In January, travel advisors using Tern recorded $170M in bookings. That puts us at nearly $2B in annualized GMV, up from $100M just twelve months ago.
Over the next several weeks, members of our team are going to write about how AI is changing the way we build this company. Real numbers, real examples, stuff that didn't work.
Our hope is that this is valuable to our Travel Advisor community as we show how we're using AI in our business. We also hope it is helpful for the startup community to be able to share learnings on our journey.
Travel Advisors are the Future of Travel
Every few years, there's a new wave of headlines predicting the end of the travel advisor. First it was OTAs. Then mobile. Now it's AI. The technology changes but the headline stays the same: "This will kill the travel advisor."
We have a contrarian take. We think the AI wave of technology will supercharge the travel advisor. Let me explain why.
What Travel Advisors and Enterprise Sales have in Common
There's a widely held view in tech that sales will be one of the last things AI replaces. Salespeople create demand, manage emotions, build trust, tell stories. Enterprise sales remains overwhelmingly human-led. For many companies, software is among their largest and most complicated purchases. The investment is significant, the organizational change is real, and the buying process requires a human to navigate it.
For many individuals and families, travel is the single largest discretionary purchase they'll make. It comes with its own set of emotions, its own overwhelm, its own complexity of managing relationships across a family or a group. The parallels are real: high-stakes purchases with real emotional weight stay human-led.
Travel advisors are not admins.
That's the connection the tech industry keeps missing.
A great advisor doesn't help you book a flight. They create demand where it didn't exist. A couple has no idea they want to do a river cruise through Eastern Europe until they hear an advisor present it at a community event. Next thing they know, they're on a group trip with twenty people from their neighborhood. AI isn't close to replicating that.
There's also a depth of access that outsiders don't appreciate. Just this past week I was talking to a destination management company focused on Italy. They have a relationship with a PhD archaeologist who also happens to be an opera singer. That person leads their small group tours. You will not find them on an OTA. You will not find them on a website. They have no online presence at all. It's a relationship that creates an experience you can only access through this channel. Travel is full of these relationships, and they are what make advisor-sold travel genuinely irreplaceable.
Advisors drove $110B in sales in 2023. They sell 67% of all cruises, 66% of tours. Their clients are loyal, affluent, and value expertise over price. These are not the economics of a dying category. But the tools advisors have been given? Those are dying. Most advisors spend more time on invisible admin work than actually advising.
Our bet: remove that invisible work, and advisors don't just survive the AI era. They grow. They take share. They spend more time doing what they're uniquely good at. Building relationships, generating demand, telling stories that inspire people to travel. The pie gets bigger.
Supercharging Tern with AI
When we started Tern, nobody in this space was saying "build one vertically integrated system that does everything." There were itinerary builders adding features slowly. CRMs adding features slowly. Back-office tools mostly not adding features at all.
We took the opposite approach. Build it all, build it fast. One-week shipping cycles. Every squad of engineers ships a feature every week. That demands ruthless prioritization and engineers who can move. We had that foundation before AI entered the picture.
Then AI got good. And we didn't just add a chatbot to the product. We rethought how our entire company operates.
On the product side, we're building toward something we think will matter a lot. Travel advisors don't trust generic AI because it gives you the average of everything on the internet. What makes a great advisor great is their context. The relationships they've built, what they've learned about their clients over years, the trusted sources they rely on. We're building tooling that combines trusted supplier data with the proprietary context that each advisor brings so that AI actually becomes a genuine partner in delivering exceptional experiences. Not a replacement. A partner that knows what they know and helps them use it. More on this in the coming weeks.
The results: 200% increase in velocity and accelerating
Since rolling out AI workflows across engineering in January: throughput up 86%, same headcount. PRs per week from 85 to 158. Merge time from 13.7 hours to 4.5. Code quality improved across the board: complexity down 29%, production errors down 38%, rework down 23%. Twelve independent metrics all moved the right direction at the same time.
80-90% of support tickets handled by AI with high satisfaction. First response time from 4 hours to one hour. We're compressing insurance partner integrations from 5 weeks to under 1 week. We're using AI to eliminate switching costs so advisors can move off legacy platforms without pain.
What we're still figuring out
I want to be honest about this because it's the most interesting part.
Our engineering team can now build faster than our product team can think. That sounds like a great problem. It is. But it's also a real one.
The hard work in travel tech isn't writing code. It's understanding the layers of complexity in how advisors, agencies, host agencies, and suppliers actually work together.
Commission reconciliation alone can involve fifteen different models depending on the relationship between a supplier and a host agency and an agency and an advisor. None of this is documented anywhere. It lives in people's heads and in manual processes spread across a dozen different tools built up over decades.
Vibe Coding & the Travel Industry
People talk about vibe coding like it solves everything. I'm skeptical. If you're building a simple SaaS product, sure. In the past two weeks we've vibe coded replacements for three different SaaS tools we were paying for.
So I get it. That's real. But in our world, the majority of the work is extracting information that doesn't exist in any system and then figuring out the right abstraction to support it.
And as engineering gets faster, that ratio only grows. The bottleneck isn't building. It's knowing what to build. AI is an incredible partner once you know what to build. But the deep product thinking required to understand these problems still takes real human brainpower.
So we're wrestling with a question I think a lot of companies are going to face: when your engineering capacity suddenly doubles, how do you decide what requires deep, careful product thinking and what you can just build, ship, and iterate on? Where do you maintain your quality bar and where do you short-circuit the process to learn faster? We don't have this fully figured out. I don't think anyone does yet.
What we are finding is that product taste and customer understanding are becoming the limiting factors. Our Head of Product used to spend one to two hours writing a PRD. Now she records a Loom walking through the feature, runs it through Claude, and has a solid PRD in fifteen minutes. The same Loom becomes a help center article, a customer-facing product update, a blog post. That part is working.
But testing still takes too long. Complex rollouts still feel like project management. And there's a deeper question about how PMs should spend their time when AI can write code but can't evaluate whether the experience is right.
- Should a PM put up a PR that produces code they aren't equipped to review?
- Should they write five well-thought-out cards that an engineer can spin up five agents against, with the engineer reviewing all that code in her area of expertise?
We think the answer is obvious, and it means the role of product is shifting toward taste and customer judgment, not toward becoming a worse version of an engineer.
The series
A lot of our leadership team built Handshake from zero to $3.5B & 800+ employees. We know the SaaS-era marketplace playbook. What we're figuring out now is how to reimagine it for the AI era. What changes, what stays the same, and where the hype outpaces reality.
Over the coming weeks:
- Pete & Brian on doubling R&D throughput.
- Molly on how AI is reshaping the product management role.
- Zach on our AI agentic platform and it’s impact on velocity
- Matt & Rodney on compressing insurance integrations with context engineering.
- Abby on rethinking customer operations.
- Gal on eliminating switching costs with AI.
If you want to follow along, give Tern a follow on LinkedIn. We'll be publishing weekly.
$170M in January. Nearly $2B annualized. Thirty-five people. And we just turned on our first booking integration. There's a lot more to build.