The other side of learning AI

The other side of learning AI

Learning to work with our new AI tools overwhelmed me. Here's what changed.


Emmie Hayes

Software Engineer

I wrote this right after Tern's latest product webinar — the launch of our biggest agentic AI release for advisors yet — when I realized how recently I'd been standing exactly where our users are now.

I'm an engineer on the Tern team, and as I watched, I had a strange jolt of recognition. We've built, for travel advisors, almost exactly the thing I live inside every day. The chat space looks the same. The customized quotes the agent spits out while it digs through data and formulates a response, the same. Even the review outputs are organized the same. It felt bizarre, and as the excitement faded it settled into a quiet ache on the advisors' behalf. Because I know what we're really asking of them. We're not just handing them a tool. We're asking them to change how they work, and that kind of change is hard. I know, because I'm not far from where they are right now.

In December, leadership told us we were going all in on AI. At the time, the AI tooling in Cursor was, for me, a half-trustworthy thing to rubber-duck with. I wanted to learn it, but, being newer to the team, I was also dying to learn our product, and a quiet conflict ran in the back of my head every single day about how to spend my time.

Then the decision got made for me. I got pulled into a problem space big enough that I couldn't tiptoe around the new tooling anymore. It was mind-boggling and, frankly, scary. But there's nothing like commission reconciliation to make new AI tooling suddenly seem approachable. It forced me to actually learn the skills, agents, and workflows our team had built, because I wasn't going to hit my deadlines if I couldn't parallelize the work and change the way I think.

That was only a handful of months ago. Baby steps, mostly, and on the good days, bigger leaps. But somewhere in there, the work changed shape. It didn't disappear, and it didn't even get easier. I just stopped being buried in it. As my process took shape, I started to see where the models are strong, where they're weak, and where my own new weaknesses are as I work alongside them.

The moment it really hit me, I was lost in the user experience of a flow, fussing over how it would feel to use, and it struck me that I wasn't thinking about query optimization at all, the model was handling it. I got to obsess over the part that actually matters to me. That was the trade: the tooling took the work I used to grind on, and I got my attention back for the work I love.

So when I watched that webinar, what came over me was empathy. I wasn't watching from a distance. I'd just stepped out of the same whirlwind, and now I was watching us hand our users something that looked almost exactly like the thing that had overwhelmed me. I kept thinking about where they are right now, at their desks, having just watched our team demo something incredible and wanting so badly to do it themselves, maybe feeling a little intimidated and unsure. I wanted to reach through the screen and tell them it felt impossible to me too, and that they're closer than they think.

So if I could say one thing to those users, and honestly to anyone learning this stuff right now, it's this: dig in. Be patient with yourself, and keep pushing even when it feels like there's never enough time. There wasn't enough time for me either. I started small, I kept at it on the hard days, and slowly it handed me back the part of the work I love most. I'm rooting for you to get there too. I really believe you will.