agent-Learning) — fixes this by letting agents:
Remember previous sessions
Retain customer preferences
Recall unresolved cases
Pass shared state across multiple agents
Suddenly, voice AI isn’t a parrot with good diction. It’s infrastructure.
The Reality of Contact Centers Before AI
Before AI, contact centers battled:
Repetition (customers explaining the same story)
Rising AHT (average handling time)
Low FCR (first call resolution)
Dropped calls
Poor agent experience
And let’s be honest: unreliable systems that looked like a 1997 Windows NT server caught in a rainstorm
Hakob explains a truth operators know well:Humans tolerate bad chatbots. But they do not tolerate broken calls.Voice failures trigger panic inside organizations.
This is why Synthflow invested heavily in:
Infrastructure teams
Sub-second latency pipelines
Reliable streaming STT → LLM → TTS
Telephony-grade uptime
Deterministic guardrails
Simulation environments
And it’s why memory matters: without context continuity, even the best AI voice falls apart.
The Infrastructure Behind Modern Voice AI
Hakob describes building voice AI as operating an airport:
“If you do your job well, nobody talks about it. If you fail once, everything stops.”
To reach sub-second latency across millions of calls, Synthflow relies on:
Highly optimized orchestration pipelines
Vendor-tuned STT/TTS layers
Route-aware telephony
Low-latency architecture
Real-time interruption handling
International failover strategies
Massive hosting capacity expansion
A dedicated reliability engineering team
This is the part nobody sees when playing with a Vonage demo on a Tuesday afternoon.
At scale, it becomes engineering mountaineering.
Synthflow’s Memory Framework (VAL + BELL)
Two internal frameworks run the show:
VAL — The Voice Agent Layer
Memory + actions + orchestration + integrations.
BELL — Build · Evaluate · Launch · Learn
This is where enterprise risk meets AI determinism:
Build: visual graph builder for deterministic flows
Evaluate: simulate thousands of calls before going live
Launch: versioning + A/B testing
Learn: analytics + auto-QA + improvement loops
The BELL framework replaces the old method of “launch and pray.”
Hakob puts it elegantly:
“You don’t want an intern freestyling on your most sensitive processes.”
So you give AI the same guardrails.
Deployment Blueprint: Pilot → Production in 30 Days
Synthflow’s rollout model is surprisingly structured:
Week 1 – Use Case Definition + Telephony Mapping
Agree on KPIs
Integrate SIP trunks
Map workflows to CRM actions
Define memory policies (read/write/consent)
Week 2 – Build + Simulate
Use Flow Builder for deterministic paths
Add memory
Add CRM/Webhooks
Run 100–1,000 synthetic test calls
Week 3 – Limited Live Rollout
Blue/green traffic split
Validate containment
Measure AHT/FCR shifts
Tune interruption handling
Week 4 – Full Deployment
Internal training
Monitoring dashboards
Weekly refinement loops
Done right, this becomes the fastest ROI upgrade in BPO/CX operations today.
KPIs That Actually Matter
Contact centers love dashboards. The problem is: most KPIs lie.
The ones that matter:
1. AHT (Average Handling Time)
Memory cuts time because callers stop repeating information.
2. FCR (First Call Resolution)
When AI has context, tickets close faster.
3. Containment Rate
The percentage of calls handled without human intervention.
4. CSAT
Even small improvements shift retention and cost-to-serve.
5. ROI (Return on Effort)
As Hakob puts it:
“If AI does even 2–3 percentage points better than humans, it’s already a win.”
Why This Technology Finally Works in Regulated Industries
Synthflow is compliant with:
HIPAA
GDPR
SOC 2 Type II
This includes:
Data locality
Non-recording modes
BAAs
Encrypted memory storage
Consent flows
Access logging
This is why healthcare and finance are now large adopters.
What used to be “impossible due to compliance” is now “mandatory due to cost.”
What Breaks at a Million Calls — and How to Avoid It
Hakob is brutally honest:
Latency spikes
Telephony instability
Rate limits
Vendor failures
Bad agent configurations
Memory conflicts
Logging gaps
Infrastructure overload
Human panic
The antidote?
1. Reliability engineering culture
“Print the reliability goal on the wall.”
2. Architecture built for black swan load
Plan for peak season on day one.
3. Deterministic flows for sensitive tasks
No hallucinations in banking or healthcare.
4. Multi-region everything
If one fails, another instant takes over.
This is how voice AI moves from “nice experiment” to “trusted system.”
The Future of Voice-to-Voice LLMs
Hakob is cautious but optimistic:
Near-term
Better reliability
More deterministic guardrails
Richer memory
Voice-to-voice inference
Longer context windows
On-device processing for enterprises
Mid-term
One-hour conversations that stay coherent
Multi-agent voice teams
End-to-end workflow automation
Domain-specific long-context LLMs
What he rejects
“AI replacing humans entirely.”The future is hybrid — with AI taking the repetitive load.
Advice for Founders and CX Leaders
Hakob’s advice is surprisingly human:
In the end, AI doesn’t reduce complexity — it raises the stakes.
FAQ
1. What is memory-enabled voice AI?
It’s voice AI that can remember customer context across calls and use it to resolve issues faster.
2. How does Synthflow’s memory work?
Through persistent storage with consent, strict governance, and deterministic workflows.
3. Does memory improve FCR?
Yes — significantly. Customers stop repeating themselves.
4. Can voice AI replace IVR?
In many cases, yes. It offers natural, fluid routing.
5. Is Synthflow secure?
Fully compliant: GDPR, SOC2 Type II, HIPAA.
6. What industries use this?
Healthcare, finance, retail, telecom, BPOs.
7. How fast is the AI response time?
Typically sub-second (≈400 ms).
8. Does it integrate with CRMs?
Yes — Salesforce, HubSpot, custom systems.
9. Can it handle peak season?
Yes, with multi-region scaling.
10. What about hallucinations?
Deterministic flows eliminate them.
11. How long is deployment?
Pilot → Production in 30 days.
12. Does it reduce cost?
Yes — through containment and FCR improvements.
13. Can humans escalate calls?
Instantly — with full context.
14. How does it handle compliance?
Selective recording + encrypted memory + data locality.
15. What’s next for this tech?
Voice-to-voice LLMs with long-context reasoning.
Internal & External Linking
Internal
Authority Sources
Give us Feedback!
Let us know who you are and what you do. Give us feedback on what we do and what we could do better. Happy to hear from each and every one of you guys out there!
The Video Podcast Will Go Live on Saturday December 6th 2025
The video is available up to 24 hours before to our channel members in what we call the Entrepreneur’s Vault.
The Audio Podcast Will Go Live Saturday December 6th 2025
You can subscribe to our podcasts here. Find our podcast on your favorite podcasting app or platform. Here are some of the links to subscribe.
