More Customers Chose the AI Agent Than Anyone Expected | Tom Chen, Aircall
Episode
56 min
Read time
2 min
Topics
Productivity, Fundraising & VC, Artificial Intelligence
AI-Generated Summary
Key Takeaways
- ✓AI Deployment Entry Point: Start AI voice agents with after-hours and overflow call handling before touching core business hours operations. This "upside-only" approach lets companies test performance with minimal risk — missed calls that went unanswered anyway — then expand coverage incrementally as confidence builds from real customer data.
- ✓Customer Choice Architecture: When one Aircall customer in Australia offered callers a choice between a human agent or faster AI-assisted service, far more customers selected the AI option than anticipated. Giving customers an explicit, honest trade-off increases both operational efficiency and customer satisfaction scores simultaneously, rather than forcing AI on unwilling callers.
- ✓AI vs. Human Performance Benchmark: AI voice agents are unlikely to match a company's top, longest-tenured human agent who carries undocumented tribal knowledge. However, they consistently outperform median call center representatives on two measurable dimensions: script adherence and infinite patience — making the median rep, not the best, the correct performance comparison.
- ✓Scalability Economics: A single AI voice line on Aircall handles 100 concurrent calls, operates 24/7 without breaks, and deploys instantly without training costs. Tom Chen estimates the fully loaded cost delivers value at roughly one-third to one-tenth the expense of an equivalent human operation, making the ROI calculation straightforward for most small teams.
- ✓Knowledge Gap as the Real Bottleneck: The primary obstacle to high AI resolution rates is not model capability but undocumented tribal knowledge inside smaller companies. AI performs well given complete information, but SMBs rarely have processes documented thoroughly. Businesses should audit and document operational knowledge before deploying voice agents to avoid underperformance and unrealistic automation expectations.
What It Covers
Tom Chen, Chief Product Officer at Aircall, covers how AI voice agents are transforming customer communications for small and mid-sized businesses. Aircall operates across 10+ global offices, handles 100 concurrent AI calls per line, and positions voice as a competitive advantage previously too costly for most companies.
Key Questions Answered
- •AI Deployment Entry Point: Start AI voice agents with after-hours and overflow call handling before touching core business hours operations. This "upside-only" approach lets companies test performance with minimal risk — missed calls that went unanswered anyway — then expand coverage incrementally as confidence builds from real customer data.
- •Customer Choice Architecture: When one Aircall customer in Australia offered callers a choice between a human agent or faster AI-assisted service, far more customers selected the AI option than anticipated. Giving customers an explicit, honest trade-off increases both operational efficiency and customer satisfaction scores simultaneously, rather than forcing AI on unwilling callers.
- •AI vs. Human Performance Benchmark: AI voice agents are unlikely to match a company's top, longest-tenured human agent who carries undocumented tribal knowledge. However, they consistently outperform median call center representatives on two measurable dimensions: script adherence and infinite patience — making the median rep, not the best, the correct performance comparison.
- •Scalability Economics: A single AI voice line on Aircall handles 100 concurrent calls, operates 24/7 without breaks, and deploys instantly without training costs. Tom Chen estimates the fully loaded cost delivers value at roughly one-third to one-tenth the expense of an equivalent human operation, making the ROI calculation straightforward for most small teams.
- •Knowledge Gap as the Real Bottleneck: The primary obstacle to high AI resolution rates is not model capability but undocumented tribal knowledge inside smaller companies. AI performs well given complete information, but SMBs rarely have processes documented thoroughly. Businesses should audit and document operational knowledge before deploying voice agents to avoid underperformance and unrealistic automation expectations.
Notable Moment
Tom Chen describes how most businesses benchmark AI agents against their absolute best human employees, which consistently produces disappointment. Experienced customer service leaders who shift the comparison to median agent performance arrive at a fundamentally different — and far more favorable — conclusion about AI deployment readiness.
