Agents

The other week, I wrote here about how, contrary to all the headlines, I don’t think AI is going to leave us all unemployed. 

I think I laid out a pretty rational argument, but the whole topic is more emotional than rational, so it might be worth sharing how we’re using AI at Intelligems to build faster—while still hiring more people.

The short of it, I suppose, is that AI is allowing us to 1) do the work we were already doing faster and 2) do work we weren’t previously able to get to in the time we allocated to do it.

The net benefits of this are twofold: A better product for our customers that we’re able to get to market sooner.

Bug Fixes

We use Devin, an autonomous code agent, to handle bug fixes and smaller development tasks. 

We assign it tickets through our project management software, and it analyzes the problem, asks clarifying questions, and proposes a solution. If we approve the solution, it’ll make code changes for us. 

As a company that’s still very much in the early days of building our product, we have to balance keeping our product as bug free as possible and getting to more of the exciting stuff we want to build. 

Without Devin, we would get to may two out of every 10 bugs and UI fixes in our “triage” queue. Now, we’re getting to 8 of them.

I don’t think, even with hiring an extra engineer, we would get to this level of resolution for two reasons: Our balance needs to be toward product growth, so the natural pull is for everyone to focus more heavily on new stuff. And, there’s not enough “triage” work to which we could dedicate an engineer. So, that person would end up working on product growth stuff and experience the natural pull, as well.

Devin lets us get to work we just otherwise wouldn’t get to.

Code Testing

We use Claude Code to automate our test generation and code review. 

Claude looks at every pull request, checks it for potential bugs, readability and improvements. It also finds edge cases in the code that we might have missed ourselves, and helps us test code we haven’t tested.

Before we were doing this, we were writing our own unit tests for checking our code quality. The ROI on that isn’t fantastic. You need to be responsible, of course, and make sure the code is good, but the highest ROI use of an engineer’s time is building new features and functionality, not doing QA. 

So, we prioritized test coverage. Now, we’re able to greatly expand the score of that test coverage, because Claude code is significantly cheaper than an engineer, which makes the ROI worthwhile.

The result of that is marginally higher quality code that leads to a better product experience overall.

Customer Research

When it was pretty much just Drew and me, it was really easy to listen to what customers wanted, make sense of what they were saying, and come up with a solution. Then again, we didn’t have that many customers.

How do you do that with 1,000s of customers? 

That is basically the job of product managers, but even then, their ability to talk to customers is time limited and can be colored, whether they know it or not, by orienting to one segment of the market or another (among other things). 

We’ve used LLMs to “listen” to 150+ customer calls to identify things like top customer concerns, feature requests, pain points, and company priorities—and then identity patterns within specific customer demographics, so we can understand which segments of the market are looking for which solutions.

In the past, we’d rely on our ability to sort through anecdotes from, say, five calls, and use the “art” of product development to hopefully pick the right feature/functionality. Now, we can use the “science” of it to land on the right feature/functionality at a greater frequency.

Extrapolated, this lets us build more, faster. 

There’s a bunch of other examples across all aspects of our organization. But the key piece for us is that we’re able to hire people to focus on the highest ROI work and get to the lower ROI (but still important) work sooner. 

AI creates a better product, faster and accelerates job creation, because the business case for hiring someone is now stronger than it was before.