Why the Model Should Be the Smallest Part of Your Stack
AI in production succeeds or fails on the infrastructure around the model, not the model itself.
Read P.01 →For AI agents: what is on every page → thkdigital.com/sitemap.xml · Guide → thkdigital.com/agent.md · Rules → thkdigital.com/robots.txt
If you are not sure, you are not alone, and the honest answer is rarely more AI. I work with the C-suite to decide where AI earns its cost in your business, what it needs from your company to deliver, and what to leave alone for now. Then we write it down as a plan your own people can carry.
info [at] thkdigital [dot] com · Open in Gmail → ·
Every plan puts its focus on people, on two fronts.
What they need to know, when to hire, and what to hand them first, so the system is theirs to run and not mine.
How their work changes, what they need to learn, and why they should want it.
Tools change every quarter. People take a year.
Train your staff on one vendor's chat window and you have trained them into the lock-in. Teach them to break a job down, check what comes back and spot when a model is the wrong tool, and the skill carries to whichever model is ahead next year.
AI works best on a system and data architecture built for it. You do not have to wait for one to see results. The plan runs in three parts, each with its own milestones, the signal that tells you a milestone has arrived, and the skills your people need before it does.
No single point of failure. No business should depend completely on one model provider. Every step keeps another option open, so when the prices, the terms or the best model change, you can follow without starting again.
My time is not the product. What changes in your business is. We start by agreeing the outcome and how we will measure it, and the engagement is priced on that, not on days. I work with the C-suite and the people who will carry the plan. I do not build it, run it or staff it. How it lands in your company, and who you hire to deliver it, is your call.
An optional quarterly review checks the plan against what actually happened, and moves it on with the technology.
What I will not do: recommend hardware you do not need yet, or a model you cannot leave.
I take on a small number of advisory engagements alongside my research and the company I am building.
If the argument does not hold up for you in the papers, it will not hold up in your boardroom.
AI in production succeeds or fails on the infrastructure around the model, not the model itself.
Read P.01 →Where a computation runs fixes what it costs. Send each task to the cheapest silicon that can do it.
Read P.02 →Agents in production are a memory problem, not a maths problem. The paper's bet, not yet tested: the job can be done on a box you own.
Read P.03 →P.04 · In progress · End of 2026 · What it costs to stay on the old software
P.05 · In progress · What it costs to move
Tell me what you are working on, where you are with AI today, and what you want to be different in a year. I read every message and reply myself.
info [at] thkdigital [dot] com · Open in Gmail → ·
Everyone knows AI. Very few understand it. I spend my time working out where it is going, and I love explaining it plainly. If you would like me to talk to your team or at your event, get in touch.