What I believe about
building technology
inside a real business.
Two advisors with identical credentials produce entirely different outcomes. The difference is never the résumé - it is how each one decides when the information is incomplete and the money is real. Here is how I decide.
Twenty years of being called in after things went wrong, and I have almost never found a technology failure.
I find the wrong thing, built correctly. A good system solving a problem the business already moved past. The code did what it was asked - the asking was the failure.
Implementation is available everywhere, at every price point. What is scarce is being careful about which thing gets built, and saying so when the answer is inconvenient. Seven principles produce that care.
Architecture is a business decision wearing technical clothes.
Every system hard-codes an assumption about how you will grow. Get it wrong and no amount of engineering rescues it - which is why this conversation belongs to the owner, not only the engineers.
A platform built for one warehouse quietly assumes there will only ever be one.
Nobody writes that assumption down, so it survives until it becomes a wall.
You never have to read a diagram - just answer what this business should be in five years.
The expensive mistakes are made in month one, not month nine.
By the time a project looks like it is failing, the decision that killed it is six months old. I front-load the uncomfortable questions, because that is the only point where being wrong is still cheap.
Month one, changing your mind costs a conversation.
Month nine, it costs the sunk work and whoever championed it.
So the first weeks are deliberately uncomfortable. You are buying the option to be wrong cheaply.
Never automate a process you haven't earned the right to keep.
Most "AI transformation" pays good money to make waste move faster. First we establish whether the process should exist at all. The cheapest system in your business is the one you deleted.
Automation is a multiplier, and multipliers are indifferent to sign.
Approval chains from a decade-old incident. Reports nobody has read since 2019.
The first pass is always subtractive: what can we simply stop doing?
If your team can't own it, I haven't finished.
A system only its builder understands is a liability with a good interface. Handover is a design constraint, not a closing phase. You should be able to fire me without consequence.
Some consultants build systems just beyond the client's reach. It is profitable and I find it contemptible.
A system your team can't operate isn't an asset. It's a lien.
So I favour boring technology and written reasoning over clever patterns.
Correctness before speed. Every time.
Speed is easy to sell and expensive to unwind. Where money, inventory, or patients are involved, fast-and-wrong is the costliest outcome available - and it arrives disguised as progress.
Correctness is invisible when it works. You only experience its absence.
A fast wrong answer is worse than a slow one, because it gets acted upon.
Shipping a month late is recoverable. Shipping a corrupted ledger is not.
Say the number out loud.
Every recommendation carries three figures: what it costs, what it returns, and what doing nothing costs. An advisor who will not quantify their advice is selling comfort.
Robust. Scalable. Future-proof. Words arranged so nobody can ever be wrong.
You get three numbers instead: cost, return, and the cost of doing nothing.
That third number is the one most consultants leave out.
The best advice I give often costs me the project.
"Don't build that." "Buy it off the shelf." "Wait two quarters." I have said all three to paying clients. It is why engagements here become decade-long relationships instead of transactions.
Told a manufacturer their inspection problem was a lighting problem. Buy fixtures, not AI.
Told a founder his platform had eighteen good months left. Spend the capital on sales.
Each cost me a project - and produced a client who now calls me first.
Who this is and isn't for.
A position worth holding excludes someone. Knowing which column you're in saves us both a quarter.
We will work well together if…
- You own the outcome and sign for it.
- The decision is expensive to reverse, and you know it.
- You want an inconvenient assessment now, not a comfortable one in month six.
- You want the person who does the thinking present during delivery.
- You invest in judgement, not headcount.
Hire someone else if…
- You have decided what to build and want it executed without comment.
- You need the cheapest quote. Someone is always cheaper.
- You want a large team on site quickly. I work with small, senior groups.
- You need a vendor who agrees with the board rather than pressure-tests it.
- The project exists mainly to be announced.
One last thing.
Disagreed with three or more of these? Useful information - don't hire me. We'd spend the engagement arguing premises instead of building your business.
Found yourself nodding? Then you already know what working together feels like. Very few consultants tell you that before the invoice.