AI & operations · Technology leadership
AI in the operating model. Not in a pilot.
Every founder I meet knows AI should be doing more in their business. Almost none can point to where it's actually saving an hour or making a dollar. The gap was never the tools, it's that nobody senior owns the question.
No hype · No pilot theatre · Payback estimated before you commit
01 The current state
Where AI is actually stalling
Not a straw man: these are the five states I keep finding inside real companies. Most are in at least two.
- 01
The company pays for six AI tools. Two get opened. None has changed a process.
- 02
One enthusiast automated their own corner. Nothing they built survives their holiday.
- 03
The pilot demoed brilliantly in March. It is still a pilot.
- 04
Everyone "uses ChatGPT", so output quality now depends entirely on who wrote the prompt.
- 05
You've been pitched an AI transformation. Nobody could tell you which line of the P&L it changes.
02 The unglamorous truth
You can't automate a process nobody has written down
AI automates processes. When the process only exists in someone's head (different every time, dependent on who's doing it), there is nothing to automate, and the pilot quietly dies the moment its champion gets busy.
So the work starts embarrassingly analogue: map the process, write it down, measure it. Then automate the steps a machine should own, and keep humans on the judgment calls. That's why AI leadership here isn't a separate service. It's the technology arm of the same operating method as everything else.
03 The work
What actually gets built
Concrete systems with named owners: priced and payback-estimated in the recon diagnosis before you commit to any of them.
CRM & pipeline automationLead routing, follow-up sequences, and the sales-to-delivery handoff that keeps falling on the floor.
On scope
Proposal & reporting generationThe documents that eat senior hours: drafted by machine, finished by humans, on your templates.
On scope
Service-delivery copilotsSOP-embedded assistants so junior people execute at the senior standard.
On scope
Quality-assurance loopsOutputs checked against the written standard before a client ever sees them.
On scope
Internal knowledge systemsWhat the company actually knows: searchable and answerable, instead of scattered across fourteen tools.
On scope
04 The filter
Worth building. Not worth building.
Half the value of senior technology judgment is the list of things you don't build.
Worth building
- Automation tied to a documented, measured process
- Work that eats senior hours and has a clear right answer
- Handoffs between departments, the seams where things drop
- Anything measured before and after, with a named owner
Not worth building
- A chatbot because a competitor has one
- AI content at volume that reads like AI content at volume
- Automating judgment calls that need a human's name on them
- Anything nobody will own after I leave
05 Answers
Asked on almost every call
Do I need an AI strategy?
No. You need an operating strategy in which AI does specific, named jobs. "AI strategy" as a standalone document is how companies end up with pilots and subscriptions instead of saved hours. The sequence that works is older and duller: document the process, measure it, automate the steps a machine should own, keep humans on the judgment calls.
What stack do you work with?
Yours. The recent builds have run on GoHighLevel, HubSpot, Zapier, Make, Notion, Asana and the OpenAI and Anthropic APIs, but the tool list matters far less than the process map. A stack audit is the first fortnight's work: what you have, what overlaps, what integrates, and what should be deleted before anything new is added.
What does AI and automation work cost?
It's scoped after the two-week recon, not guessed on a sales call. The diagnosis prices each build and, more usefully, estimates its payback, so you're comparing cost against named hours saved rather than against a feeling. If nothing on the ranked list pays for itself, that's the finding, and it costs you the recon.
The engagements page shows how scoping works across the practice.
Will this replace my team?
In companies this size, it mostly returns hours to people you can't hire enough of. The economics of a 15-person firm are rarely improved by removing people; they're improved when senior people stop doing $15-an-hour work. If your plan is to automate the team out, I'm the wrong hire. That's a stated boundary, not a sales line.
Next step
Find the hours AI is actually hiding in your business.
Bring the stack, the half-dead pilots, the subscription list. Forty-five minutes later you have a written read on what's worth automating first, and what isn't.