We coach your subject matter experts to build their own internal tools and analysis with AI, from picking the right first project through to deployment.
“I started building and it was going great. But it got more and more complicated, and now I can’t audit how it works myself. So I ask the AI, but I’m not getting what I need, and I feel stuck. What do I do now?”
We hear some version of this a lot from savvy analysts building with AI. It’s fixable.
In 2026, a few of our customers, and other friends and colleagues, started building their own internal analytics and software with AI tools. Almost universally, it went like this:
“This is incredible, humans are cooked.”
“Wait, a bunch of stuff might be wrong under the hood.”
“OK, I get it. It’s an amazing tool, but a person still needs to guide it.”
The transition from Stage 2 to Stage 3 is where the real work starts, and it’s where we help.
One of our customers asked one of the big systems integrators to bid on some internal pilot projects. The bid came back at multiple millions of dollars. So they said, “you know what, let’s see how far we can get with some subject matter experts and Claude.”
Before AI coding tools existed, we spent a decade building software and machine learning products by hand. We know what “good” looks like, and we can tell when the AI’s work falls short of it.
We’ve built River, our supply chain simulation platform, with LLMs from day one, and we use billions of tokens a month. Two years ago, AI would regularly “accidentally” delete half our codebase trying to fix a bug. We’ve since gotten very good at getting the most out of these tools at the lowest cost.
During their build, the customer started asking us for advice, and eventually we signed up for a consulting project to help their SME with the pilot. In their words, the project was “wildly successful.”
What types of projects can we realistically build internally? How should we even think about this? What about maintenance?
How far can we get with an SME and Claude? When is it appropriate to hire it out?
What are the best practices for working with modern AI? What kinds of failures do we need to watch out for?
We don’t know a lot about software development, and we’re worried about security and data privacy. What should we do about it?
Which roles and people should we enable to do internal builds?
What do we do if we get stuck along the way? Or if it’s gotten so complicated we can’t tell how it works, and the AI gives a different answer every time we ask?
These are the questions we help with, and we can guide you through scoping, the build, and deployment.
A few working sessions on what to build in-house, what to buy or hire out, and the security and maintenance rules to set up front. You leave with a scoped first project.
A few hours a week with your SME through build and deployment: design reviews, getting unstuck, and habits that catch the AI’s mistakes. This is what we did for the customer above.
A structured program for a group of SMEs: working practices, failure modes, security guardrails, and office hours while they build.
Not sure which fits? Tell us what you’re trying to build and we’ll suggest a shape.