The audio and video is available up to 24 hours before to our substack members in what we call the Entrepreneur’s Vault.
Tune in to our Internet Radio Station here:
Be one of the people smartening up with our content, as well as that of many media partners, including but not limited to Tech.eu and Stanford University Radio Show Laptop Radio
Get Our Content to Your Inbox
Decide what you want to read and when. Subscribe to our monthly newsletter here: https://startupradio.substack.com/
Find All Other Channels Here
Find all options to subscribe to our newsletter, podcast, YouTube channel or listen to our internet radio station here: https://linktr.ee/startupradio
Explore More from Startuprad.io
Want direct access to every founder insight, metric breakdown, and lesson we’ve ever published — without scrolling through episodes?
Try our Startuprad.io AI Chatbot, available on our website, where you can query insights from hundreds of interviews in seconds.
And if you want the unfiltered founder stories we can’t publish anywhere else, join our premium community:
Substack (Founders Vault): https://startupradio.substack.com/subscribe
YouTube Membership: https://www.youtube.com/startupradio/join
Unlock deep-dive conversations, private Q&As, and exclusive insights straight from Europe’s top founders and investors.
Interested in Working with us?
Welcome to Startuprad.io™, your premier source for news and insights into the vibrant world of German, Austrian, and Swiss tech innovation! As a decision-maker, you’re invited to dive into our content, featuring Emmy award winners, New York Times bestsellers, and Forbes top lists luminaries. Our platform offers a curated selection of interviews with industry leaders, game-changing entrepreneurs, and influential investors. Our focus on startups around Series A funding ensures that you get exclusive insights into the rising stars of tomorrow. By advertising with us, you tap into an audience that values forward-thinking and impactful partnerships. Join us at Startuprad.io and let your message reach the innovators shaping the future: partnerships@startuprad.io
Subscribe & Follow
Don’t miss out on our latest episodes and updates! Follow us on social media and subscribe to our newsletter for more exclusive content: https://linktr.ee/startupradio
The Host & Guest
The host in this interview is Jörn “Joe” Menninger, startup scout, founder, and host of Startuprad.io. And guest is Stephan Schubert, Investor & Managing Director at STS Ventures.
📝 About the Author
Jörn “Joe” Menninger is the founder and host of Startuprad.io — one of Europe’s top startup podcasts. Joe's work is featured in Forbes, Tech.eu, and more. He brings 15+ years of expertise in consulting, strategy, and startup scouting.
Automated Transcript
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:00:00]:
If you run a contact center or business process outsourcing, you know the pain. Customers repeat themselves, average handling time aht creeps up and your AI can't remember anything. Today's guest hey Cop Astabastian, CEO of synthflow, the Berlin voice AI platform backed by Excel, just launched Memory for voice agents so AI can keep contacts across calls, add sub second latency inside your existing CRM telephony stack. In the next few minutes you'll unpack how memory turns conversational AI into enterprise infrastructure. So use costs, lift FCR and stop making customers repeat their story.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:00:57]:
Welcome to Startup Rad or IO, your podcast and YouTube blog covering the German startup scene with news interviews and live events.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:01:11]:
Our guest today is Hakob Astabatsyan, co founder and CEO of Synthflow, the enterprise voice AI platform out of Berlin that's scaling like few others. The company powers human like phone conversation with for 1000 plus customers, serves 100 enterprises, integrates with 200 tools from telephony to CRM and runs at sub second latency with 99.9% availability. In 2025, Synthfloww raised US$20 million Series A LED by Excel, cementing its position among the world's top AI agent platforms, recognized by G2 for customer satisfaction, fast implementation and best ROE return on investment. Today we dive into Synthflow's newest capability memory. The VAL framework lets voice agents remember details, preferences and unresolved cases across sessions, support seamless human handoff and even share state via memory groups across multiple agents. If you tried AI in your call center and hit the context wall, this episode is for you. Hey Cop will take us from pilots to global rollout. How to deploy, measure and govern memory enabled voice AI without adding risk.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:02:37]:
Hey Cobb, welcome to Startup radio.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:02:40]:
Hey John, thanks for having me here. Glad to be here.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:02:45]:
Before we get started, for a little bit older people, I was wondering, you remember Space Odyssey where the AI hell goes crazy and tries to kill all the people? Did you have at any moment the intention also to have a hellboat in your voice agents?
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:03:08]:
Yeah, it's a good question actually. When we started building this in the early 2023, everyone was talking about AGI, right? The first version of GPT came and everyone was saying oh now you're just going to click one button. AI is going to do everything. And then everyone, you know there was like scandals everywhere, like people were warning against AI AI. But I think in the last two years everyone came to this realization, oh actually the LLMs are not that smart as we thought, right? I mean it's an incredible progress what we had, right? And, but, but I think everyone realizes there is also as much as this technology can do and we are actually very far away from that AGI. So I think the expectations are very way lowered right now. And we're entering into this phase of soberness right after the, after the, after this excitement and euphoria, I think people are becoming sober in terms of being realistic. What can actually this AI do? And the same applies to the voice AI, right?
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:04:17]:
I also realized that inside fiction there have been always the bad, the mean AIs think about Skynet, but I do believe simply for the reason they make way better stories, different topics. So let's take us to the start. Like not necessarily when you founded the company, but the original Spark. And when did, like what was the moment, you may remember when you conceived the original Spark for Synthfloww and when did it click that voice AI agents needed to be infrastructure, not just smart stuff?
