Notes from real engagements, for the decision you are about to make.
For the people who decide what to build, what to move and what to stop. So far there are 22 of them, each written from work we delivered, including the parts that went wrong.
Before you build or buy
Owning the code, comparing quotes, deciding between a product and a build.
- Acceptance criteria written so that done is a fact, not an argumentMost delivery disputes are not about quality, they are about whether the work is finished. Acceptance criteria written before the build make that a matter of record rather than opinion.
- The project ran over budget. Was it the estimate?When a software project overruns, the estimate takes the blame. Sometimes it deserves it, usually it shares it. A four-bucket test tells you which, and what to change next time.
- Writing requirements that produce comparable quotesThree vendors, three quotes, three different scopes and no way to compare them. The problem is upstream: five sections decide whether quotes are comparable, and length is not one of them.
- Build or buy: the calculation almost nobody actually runsOff-the-shelf always looks cheaper on day one. The five criteria that genuinely decide it, the three-year arithmetic no vendor shows you, and the trap of the 80 percent fit.
- Who owns the code your contractor writes? Probably not you, and paying does not change itUnder US copyright law, software written by a contractor is not work made for hire by default. Without a signed assignment, the developer keeps it. The four clauses to check.
Before you sign, or switch vendors
Fixed fee or time and materials, exit clauses, handover: what to settle before signing, and how to leave without losing the system.
- Changing software vendors without losing the systemThe decision is rarely the hard part. The order of operations is. Six steps, the one that must never be skipped, and why you tell the outgoing vendor before you think you should.
- The exit clause you negotiate before you need itAn exit clause is only obtainable while you can still walk away. The five things it must name, the three parameters that decide whether it works, and a clause you can reuse.
- Fixed fee or time and materials: what each one makes your vendor optimize forNeither model is honest or dishonest. Each one rewards a different behavior, and knowing which behavior you are paying for is the whole decision. Plus the two hybrids that actually work.
Before you automate or add AI
What pays for itself, what quietly does not, and how to measure the difference.
- Self-hosted model or vendor API: the five-line calculationRunning a model on your own hardware costs little per request and a lot in human time. The five lines of the comparison, and the three situations where self-hosting genuinely wins.
- Measuring what an AI agent actually saves, six months inThe number promised before the project and the number observed after are never the same. How to measure without flattering yourself, and the two gains everybody forgets to count.
- Why AI pilots stall before production, and the three questions that predict itThe pilot works, everyone is pleased, and nothing ships. The cause is almost never the model. Three questions asked at the start predict the outcome better than any proof of concept.
When nobody senior owns the technology
Fractional or interim leadership, what it costs, and the risks nobody registers.
- Delivering a platform, then delivering the people who will run itOn a national addressing program the condition was not negotiable: code and hosting in the institution's name, a local project manager and two developers trained on the project. What that changes about how you build.
- When one person is your only backup, and nobody put it on the risk registerEvery small IT setup has one person who alone can fix a particular thing. That is not a staffing problem but an availability risk, and it shrinks in weeks rather than by hiring.
- What a fractional CTO costs, and the three ways it gets pricedDay rate, monthly retainer, fixed scope. The three models make your vendor optimize for different things, and the cheapest headline number is usually the most expensive arrangement.
- Fractional CTO, interim CTO, or consultant: what you are actually buyingThree arrangements that look alike on a résumé and differ on one thing: who holds the decision. How to tell which one your situation needs, and the drift to watch for.
When the old system has to move
Migrations, renewals, backups and debt, priced before you decide.
- Forty-seven alerts one morning, forty-three of them falseThe morning we checked forty-seven alerts by hand and forty-three were noise, the problem became the instrument. What we built instead, and the five things a status check never sees.
- Rebuilding more than ten business applications without ever stopping the businessInside the IT department of a multi-country distribution group, more than ten applications were rebuilt. One at a time, never all together. What that choice costs, and what it avoids.
- Cloud or on-premise: the five costs the comparison always leaves outThe comparison is usually hardware price against monthly subscription. Five costs are missing from it, in both directions, and three cases where staying on your own hardware holds up.
- Making the case for technical debt without using the words technical debt“We should rebuild this properly” never gets funded, and that is rational. Three ways to convert technical debt into delay, risk and money, so it competes with the requests that do get funded.
- Modernize or renegotiate: price the exit before you decideEveryone advising you earns from the answer being yes. The four inputs that settle it, and why a costed migration is worth having even when you decide to stay.
One of these describes your quarter? Tell us the rest.
Describe what you need built or decided, and where your data and customers are. A person reads it and answers within 48 hours.




















