Dec 13, 2023 · 31m · mad
AI vs. Call Centers: ASAPP CEO Gustavo Sapoznik on the Future of Customer Service
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of The MAD Podcast, ASAPP Founder and CEO Gustavo Sapoznik joins Matt Turck to discuss how generative AI and custom machine learning models are transforming the $600 billion contact center industry. He shares insights on ASAPP's hybrid AI product architecture, enterprise go-to-market execution, talent acquisition, and scaling strategies for AI startups.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 14.3% of the talking time here. How this is scored →
Matt as informed peer 3.7 Guest teaching 4.7 Guest disagreement 1.4 Matt pushing back 2.1
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 19:38 Calling Stealth Mode 'A Very, Very Stupid Idea'
Gustavo forcefully rejects the host's framing of their early stealth strategy, calling stealth mode a 'very, very stupid idea' for enterprise software companies.
Hardest push from Matt ▶ 3:25 Challenging the Assistance vs Automation DichotomyMatt directly challenges Gustavo on whether ASAPP's core design philosophy favors co-pilot assistance over complete automation.
Biggest teaching moment ▶ 14:55 The Economic and Latency Impossibility of General LLMs in Call CentersGustavo educates the host on practical call center limits, showing that using GPT-4 would cost $50 to $100 per call and add 40 seconds of unacceptable delay.
Matt holds his own ▶ 12:56 Mapping ASAPP's History to Deep Learning MilestonesMatt displays deep domain context by mapping ASAPP's founding to 2014 deep learning advances, Transformers, and the emergence of GPT-2.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| Video Title and Event Host Graphic Cards | 2 | 4 | 1 | 1 | Matt opens with foundational questions about ASAPP's origin and the call center market landscape. Gustavo outlines the $600B market size and why ASAPP targeted giant enterprise accounts from day one. | |
| Legacy Software Limitations and ASAPP's Hybrid Strategy | 3 | 5 | 2 | 2 | Matt asks if ASAPP operates on an agent co-pilot model rather than automation. Gustavo reframes the premise by explaining that ASAPP automates as much as possible while using agent augmentation to collect data for further automation. | |
| ASAPP AI Services: Auto-Compose, Transcribe, and Coaching | 3 | 5 | 1 | 2 | Matt prompts Gustavo to explain ASAPP's specific AI product modules. Gustavo details why standard commercial ASR fails in noisy contact centers and how real-time AI coaching replaces 1% manual QA sampling. | |
| Redefining Customer Service Economics with Generative AI | 3 | 5 | 1 | 2 | Matt asks how customer service value and ROI are measured across complex stakeholders. Gustavo breaks down the historical trade-off between cost cutting and customer delight, illustrating how AI eliminates that zero-sum compromise. | |
| Custom Model Training, Latency, and Agent Models | 6 | 5 | 1 | 3 | Matt shows strong technical literacy by tracing deep learning history from 2014 through Transformers and bringing up 'agent models'. Gustavo explains why GPT-4's high cost and 40-second latency necessitate custom, domain-specific models. | |
| Enterprise Go-To-Market Execution and AI Talent Recruitment | 4 | 5 | 3 | 3 | Matt brings up ASAPP's extended period in stealth mode, which Gustavo forcefully rejects as a 'very, very stupid idea'. Gustavo then details enterprise GTM transitions and recruiting strategies against tech giants like Google Brain. | |
| Research Management and High-Profile Board Leadership | 5 | 4 | 1 | 2 | Matt asks how a young founder managed high-profile board members like John Doerr, Frank Slootman, and John Chambers. Gustavo discusses balancing pure AI research with enterprise products and shares an anecdote about an introduction to Tim Cook. |
Statements from this episode (13)
Sapoznik: Call center agents face an annual turnover rate near 100%
“And then agents have an average attrition rate of about a hundred percent per year, depending on the year.”
Sapoznik: The US employs roughly 3 to 4 million call center agents
“I think in the U.S., most estimates have agent counts about three to four million agents.”
Sapoznik: Legacy contact center incumbents are overwhelmingly underwhelming technology companies
“Most of the legacy incumbents are overwhelmingly underwhelming technology companies.”
Sapoznik: Switching from Siebel to Salesforce gives agents zero productivity
“If you survey the agent that spent eight hours looking at Siebel and now is spending eight hours looking at Salesforce and you ask agent, how much more productive are you? The answer is effectively zero.”
Sapoznik: ASAPP automates customer service via bottoms-up human agent augmentation
“So our approach is automate as much as you can, augment the rest, and the more you augment, the more you can actually automate in a bottoms-up way, not in a rules-based way, which is traditionally how things have been automated in this space.”
Sapoznik: Agents click ASAPP's AI suggestions 80% of the time
“And today in our most mature customers, roughly 80% of everything an agent does is click on those suggestions that the system's generating.”
Sapoznik: Under 1% of traditional call center interactions undergo manual QA
“So as a result of that pretty burdensome and silly way of doing quality assurance, on average, less than one percent of interactions end up going through that process.”
Sapoznik: GPT-4 real-time call center suggestions cost $50-$100 per conversation
“In an average conversation, there's thousands of pings to this model. You can say, well, couldn't GPT-IV, for example, make those predictions? It actually could, and it'd probably do a very decent job out of the box. The problem is your conversation will cost …”
Sapoznik: OpenAI is aggressively commoditizing NLP and machine learning capabilities
“Where we see this going is what open AI and all that we're doing is really a really fantastic and spectacularly aggressive commoditization of NLP and some other machine learning capabilities.”
Sapoznik: Current LLMs cannot execute UI-driven action workflows for call centers
“Which large language models can do very good jobs at predicting language and things of that nature, but they really can't do very much on the action space, especially in that action space, not API driven, but it's literally a UI workflow that an agent's doing.”
Sapoznik: Operating enterprise startups in stealth mode is a stupid idea
“That's a very, very stupid idea. Do not be on stealth mode. There's very seldom good reasons for doing it.”
Sapoznik: Great board members advise but recognize the CEO is driving
“One is being a sounding board that recognizes you're driving the car. I'm not driving the car. I'll tell you what I would be doing if I were driving it, but even if you disagree with what I'm saying, I recognize you're driving it.”
Sapoznik: John Doerr introduced ASAPP to Tim Cook in 12 minutes
“And I asked John Dole, do you think you can help me with someone at Apple? Cause there's this Apple, but he's like, what do you need? Can you drop me a note? And then, 12 minutes later, I had an introduction to Tim Cook saying, can you help this kid with like …”