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:04:59]:
Yeah, very good question. Actually.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:05:03]:
We started experimenting with LLMs around like March 2023. This is when the first, I would say proper version of GPT was released with an API that you could start working with. And everyone was building like text bots on this LLM, basically this wrapper around LLM, right? And then you would have a text bot. And for us the first idea was that we wanted to build a no code layer so that business folks could interact with it, right? Because. Because this was like a lot of developers and in developer forums everyone was trying to work with the API, but we also wanted to have this no code layer. And we built the first version for Chatbot. And I think it was around summer of 2023, like few months after the release of GPT that the first ideas of this voice orchestration started resurfacing on Internet, right? And the orchestration idea is actually very simple. You take the speech of a human, you convert it to text, you do some work there and then you send it to LLM.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:06:10]:
This text LLM gives you an answer back in a text form. Then you take this text, do a couple of things and then convert it back to speech. You do it and you have to do it very fast so that you play back the answer to the human. That's where latency comes into game. This was the first orchestration. Everyone, not everyone, but I would say a few teams actually, because it's very hard thing to do. Text is easy, but. But voice was very hard.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:06:37]:
And we started building this orchestration. And I remember around end of the summer of 2023, we had the first version that you could talk on the phone and it was terrible. If I look back, it was terrible. The first version and it would wait for two seconds to answer. For example, right now when we Talk, we have 400 milliseconds of latency humans, we are perfect because the moment you interrupt me, I immediately process it and stop and then vice versa. That's why we can have such a fluent conversation together. But it was not the case for AI, right? It would pose for a long time and then when you would start speaking as a human, it would speak over you. Right.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:07:16]:
At some point you just hang up as a human. Right. This was the very first version, but it was an interesting revelation for us because we realized this is a very hard problem to solve and it's really worth it investing our time because the demand was very high for this. And then we started working on this and we released in early 2024. So we worked six months on this. We released the first version, which is way worse than what it is right now. Right. But it was already pretty good.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:07:47]:
And this is where we actually started commercializing. We had our first customers and we started iterating with them. And since then, basically it, the technology has been progressing very fast. Right. New LLMs have been coming out and lot of infrastructure around this orchestration has developed like Deepgram, you know, speech to text and all this kind of solutions. And at some point, I think in 2025, early 2025, I think the entire voice industry crossed the chasm of actually voice. Because if you asked me in the early days, I was still not sure if this is actually going to be a technology that humans will be actually adopting on mass on scale. But this year we passed that chasm fortunately and it's just being now deployed in production like all around the world.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:08:41]:
Yeah, I also experienced that from a very different perspective. For example, I'm member of a few platforms where you can book voiceover talent. And I've realized that the number of requests are going way down and a few still pop up where you basically hand over your voice until eternity to an AI, which is not necessarily attractive for people who make a living. From that. Can you paint the reality of contact centers, business process, outsourcers, pre memory, what do aht? So the average handling time or the first contact resolution, also called FRC or the CSAT customer satisfaction course and call deflection, typically stall without the persistent contact. So.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:09:39]:
How did it work before your memory introduction and to all of our audience, yes, we have tlas Three letter acronyms.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:09:51]:
Yeah, exactly. Yeah, it's a, it's a specific industry with specific acronyms. Look, so I think before, before LLMs, like the way all these conversations were handled is through IVR trees, right? So you would call like, I don't know, I used to experience that a lot in Germany. I would call like insurance company or whatever and they would say, if you want to ask question about your financial thing, press one. If you want to talk to customer support, press two. Right? So you would have to go and then you would go into wait line and wait forever. So it was actually a terrible experience for humans, but sometimes you just didn't have a choice, right?
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:10:35]:
And I vividly remember they gave you, for example, when you talk to an insurance hotline, in the beginning, they gave you options where you had totally no clue where you fit in and you had to try through all the menus. Yeah, was really bad experience. And then it asked you, would you still like to stay online? And you yelled, yes. And it said, I understood. No, I'll hang up now. There was level of experience, right?
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:11:01]:
Yes, exactly. And it was like a lot of NLP NLU technology involved there, sometimes capturing information. I mean, it was a very basic technology, right? And the human experience, it was not built for human experience. It was built for just doesn't matter. Your experience doesn't matter. But at least we will get something solved rather than never solved. So it was, the mantra was better late than never, sort of.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:11:32]:
But it was the state of the technology.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:11:35]:
And now.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:11:37]:
With this conversational AI, the thing is that now it's all about experience as well, right? Results, but also experience, right? It's not like, oh, I just reduced my deflection rate or I just increased my resolution time, but what you just mentioned, csat, right? The satisfaction score, this is very, very important, right? So when, when we started building Synthfloww, actually for us in the very, in the beginning it was all about customer experience. So this is the only thing we cared about, right? Like how is your experience with AI, right? Are you, are you having a good con? And that's why, you know, the entire industry tries to build this empathy into AI, right? Better voices. So the early feedback was many voices are robotic, right. And then you try to build more empathetic voice. Right? And all these things is all about human experience. Ultimately it's all about humans here, right? And then the second stage of the technology, when we started working with contact centers, of course they have to have clear, as you mentioned earlier, roe, return on investment. And this is where is the second part of the conversation? Right? When you have a conversation with voice AI, the first part is the experience, the voice experience. The second part is the rpa, right? This business process automation, like what, what is the work that AI does, right? Does it, does it answer your question? Does it route you to a human? Does it send you an SMS with some information? Whatever the job of the AI is, it has to be done right because only that way the contact centers can measure the success.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:13:18]:
And this is what, what we measure at the sink flow. For us, the most important KPI for our customers is that, that they get their results, roe but also the end, end customer, right? The person on the phone has to have an amazing great experience because the entire purpose of this technology is in vain if people don't have a great experience with this.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:13:41]:
Totally. Yeah, exactly.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:13:47]:
I was wondering what feedback, what signals.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:13:51]:
Or.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:13:53]:
Any other input did you receive from your customers that you had to prioritize this multi turn memory over adding yet another feature? Because looking back now, it does make sense. But if you do have to make the decision, let's do A, B, C, D, E. How did you decide? And decide based on what?
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:14:24]:
So the way we build our product is always based on our customer feedback, right? We talk to our customers and we try to understand what is important for them. That's one. And second, we always try to understand is it feasible, right? Because as I mentioned earlier, right? When this technology came, everyone was expecting AGI, right? Like general intelligence and you just click one button and everything gets done. And then I think suddenly everyone woke up to the reality that the context is everything. What is the context, right, of the agent take like where is the data? What does it have to do? And this is for our customers, right? Being like in a B2B context, right? Basically you can think of Synthlow like an AI contact center, right? And the whole purpose of doing a work end to end, right? AI has to do a work end to end. What I mean by that? Think about if someone calls and wants to get.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:15:29]:
Some number sent, right? Or some information sent per sms, right? If the AI has a good conversation but doesn't send that SMS to the customer, then the job is not done, right? The action is not executed. So this is where we started working, right? And you need like memory and you need, you need context, right? For AI to be able to do the job and for large companies, enterprises, very often it needs that Synthfloww agents have to be integrated into their existing tech stack ecosystem, right? Maybe they use one telephony system Right. They have their own CRM and they have whatever they have that agent takes inflow agents have to interact with. Right. So that they can actually resolve customers problem end to end was. Was the key. Right. Otherwise you don't have that ROE we just mentioned.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:16:24]:
And this was the reason.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:16:25]:
They're just an interaction layer. They make the data from your customer, meaning their tech stack, their problems, their interaction. They make it feasible. They just a layer of interacting instead of sitting there trying to navigate through the systems.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:16:42]:
Exactly. But it also does, it also navigates. Right. So you can think of it like. Basically it's like a knowledge base. Right. Like existing knowledge base. Like you connect it to the agents and when the agent interacts with the customer, it has the information that it needs to answer customers questions or send information or even capture.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:17:04]:
Sometimes it's the other way around. Right. The agent captures an information from customer and sends it to the CRM. Right. To the existing CRM and updates, let's say the database with the data. Right. So it can go both way. Either it shares information with the customer or it gets the information from the customer and saves in the existing database.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:17:26]:
Right. And without the context. Right. All this ecosystem, the agent is basically it's a demo. Right. It will not be able to do an end to end work that humans do nowadays. Yeah.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:17:38]:
We, we talked about the latency that it's below a second sub. Second latency and like having really persistent sessions. Not that it's interrupted and it's, it's not useful anymore. So how, what was the.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:17:59]:
What was the hardest challenge there from an engineering standpoint to get, to get this workable?
