I help knowledge teams - sales, marketing and operations - put AI to work in the jobs they already do, as agents that live in Teams, Slack and the inbox and make the day lighter. When the right one doesn't exist yet, I build it.
I'm Eddie, and I'm not an engineer. I run sales and marketing for a living and build with AI because I'm hooked on it. That mix is the point: I know how normal teams actually work, where the day goes, what they'll never open. So when I put AI into the work, it fits. In my own commercial team I've built a sales co-pilot, an agent SDK, and an AI chief of staff - not demos, but agents people open every Tuesday morning, right inside Teams, Slack and their inbox. I don't think AI is magic. I think it's plumbing: useful when it's wired into how you already work, and quiet when it's done right. Most teams don't need a strategy deck. They need someone who speaks their language, finds the three jobs worth automating, and ships them.
I work with non-technical teams - sales, marketing, ops - to put AI where the work already happens, in Teams, Slack and the inbox, so it shows up in the day instead of in another app no one opens. We start from the jobs you already do, not from a tool you read about. Then I teach your people to run it themselves.
e.g. a marketing team going from "we tried ChatGPT once" to an agent in their Slack that drafts, summarises and chases, used every day.
When the off-the-shelf version doesn't exist, I design and ship the agent that does - an "AI employee" that handles a real slice of the job end to end and lives where your team already works. Scoped tight, built to be used, handed over with the keys.
e.g. Donna, Sdkonna and Ana - three production systems, all in daily use. See below.
Two I built at work, one I built for fun. All shipped, all in real hands - not pilots left on a shelf.
A field-service app for the support team - tickets, spare-part orders and warranty cases from every dealer, in one place. Built so a technician can close a case from the field in a few taps.
An interactive configurator for the seeder - choose crop, rows and spacing and watch the field layout, seed count and price update live. Turns a sales conversation into a self-serve spec.
A playful recipe app for homemade gelato - nutrition, ratings and flavour notes wrapped in a character-driven interface. A side project, shipped for the joy of it.
The hard part of useful AI isn't the model, it's the context. An agent that can't see how your team actually works will guess, and guess wrong. For a small team you don't need a data platform to fix that. You need to capture what's in people's heads and put it somewhere the agent can reach.
The knowledge that runs a small team lives in a few people's heads: how you qualify a lead, what a good reply sounds like, which rules actually matter. Write it down in plain language. That, not a bigger model, is what makes an agent sound like you.
Spread across ten tools, knowledge is invisible to an agent. Give it a single source of truth, a folder or a wiki it reads on every task, so the context is the same whether the work lands in Slack, Teams or the inbox.
Knowledge rots the moment the work changes. Make updating it part of the job, not a side project, so the agent gets sharper as you go instead of drifting out of date. A small team can do this without a single new hire.
We sit with the actual work. I watch how your team really spends a week, find the jobs worth automating, and say no to the ones that aren't.
I design and ship it - fast, scoped tight, landing right in the tools your team already uses. You see it working long before it's "finished".
I give you the keys: documentation, training, and a team that can run and extend it without me. No lock-in, no mystery box.
Got a job that AI should be doing and isn't? Tell me about it. No form, no funnel - just send me an email.
eddie@