Will you just recommend rebuilding everything?

No. Rewrites are the expensive answer and usually the wrong one. No recommendation is issued without naming the alternative that was rejected and why.

AI Strategy Consulting, Architecture Review and Fractional CTO

Most technology problems turn out to be ownership problems wearing a hoodie.

   what is actually stuck?
        │
        ├── nobody will own the decision ──────►  it is an ownership problem,
        │                                         not an architecture one
        │
        ├── the system can't absorb change ────►  architecture review   ~1 week
        │
        ├── "should we be using AI here?" ─────►  AI strategy          1–2 weeks
        │
        ├── we ship, but slowly ───────────────►  process audit        fixed scope
        │
        ├── permanent calls, no senior eng ────►  fractional CTO       ongoing
        │
        └── buying it or investing in it ──────►  due diligence        fast, discreet

 what is actually stuck?

 nobody will own the decision
   ▸ an ownership problem,
     not an architecture one

 the system can't absorb change
   ▸ architecture review · ~1 week

 "should we be using AI here?"
   ▸ AI strategy · 1–2 weeks

 we ship, but slowly
   ▸ process audit · fixed scope

 permanent calls, no senior eng
   ▸ fractional CTO · ongoing

 buying it or investing in it
   ▸ due diligence · fast, discreet

The first branch is the most common one and the cheapest to fix — which is why the first call is free and sometimes ends with you not needing an engagement at all.

What it covers

Architecture review One system, read closely. You leave with a map, a list of what is load-bearing, and the three things to fix first.

AI strategy Whether a model belongs in this product at all, which decision it would augment, and what the boring version costs instead.

Fractional CTO Senior judgement part-time, for teams between their first engineer and their first VP of engineering.

Technical due diligence A second pair of eyes on what you are about to acquire, invest in, or ship. Discreet, fast, unsentimental.

Process audit Where the work actually slows down, named without blame. Usually the dependency graph, rarely the people.

Hiring the first engineers What roles you actually need next, what to test for, and how to tell a strong answer from a confident one.

How it runs

Three questions

What constraint are we removing, what does success look like in six months, and who owns the result afterwards. None are technical; all decide whether the work survives.
Read the system, not the deck

Code, schema, deploy process, incident history. What a team says it does and what the repository says rarely match exactly.
One page, then the detail

The recommendation fits on a page: this is the decision, this is who owns it, this is what we will regret if it is wrong. The rest is appendix.
A walkthrough with the team

Delivered live, to engineers as well as founders, so people can argue with it. Advice nobody challenged rarely survives the quarter.

What you end up with

A one-page recommendation, with the decision named
A map of the system as it actually is, not as documented
The three things to fix first, in order, with reasons
The alternatives that were rejected, and why
An owner named for each recommendation
A recording or notes from the walkthrough, for whoever was not there

How recommendations are made

The alternative is named Nothing ships without stating what was rejected and why. Advice you cannot audit is decoration.

Someone owns it No deliverable without a name attached to keeping it alive after the engagement ends.

It has to fit your context A framework that does not survive your actual team, budget and deadline is not advice.

No is a valid answer Including “do not build this” and “you do not need me”. Both get said on the first call when they are true.

Terms

Architecture review

~1 week, fixed scope

AI strategy

1–2 weeks, fixed scope

Fractional CTO

Ongoing, monthly retainer

First call

Free, and sometimes ends in “you do not need this”

Not a fit if

You need a report that confirms a decision already made
The engineers are not allowed in the room
Nobody will be accountable for the outcome afterwards
The real request is extra hands rather than a judgement call

Questions that come up first

What do we get from an architecture review?

One system, read closely over about a week. A map of how it really works, a list of what is load-bearing, and the three things to fix first — with reasoning, so your engineers can disagree rather than just receive it.

What does a fractional CTO actually do?

Sets the quality bar, makes the architectural calls, and is accountable for them. It fits teams between their first engineer and their first VP of engineering, where decisions are permanent but headcount is not yet.

Can you tell us whether we need AI at all?

Often that is the entire engagement. The projects that fail share a tell: good eval numbers and nobody who wants the output. The question is never whether to use AI, but which human decision it augments — and what the boring version would have cost.

Will you just tell us to rebuild everything?

No. Rewrites are the expensive answer and usually the wrong one. Every recommendation names the alternative that was rejected and why, which makes it much harder to hide behind one.

Who is in the room for the delivery?

The engineers, not only the founders. A recommendation nobody in the room challenged usually does not survive the quarter, and the objections are the most useful part.

What if you cannot help?

You get told on the first call and pointed towards who can. It costs nothing and saves both sides a month.

Technical strategy