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:18:06]:
Yeah, it's a very good. So actually latency was the number one problem when Fox started building voice AI because by following that orchestration, right. Like the speech to text and text to speech, right. You had to do.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:18:25]:
Was very rudimentary. The first version that was built, it was very basic. Right.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:18:33]:
And also like everyone was building, it was like almost for demos. So you were just trying to optimize it for one call, etc. But the problem became really complicated. Right. When we started growing very fast. We started growing very fast. And imagine like processing millions of calls every month or hundreds of thousands, like tens of thousands per day and all these calls together. You have to make sure that your entire infrastructure.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:19:00]:
Right. Can support that. And there are also so many factors that can affect latency. Right. It could be related to telephony, it could be related to connectivity, it could be related to even how you set up the agent. Right. And it was basically. But we realized that this is the number one challenge to solve without this, like nothing really matters in this business business.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:19:27]:
And what we did is we hired engineers who are, who have been specialized in this area and we have a team in Synthfloww that's called like it's basically our infrastructure team that only works on these topics, right? They work on reducing the latency, they work on improving the interruptions and then some point. It's basically two components that affect this one. It's an empirical work, right? You constantly collect data, you constantly improve your system and infrastructure. And the second is like how you build your architecture, right? Which vendors do you use, which LLM do you use? Right? How do you send the data between A and B? So there's a lot of architectural decisions you have to make so that when you start scaling this, you don't. It doesn't break. Because everyone that works in voice AI knows how hard it is to ensure reliability when you process millions of calls. I always bring this example, I say it's. It's similar or analogical anecdotal example could be for folks who work at the airport and have to ensure that the planes land safely, right.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:20:38]:
One coordinators. And of course if you do this job well, no one talks about this, but if you do one mistake, it's a huge disaster, right? Everyone will. And the stakes are very high in our business as well, right? You can process million calls and it's expected. But let's say 100 calls break, right? Something happens suddenly people or contact centers panic, right? Because I don't know, there's something about calls that people get really angry when these things don't go through, right? It's not the case with text bots. Sometimes people get frustrated and close it and go. But there's something about calls that people make huge buzz around this. They close down operations and this. And that's why reliability is one is basically we print it out and put it on the wall at the office.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:21:32]:
For our engineering team in the early days, this is the most important thing we have to solve in the very beginning. Yeah.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:21:41]:
Did you ever talk to psychologists why it is that voice and calling is so special here?
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:21:49]:
I haven't talked to, but I have read about similar kind of concept. It's like it's basically psychologists call it anthropomorphism, which means that humans basically like humans have this tendency to attribute human qualities to machines or non. Non living things, right? Also you see it like in children, books, etc, when animals, animals talk, robots talk. Robots became humans friends and there is an empathy, right? So there is definitely an anthropomorphism, which means humans in the beginning have this tendency to be creeped out when the technology machine behaves like human. It scares humans. Right? So there is a, There is an adoption. There is an adoption window for humans actually to get used to it. Right? So it has been with many technologies.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:22:41]:
Like for us, everything we use is normal, right? The phones and laptops and everything. But the folks that saw it for the first time in their lives, there is always some, I would say.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:22:55]:
Adoption time needed to get used to this technology. It was same thing with voice AI, right? People thought, this is dangerous.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:23:04]:
I don't know, it's weird that the AI speaks like human, right? But that's why I was saying this year we crossed the chasm, because right now there's a large amount of humans that are actually used to it or already had their first experience talking to AI. And there were so many. Like, I always bring this example of this movie called her, right? When, When Joaquin Fenix is talking to the, To, To. To the AI, right? All the times, basically, it's like a human. And, and I think, I think humans have also been prepared for this a little bit through, through popular media, et cetera. But it's still like in the early days, I remember it was like it was very hard to make certain type of calls.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:23:48]:
Yeah, those, those people who warn about something. I vividly remember that I once read there were doctors recommending not to travel by train because they thought it would be harmful to the human body to move faster than a horse ride.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:24:06]:
Yes, yes, exactly.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:24:09]:
Yeah, so we, we've seen that. Let's be a tiny bit geeky here. Talk about here. What did you try that didn't work? When building memory prompt only hacks external key value stores, native retrieval, augmented generation, and what replaced it.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:24:31]:
So.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:24:33]:
Look, so the very first I would say approach was put, put everything. So if we go back to the very beginning, two chatbots, right? So it was the, this was the rug, right? Like you connect a knowledge database, right? But in, in, in some situations, you know, in some context, I would say it's for, for, for. Especially in the Enterprise situations, each situation was different, right? Each situation was different, each document was different that you had to reference, right? So there was some complexity involved to make it work for each, especially for the Enterprise. And then I think there was a moment where we solved this by having the prompt, right? You have the prompt field and everything that agent needs to know is within the prompt, the open prompt way, right? And you just paste it there and the information is there. But of course, there is also a limit there, how much you can put. Because if you put too much, right, if the text is huge, that also increases the probability of hallucination, right? Basically, think about it. The more the AI, the more information you give to AI, there are more paths that the AI can go in the conversation, right? So it increases basically the probability that it can deviate or it can get confused from the main thing that you set the AI to do, right? And I would say honestly, the answer is it depends because right now we have that like the open prompt thing, but we also just released the Bell framework and one of that Bell framework, the first one called Build, is the Flow designer. So the Flow designer allows you like, it's a visual graph that you can build the conversational paths, right? And with that we want to address the cases where you really know, you really know what the AI needs to do, right? Imagine you just hired an intern, right? You don't want that intern to go freestyle, right? You know your business, you just sit down and explain one hour and say to the intern, look, this is the script, this is what you say in the call, right? And this is what you don't say in the call, right? And I want you exactly to follow this, right? And this is like.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:26:52]:
And you can create it. The benefit of this, it's very deterministic, it reduces the hallucinations, it puts guardrails in the place. And basically AI does the job that it has to do if.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:27:07]:
For the open prompt version, right? When you have, you basically just paste the same way you prompt GPT, right? You don't build the graphs, but you just prompt it for. In that this is better if you want to keep it more open ended the conversation. If you want to keep the conversation more open ended, right? You can give it some instructions. But you also hope that AI would freestyle a little bit, right? The same way you spend some time with GPT and you just wanted to give also creative answers, right? So I think having that, that's why we've built both, right? So that our customers can kind of have more deterministic for more specific cases, especially in B2B context. Determinism is actually very much appreciated nowadays. And of course you can also leave it a bit more open ended if your use case is such. You don't to want it to do exactly that way. And, and this is like the spectrum that we leave actually our customers to choose how they Kind of create that, that instructions and memory.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:28:08]:
Yeah.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:28:09]:
You just talked about leaving your customers choices. Where do they get burned by a or did get burned by AI voice, promising promises, God rays, evaluation, handoffs or security. And how do you de risk them?