Episode Transcript
I've had the experience of stumbling onto an AI voice agent. You can hardly tell at this point that it's not a human. Typically, a human agent can only take one call at a time. But with an AI voice agent on air call, they can take a 100 concurrent ones at a time. I'm still getting phone trees or chat bots that have canned answers, and it's just frustrating. It shouldn't be like that. How do you see the market? Is an AI voice agent going to be as performant as your smartest agent who's been around, who kinda has all the knowledge of the things that are never documented? Probably not yet. But I do think that it is better than the average rep in call center. Can you start by introducing yourself to listeners? Give a little bit of your background so far as it's relevant and how you got to Aircall and then what Aircall is, and and we'll start talking about it more deeply. So I, Tom, I work at Aircall as the chief product officer. When I first joined Aircall about two years ago, it was a little over two years ago, almost getting two and a half. We were in a completely different world. I mean, this was about a year after, I think, ChatGPT had launched. There was some early promise, but I don't think anybody sitting in my shoes or most people's shoes would have guessed how quickly it might have moved. Of course, there were folks who are maybe more prescient than, or or more future looking, better future predictors than than some of us. But I think it probably caught most of us by surprise in terms of how well and how fast it's moved, especially in certain, general AI spaces like coding and others. And so since my time, you know, joining, Aircall to now, you we've really turned from what you would describe as a software SaaS business to much more of an AI business, because almost all customers around the world no longer look at, you know, a phone system or customer communication software as just the software itself, but more around, you know, what can you help me actually resolve? Whether it's tickets from conversations or, you know, can can you actually autonomously handle conversations? And while deployment and, you know, uptick is not across the entire world uniformly at a 100%, we're still in the single digit early innings. You can bet that all businesses are thinking about it and trying to find, you know, a future partner who have the capabilities that they can really lean on and trust, moving forward. And trust is a is a is an interesting thing because it it runs across the gamut of not just the AI services, but also your base reliability in, you know, whether it's your chat or telephony services, your ability to navigate, the different telco regulations around the world. All of those …
Get the full transcript (9,554 words) + summary by email — free
One-time email with the complete transcript and AI summary of this episode. No account needed.
One email, no spam. We’ll also show you what SignalCast does.
You just read a 3-minute summary of a 53-minute episode.
Get Eye on AI summarized like this every Monday — plus up to 2 more podcasts, free.
Pick Your Podcasts — FreeKeep Reading
More from Eye on AI
In 5 to 10 Years, Using Weapons Without AI Will Be Considered Unethical | Yaroslav Azhnyuk, The Fourth Law
Aug 31 · 53 min
a16z Podcast
The Self-Improving Company | Kavak's AI Playbook
Aug 10
More from Eye on AI
Inside Ukraine's Azov Drone R&D: The Engineer Building AI Weapons 18 km From the Front Line | Alexander Palamarchuk
Aug 27 · 41 min
SaaStr Podcast
SaaStr 844: The Top 5 Issues Managing Multiple AI Agents in Production with SaaStr's CEO and Chief AI Officer
Mar 4
More from Eye on AI
We summarize every new episode. Want them in your inbox?
In 5 to 10 Years, Using Weapons Without AI Will Be Considered Unethical | Yaroslav Azhnyuk, The Fourth Law
Inside Ukraine's Azov Drone R&D: The Engineer Building AI Weapons 18 km From the Front Line | Alexander Palamarchuk
95% of AI Agent Projects Fail to Reach Production. Here's Why | Manoj Saxena, TrustWise
From Zero to 150 Robots in Just 20 Months | Mike LeBlanc, Foundation Future Industries
Why People Are Paying 10x More for AI - and What That Means for the Chip Market | Sid Sheth, d-Matrix
Similar Episodes
Related episodes from other podcasts
a16z Podcast
Aug 10
The Self-Improving Company | Kavak's AI Playbook
SaaStr Podcast
Mar 4
SaaStr 844: The Top 5 Issues Managing Multiple AI Agents in Production with SaaStr's CEO and Chief AI Officer
NVIDIA AI Podcast
Jan 21
From Warehouses to Robot Shoppers: Jason Goldberg Talks Retail’s AI Makeover - Ep. 286
SaaStr Podcast
Nov 26
SaaStr 831: How We Use 20+ AI Agents for Marketing & Go-to-Market with SaaStr's Chief AI Officer and CEO & Founder
Masters of Scale
Apr 21
A first look at Samsung’s blueprint to win the AI era, with Mauro Porcini
Explore Related Topics
This podcast is featured in Best AI Podcasts (2026) — ranked and reviewed with AI summaries.
Read this week's AI & Machine Learning Podcast Insights — cross-podcast analysis updated weekly.
You're clearly into Eye on AI.
Every Monday, we deliver AI summaries of the latest episodes from Eye on AI and 192+ other podcasts. Free for one show.
Start My Monday DigestNo credit card · Unsubscribe anytime