About

David Kotler

David Kotler

AI and automation engineer

I build AI systems that go into real work, then take them apart into an explanation you can actually use — not tool demos, but what enters a team's routine and stays there.

I'm a software engineer working on AI, automation and integrations. Most of what I do sits at the seam between a model and an existing system: how to connect them, what an agent is allowed to do on its own, and where a person has to stand before anything actually happens.

What I've learned from building this is that most of what gets sold as an "AI project" is a new tool pushed into a process nobody bothered to map. It works in the demo and breaks in the second week, and when that happens people conclude AI isn't a fit — when the problem was somewhere else entirely.

Lea AI is where I share that: articles about what I built and what didn't work, case studies of real systems, and also work with businesses that want to turn a technical capability into a process that works day to day. Same angle, two audiences.

The work doesn't end when the automation goes live. If the team hasn't adopted it, there's no project — just another tool open in a browser tab.

How I work

Four things I don't move on

  • 01

    Build before opinion

    I write about what I've built, tested or taken apart — not about what I've read that someone else built. If I haven't run it, it doesn't go into an article as fact.

  • 02

    Sources, or silence

    A figure without a source isn't a figure. If I don't have a number I can verify, I'd rather describe the mechanism than invent a percentage.

  • 03

    Limits and permissions before capabilities

    Before an agent performs a real action, someone has to decide what it reads, what it writes, and what needs human approval. That's the part most demos skip.

  • 04

    Production over demo

    A demo succeeds when it runs once. A system succeeds when it's still working in month three, after someone else has touched it.

This site and the diagnosis were built the same way I recommend to clients — including the same first question: what's the process, and only then, which tool.

Where to go from here

What's here to read, see and do

  • Articles

    What I've learned from the building itself — models, agents, MCP and production.

    Read

  • Projects

    Case studies of real systems, including the decisions and the limits.

    See what I've built

For businesses

Is there a process that keeps repeating?

The diagnosis takes a few minutes and returns a readiness score for every area of the business, plus one focused recommendation to start from.

Start the AI diagnosis for your business