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:28:30]:
Yes, that's an excellent question. Right. It's basically for enterprises and contact centers in B2B context. Right. Where ROE is everything. Risk has been the biggest challenge. And basically think about early days of agentic AI is you build the agent and you hope it will work. You hope it will do what you want it to do.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:28:55]:
Right. It's, it's a hope. You just launch it and then suddenly you come and see, didn't do what you wanted to do and then you iterate with it. Right. It's very similar. When you chat with GPT, you ask a question, it doesn't answer and you go back and say dear chat GPT. My question was this and this, please be very right. You just, you, you push back a couple of times until you get exactly what you need to do.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:29:17]:
And it was very similar. Right. And we said look, we have to solve this.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:29:26]:
We can't leave it to chance these things. And that's where we started working on Bell Framework, which is like build, evaluate, launch and learn. The entire idea of that framework is we wanted to derisk enterprise deployments. Right. It means you can build more predictable conversational paths with this designer. For evaluate part we build simulations. It means you don't have to go and call real people to see if AI does what you want to do. You can do it with synthetic data.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:30:02]:
You can do a simulated in the same environment before you actually release it. So it's way faster and way safer. Then there is the versioning part. Maybe you want to have a couple of versions that you want to test and lastly you want to do some auto QA and analytics. So basically we created an environment within Synthflow for customers, enterprises where they can go and they can do all the work there in a very safe environment, make sure that the agent does what they actually intended to do and only then they release it to the public. And we have been working for this for many, many months and we're releasing right now. And that's the entire purpose, risking and accelerating actually the deployment of these agents. Yeah.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:30:54]:
We also going to record a Founders Vault episode after this one and they'll tell us about the toughest enterprise deployment and where you failed initially, what broke in production and who escalated and who you salvaged it.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:31:12]:
Let's talk about your current playbook and how to move from a pilot to production in weeks, team setup, no code versus API. And the first three workflows you always target.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:31:28]:
Yes. So look, basically the older technology required a lot of human, human touch, right? And like especially the NLP NLU technology, people would have to go understand, map out the conversation and then go and build it, right? With Synthfloww approach, we build this layer, no code layer that allows customers to basically follow like three, four steps and you actually have your agent, right? So integrations is a bit different thing. This is where our forward deployed engineers work with the customer to integrate, especially on the telephony side and other tech stack like CRMs, et cetera. We integrate with an API or with SIP trunking basically. But once that work done, and this can be done very quickly basically within few days or weeks, depending on the case. But the agent creation part is very straightforward. So we have reduced it to an absolute, absolute minimum and basics, right? You have to prompt or with a visual graph builder, you just have to build a conversation and then you have to choose an action that AI has to do. You can also build a custom action if you don't have that.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:32:50]:
And then of course you can always leverage the Bell framework for that to test everything, etc. Before you deploy and then you can actually go live. So we have simplified it.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:33:03]:
I'll say to the limit, right? And so it's very easy actually to do and we continuously building new things around that so that we take out this, I would say this complex part from agents, right? So that it becomes almost like a deterministic thing you do, you just go, you build it and it works, right? That has been the, the whole idea. And in terms of use cases, the most common ones or workflows.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:33:36]:
That we see across the board, we see a lot of transfer case where AI is the first line of defense and then it weeds out the irrelevant calls and forwards the important ones, which actually reduces the workload. With humans we see a lot of customer support, or in other words, I love to call it faq, right? AI answering a lot of questions, right? For customers, which are very basic, so humans don't have to waste time on that. AI can answer all this, it's very solid. There is also a lot of qualification is happening. AI reaching out to humans, asking couple of things and qualifying and booking next steps. Basically these are very common, widespread cases we see in our, right. B2B AI contact center case, mainly on the phone, right?
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:34:29]:
Yeah.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:34:31]:
Talked about those KPIs. Can you share some specific numbers? Typical containment, the Reduction in average handling time, the lift in first call resolutions, meaning the customers only need to call once or human hour savings that you comfortable stating by vertical.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:34:56]:
Look, there are no single numbers, right? It very much depends on the customer and the use case, right? So for example, we have how complicated the case is, right? For example, we have one customer in healthcare and they basically benchmark AI with the work that human does today, right? In terms of deflection rate, right? So, and I don't remember what the exact number was, let's say 17 or 18%, whatever the number is, basically that customer says if AI even like 1, 2, 3 percentage points better than this, that's already a big win, right? Because humans, humans time is more valuable than AI's time, right? AI is 24, 7, you know, it's cheap, you can just put it there and it can do the work, right? The only reason you wouldn't put AI, if it's significantly worse in terms of ROE than humans, then you still justify. But if AI is even slightly better. So for example, in that particular case they were targeting 20% plus rate, then they went with AI and the human basically moves one step back and takes important calls, calls that are a bit more complicated, maybe they are more of an emotional nature or there is a complicated things that they have to do afterwards, right?
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:36:23]:
So that's what we're seeing, right? In other cases, for example, if someone does an outbound campaign and does like thousand calls, even if you get, I don't know, five, six calls answered, it could be a big win, depending what's your case, right? Maybe each call can result in a lead that can result in tens of thousands contracts for you afterwards, right? It's a big win. But maybe you call thousand times and everything less than 500 answers is already bad, right? And then in some cases AI works, in some cases it doesn't make sense, right? The roe, the simplest things to measure are more like, I would say the cases that I mentioned, right? Let's say the transfer and et cetera, these things are very easy to measure. And in some cases you see marginal improvement, right? AI is several percentage points better. And it's already a good case to put AI because of the cost savings. But sometimes we also have customers who never had anyone there, so they didn't even have resources. And just putting AI results in 80% improvement in time reduction, for example, right? Like some crazy numbers, because they never had that, right? They didn't have the resource and now they just put AI and it does something that was never being Done. And there you see like very like large numbers as an improvement.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:37:53]:
For our audience. If you run a CX or bpo, what's the one place you'd want voice AI memory like tomorrow, Bookings, claims, returns, tech support. Drop it in the comments. By the way, if there's anything interesting, Jacob, of course we'll share it with you. And let's talk a little bit about the framework. What's your memory read, write policy framework, what to store, for how long with what consent and what controls.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:38:28]:
This is a very important question, right?
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:38:31]:
We touched upon anecdotally before like why are phone calls so dramatic, right? Have such a dramatic impact on humans? And it's also same in terms of compliance, right? So for example, there is a HIPAA compliance in the US or GDPR compliance in Europe. And with a Synthfloww very early on in our journey, invested significant resources of our company actually to undergo all these compliances, right? SoC2, type 2, GDPR, HIPAA. And for example, with HIPAA compliance, right, there is the toggle bar in our tool that our customers disable so that the calls don't get recorded. Because certain types of calls in certain regions you're not allowed to record, right? So in gdpr, right, for example, the data hosting is a very important, important thing, right? And so calls are very sensitive and we always work very closely with our customers, right? So that we ensure we sign like baas with our customers, right? Business associate agreements we have basically right now on the Synthfloww website there is a trust vault that you can access and we have all sub processing agreements baas. We take this topic very seriously because in some industries, and I would say in many cases like the data storage and recordings is a very sensitive topic and it has to be 100% compliant. This is very important in our market.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:40:00]:
I was wondering for callers who get interrupted, switch or something, how do you maintain your context coherence and avoid hallucinated calls? For example, you call for your VW Golf and actually they tell you yes, you're bantling is ready.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:40:23]:
Yeah, that's a very good question, right? And this is comes back to the to the point, right? It's also like kind of the idea of Bell framework is we want to make sure, right, the customers can build more deterministic agents if its topic is sensitive, right? So for example, you can leave the AI agent a bit open if it has to answer like I don't know, some low stakes situations, let's say like hypothetically has to say what's the weather or something like that. Right. But if it's about the order or Golf and Bentley, right. You want to use the visual graph builder that we built because that way you completely eliminate the hallucinations. Right? Because that's what we've built. It's a very deterministic environment. Right. That doesn't allow AI to deviate from your.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:41:19]:
If you know, as a Volkswagen. Basically too that this is exactly what customer is going to ask and this is exactly what the AI should answer. This is the right way to go. Right. And then of course you just simulate this, you create this before actually releasing this. This is very important. Right. And we have invested lot of resources in building the Bell framework so that our customers, these enterprises can actually serve these use cases.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:41:48]:
And I can go one step deeper and tell you it's even more important in financial industry, for example, right. You cannot quote the wrong percentage or etc. Right. So you have to adhere to the things. And the short answer, basically you reduce the hallucinations through graph builder and of course the simulations and versioning in qa.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:42:11]:
That is great guys. Right after the ad break, hey cop will unpack the deployment blueprint. ThinFlow users to get memory enabled agents live in under 30 days plus one mistake that kills ROI.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:42:32]:
Guys, welcome back after our short ad break. Hey cop, can you walk us through through the 30 day deployment blueprint from data mapping and integration selection to evaluating metrics and blue green rollout.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:42:48]:
Definitely. So the way we work with our customers is in the very first step we try to understand the use case and the value that the customers seek to get out of this technology. Right. Because there is a lot of noise around AI and we want to make sure that our customers have the right expectations from this technology and we can actually deliver them in a promised time frame. Once we have done this part, we actually get into a.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:43:23]:
Period working period with customers where we help them with integrations. Right. We advise them on a lot of topics. Our forward deployed engineers stay very close to the customers and we advise them on a lot of things like for example on the agentic part we demo them the Bell framework so that they can deploy this in a de risked way. But also we help them integrating this into their existing ecosystem. Right. Especially many customers have their own telephony system and we have to integrate into that with the SIP trunking and there are a lot of nitty gritty details sometimes and our forward deployed engineers work with them to make that happen. And generally during this period we do these parts and the customers would go live and start testing.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:44:11]:
Right.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:44:13]:
And Then they would get their KPIs they want. Right? And this would be like a success. But right now with the Bell framework, that's not even needed to go to live. They can even do it like in a more controlled environment. And we even expect to reduce this more. Right.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:44:32]:
But that's generally like roughly how we work with our customers to make sure actually they get out of this technology what they want to.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:44:41]:
You've been talking about the nitty gritty details and that's usually where you win or lose in the startup game. Walk us through growth Methodology. Bell methodology. Bell framework. How do you scale from 1 use case to 5? What's the sequence that compounds return on investment?
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:45:00]:
Yes. So I think the very first step for us scaling the business, the first challenge we started facing once we scaled was the volume, the infrastructure. Right. Basically we were a startup of couple of folks and then basically in few months we started processing millions of calls on the platform. And it was very unexpected. So we had to do a lot of infrastructure investments and expansions to be actually able to process this. And the way we scale this is we generally are very selective with whom we work. Right.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:45:38]:
That's why we do this. Step one, what I mentioned earlier is we always try to understand the customer's expectations from this technology and kind of have the same expectations, right. So that we can actually help customer to get there. But the work on our side is more like consultative and being as a sparring partner for the customer. So it doesn't require large headcount on our side to work with our customers. And we have built the product in a very scalable way and we continue doing that. Right. We release the Bell, but then we also are cooking a couple of very advanced features which I'll not talk about today, but we are going to release soon in the future.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:46:25]:
And basically we'll scale it with the product. Right. So there is no like an inherent, I would say limit to why our platform at this moment. Right. Cannot process hundreds of millions calls instead of millions of calls. Right. And, and this is how we want to continue at this moment. Of course we have to grow our team.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:46:48]:
It's natural as a company as we grow and the revenue grows and the usage, we have to build different departments and be better, also offer better service to our customers.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:46:59]:
But there is no, I would say a fundamental blocker to scale this business for us at this moment.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:47:06]:
I've seen some logic sprawl in such projects. How do you recommend no code ownership vs ops vs API for engineering? And.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:47:21]:
Would governance Prevents this logic sprawl.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:47:27]:
Yeah. So I would say of course the API gives more flexibility. Right.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:47:35]:
And I would say for smaller cases, right. So we have to, I would say maybe you can say like two types of customers that one type would prefer the no code, the other type would prefer the API. So let's say one of our customers, a very large contact center bpo, they basically use the API because they integrate synth flow into their operations, basically into their entire tech stack so that they can offer voice AI to their customers. Right. So it's like a white label integration there. And of course the API is the way to go there because they want to expand on this massively. But if there is, let's say for smaller customers, right. If you have one particular use case in mind and you want to quickly build this, it's very easy to go and build with the no code layer you have there because it's just click, click, click right and you have the case.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:48:35]:
But we also have built some sophisticated features there. It's called like multi tenant, basically.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:48:45]:
Portal where you can become the owner and you can have many users invite your team members and have sub accounts. Right. For different use cases. And we have many customers who just went chose this way. Right. Because it's simpler, it requires less engineering resources and it's more than enough for their particular case. Yeah.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:49:09]:
You do have.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:49:13]:
Some case studies on your website, Med Bell, SmartCat, what ingredients made those outcomes work?
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:49:22]:
Yeah, so I think so. One important thing is that this initial period that we actually are close to our customers.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:49:37]:
We set their expectations right, we help them define the outcomes that they want to get. Very often customers know it themselves, but if they don't, they're more explorative. We help them understand that. And once it's done, then the rest is all about the right integrations and the right agentic part, the right agent setup. Right. So the agent setup we have simplified to minimum. But there still could be some questions. Do I go with the visual graph builder or do I prompt it myself? But of course our customer support is always there to answer customers questions.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:50:15]:
But the integrations part, this is something that often we sometimes have to share something like our documentation or etc so that they can integrate into their telephony system. Or if they don't have telephony system, they use us. Right.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:50:33]:
Or maybe they have another tool stack like a CRM Salesforce or something like that they need to integrate and these are the main components. If this is done well.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:50:46]:
It should start actually delivering the value. You can go to production.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:50:51]:
I was wondering a little bit why you talked what breaks at 1000 parallel calls, speech quality, latency spikes, rate limits, integration quotas. And how do you harden for peak season? I mean especially if you are into some kind of business that has a rush towards the end of a year towards Black Friday, e commerce Christmas is what I had in mind. How do you prepare for something like that?
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:51:24]:
That's a very, very good question. And this was honestly at the very beginning this was our main problem. This was like when we started scaling because and thousand calls are fine, it's nothing. We're talking about millions. This is okay, totally fine with me.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:51:43]:
Let's talk about millions.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:51:45]:
That's where real challenges start happening. And honestly one of the first challenges we had was the capacity. So we had to.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:52:00]:
Get more and more hosting capacity so that we could digest all this, all this processing, right? And so this is one thing you have to make sure that your infrastructure, right, the cloud and the hosting, you already have that capacity. It's set for scale, right? And for us, because we grew so fast, we were not very much prepared in the beginning. We never thought we would go in few months basically from hundreds to hundreds, thousands or meeting calls.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:52:33]:
And this was the very first challenge, like architectural challenge, right? So the number of calls per se didn't affect in our case like the latency directly. But if something else affected latency reversely that would mean that million calls were affected by that, right? So the stakes are becoming very, very high. So you cannot work operate as a startup anymore, right? By pushing Things and Version 1 you don't care if anything broken. So suddenly you have to graduate very, very quickly to actually having a high performing, high reliability like triple 9s 4nines at some.59 reliability company which is actually to be very frank, is a huge challenge because you just have an engineering team of like eight people and suddenly and now we are like 30 plus right? Engineering team and you have to professionalize like overnight. And that was like a very big challenge for us. And it's very hard. This is something, this is something.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:53:43]:
Even in the era of AI hasn't changed. It takes time to hire your team, it takes time to build your culture, it takes time to have quality people. And that's kind of was a big challenge to get there. So to summarize, basically one thing is you have to make sure that the architecture is there, but you also have to build a very world class team because your business processes, your engineering processes are going to affect if you push something to deployment and it affects even minimally one latency or et cetera. It means 4 million calls at the same time and then you will have too many fires you will not be able to extinguish. Right. So you have to build a very strong system to have to ensure that reliability.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:54:40]:
Getting professional, getting a big team very fast. Always a challenge. As you said, you can change it simply by applying AI. In the Founders Vault you'll share some internal evaluation dashboard used for memory co KPIs, guardrails and so on and so forth. Find the links down here in the show notes to subscribe to our premium offering the Founders Vault. Just five bucks a month.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:55:09]:
Let us go a little bit from what is now working into the future outlook. What's next for voice to voice LLMs reasoning and on device on edge. And where does memory evolve like agent teams, shared state auto summaries.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:55:31]:
So I would say like the near future in terms of like roadmap is clear.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:55:38]:
Longer than that. It's always hard in AI. It's always like basically I think humans have been historically terrible at predicting the future. Right.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:55:49]:
I usually say in such cases in our interviews, I know predictions are hard, especially concerning the future. Right. But, but, but you're here and we'll give it a shot.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:55:59]:
Yes, no, no, I'm happy to share actually. I am one of the guys who always has an opinion on these things. But humans are generally. And like for example, everyone 2023 was saying AGI et cetera and now it's like very, very different. So look, so that the near future, I think at least for our market, right. For our product, it's all about basically reliability, quality.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:56:28]:
Ensuring outcomes like deterministic outcomes. Right. So that reducing the, I would say the chance, right? Reducing the chance in the agentic space. Right. So that the ROE is evident. And there are many things, right? It's there. You build architectural stuff around your infrastructure structure, but you also build features. This is the Bell framework which we will announce very soon.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:56:55]:
And many similar features, right. You built now and.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:57:02]:
On the more I would say down the road in the future. I think I'm really curious kind of what's the next step from LLMs? Because basically, basically products like Synthfloww very much depend on that, right? Like what's next in that, what's next in speech to text. And there is also like the voice to voice model from OpenAI. Right. The reason we're not using that because it's like commercially not viable yet. It's very expensive and it also has more limitations when it comes to actually agent doing tasks. Right. The RPA part.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:57:41]:
So it will be I think there soon. So that's one of my predictions is that voice to voice and the voice to voice is amazing experience. Right. And you have many of the problems that you have with the current orchestration, you don't have it with there. So I'm pretty sure this will be also state of the art very soon.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:58:04]:
I think that's an interesting part. I think we'll see also like huge progress in terms of long format. So right now AI is very good for shorter conversations, right. And finishing something and doing a job like I don't know, one minute conversation, customer support. Right, but what about one hour conversations? Right.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:58:28]:
Because the longer these conversations go, the harder it actually for AI to stay on track. So I think that's an interesting to see like what's on the long format part and of course like the complexity of the tasks. The tasks that AI does nowadays are relatively simple or straightforward, right. Like talk, capture something, share something, update something somewhere. Right. But going forward maybe it can like do something end to end, right. Go and book and I don't know, follow and really complicated tasks. I would say this is what I would expect this to develop.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:59:10]:
However, it's not going to be an overnight transition. It's going to take few years at least.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:59:19]:
Not only that, people don't shy away from interacting when somebody calls, hey, this is a booking service and you call a restaurant.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:59:30]:
In 24 months.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [00:59:33]:
What percentage of customer calls will AI handle end to end in regulated industries? And what will still require human?
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [00:59:44]:
Look, I don't think AI is going to completely replace humans. I keep saying this everywhere, that AI is going to augment humans. I don't think humans will be out of the loop in the near future. That's not going to happen because I don't see humans trusting. Like I would say.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:00:09]:
Not only trusting but AI. It just doesn't have that capacity to do very complicated things. Right. And very long, I would say RPAs automations. Right. So.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:00:25]:
I think in the next five, so look for some cases it's going to be 100% to answer your question, it's going to be 100% of the calls. Like for example for transfers, routings, FAQ, first line of defense. I think it's going to be 100% is going to be done by AI.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:00:45]:
For some cases, more complicated ones. Right. Processing, et cetera, Maybe even on the governmental side, I think it's going to be to be way slower. Right. Like I don't Know, let's say taxes or this kind of areas. I don't see it like adopting very quickly. That said, regulated spaces. Also like we already see like crazy adoption in healthcare and financial institutions despite.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:01:09]:
You just need to be compliant, right? You have to have everything like you have to have HIPAA compliance, you have to have pen tests, right. You have to ensure there and, and, and you're signing SLAs and this kind of things with the customers, right? If the state of the technology is there that you can guarantee that, for example with visual graph builders, then.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:01:32]:
Basically if the AI works for 10% of the calls, there is no reason for them not to deploy it for 100% of the calls. And I think it's going to be very different case by case, like use case from use case. For some use cases going to be 100% for some probably very low. Not even.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:01:50]:
1%. Yeah.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [01:01:53]:
A little bit of contrarian view when we go close to the end. By the way, you're holding up pretty well and we're recording for no more than an hour now. I was wondering what's an overrated metric or myth in an AI contact center deployment that teams should ignore or that it's still missing the return on investment?
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:02:17]:
Yeah, it's a very good question. So look, I don't think that the classical ones like CSAT.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:02:28]:
Or like things like deflection rate or average resolution time or time to answer these things are actually very important, right? And they do matter, right.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:02:46]:
I think one thing that I always look at, but I don't care so in a sense is that is the subjective feedback that we get very often. For example, sometimes people say oh, it was too robotic.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:03:01]:
And of course at this stage it still could be some elements of that it's an AI. And also I tell to all my customers, customers, you know, like you have to disclose that it's an AI. It's not built to lie to people that it's human. That's a bad practice, right? So when humans talk with AI, they should know they are talking to AI and they should have the right expectations, right? If someone tells me, ah, it didn't feel like it was a bit robotic, I think that's fine. Of course, as long as you know it's an AI, as long as it does the job for you, right? You have a good experience, right? Better than the Ivr trees or etc of the past. We're making a huge progress as a humanity in that area. And this is like for me, I wouldn't care that much about that we're getting that subjective feedback that it's. Because first of all, it doesn't have to be exactly like humans.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:03:57]:
Right. As long as your experience is good, you know, you're talking to AI and your problem is solved on the phone. What's the problem? Right.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:04:08]:
So that's kind of what I look right now. And I'm very, very tolerant in the subjectivity towards AI.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:04:15]:
And I just think we just need more time as humans to actually get there.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:04:21]:
To say, okay.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:04:23]:
This is where we are. Right. Yeah.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [01:04:27]:
Interesting. Yeah. Bottom line, the slower developing part is not AI, it's the humans in the. Yeah. For our audience, I was wondering, should AI remember you across calls if it improves your service? Yes or no? Tell us why in the comments.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [01:04:47]:
An advice for founders and CX leaders.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [01:04:52]:
What's your three step guide here? And can you share some advice for our listeners who are mostly founders and executives from this plan?
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:05:04]:
Yes. So look.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:05:09]:
I would share some advice on a bit like higher level. Right. So I think one thing that.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:05:17]:
It'S very different right now in the AI era compared to before AI is that it's very stressful. I don't think that founders have been under such a pressure and stress ever, ever in the history of startups because things are moving basically on weekly basis. Like every week, every day something new is happening. Like I wake up in the morning, I look into my phone. Some, someone, something, someone new model is out there. Someone released that, someone released this. This is changing. OpenAI is doing that and does so I think like these skills, navigating skills.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:05:59]:
Right. Filtering out through the noise is a superpower at this stage and founders need to learn actually not to get too distracted and too stressed down by the noise. It's going to be there for a while. It's a new technology. We're experiencing an immense progress as a humanity. I think this is already way bigger than Internet at. Right. What's happening right now.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:06:27]:
This is, we have never seen this, this type of progress. Right. And, and on a weekly basis. So really understanding and focusing on value building business is, is, is harder right now because you. There are many shiny things, right. You always confront these decisions. Oh, I know. This new model came.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:06:47]:
Shall I use this and that or do I go this way? This way? What about this? Right? And, and I think being very close to the customers, very, very close to the customers and focusing on building long term value for them for your customers is the key.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:07:05]:
This is my view, this is what I will highly recommend based on my own experience in the last Three years, I would say. Yeah.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [01:07:15]:
Three more questions for you.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [01:07:18]:
Closing Bold Insights Finish this sentence. In three years, voice AI with memory will be the default interface for for.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:07:30]:
It will be default interface for contact centers, both external and internal. It means external BPOs, right? Contact center as a service and internal contact centers for insurance, financial institutions, companies. Because we already see the ROE and the case is very clear. It's just that it's just a matter of time until enterprises start deploying and adopting this in their companies.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [01:08:01]:
And two closing questions.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [01:08:06]:
Are you open to talk to new investors?
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:08:09]:
Yes, definitely. Always happy to.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:08:12]:
I think the best format is always I say it Coffee Chat.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [01:08:16]:
Yes, we link your LinkedIn profile down here in the show notes and I do believe every investor that has been sticking around for over an hour, they'll be interested to talk to you. And of course you're looking for hiring talented employees.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:08:31]:
Yes, always. Right. And we're hiring on a rolling basis so we are not a company that just fills in gaps. We always look for exceptional talent whether it's AI, engineering, voice AI or also commercial roles, sales, solutions, engineering, forward deployed engineers. We're always open on a rolling basis for exceptional talent and we guarantee you and promise you also an exceptional environment to work on cool things and grow as a professional.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [01:09:09]:
Awesome. Hey Cobb, with such a pleasure having you here for our premium subscribers. We'll be back shortly with the Founders Vault. Thank you, thank you.
Hakob Astabatsyan | Co-Founder & CEO at Synthflow AI [01:09:20]:
Thanks for having me.
Jörn "Joe" Menninnger | Founder, Editor in Chief | Startuprad.io [01:09:26]:
That's all folks. Find more news, streams, events and interviews@www.startuprat.IO. remember, sharing is caring.
📝 Copyright: All rights reserved — Startuprad.io™
Comments