AI, business software and automation

You measure it before you trust it. On every AI project I run the pipeline alongside your existing process for a few weeks first, extracting or answering in parallel without touching anything live. That gives a real accuracy figure on your actual data rather than a vendor benchmark, and it makes the go/no-go an evidence-based decision. If the numbers are not good enough, you have spent a fraction of the budget finding that out.

Usually not, but the useful projects are smaller than the hype suggests. The ones that reliably pay back are narrow: reading documents so nobody has to key them in, answering the same repetitive questions from your own documentation, sorting and routing an inbox. Those start around $6,000 and pay for themselves in staff hours. It is the sprawling "an AI that knows everything about our business" projects that burn money, and I will steer you away from those.

A simple marketing site starts around $2,000 and small automations around $3,000. A solid first version of business software is usually $12,000 to $60,000, driven mostly by how many systems it must integrate with and how complex the permissions are — not by how many screens it has. I give a fixed price for phase one after a discovery conversation, so you are never signing a blank cheque.

Automations typically take one to five weeks. A marketing site takes two to six weeks, most of which is content. Larger business software runs from six weeks to five months depending on scope. I ship in phases, so you have something real in your hands long before the final phase lands.

I do. The person you talk through the problem with is the person who designs and builds it — you will not be handed to someone else after signing. If a project needs a specialist I do not have, a designer say, I tell you upfront and you approve them.

Yes, completely. The repository lives in your account, the domain and hosting are in your name, and your data exports without asking me. This is arranged at the start of the project, not negotiated at the end.

If I underestimated, that is my cost and I absorb it. If the scope genuinely changes — you want something we did not agree — we price that separately before I build it. The distinction is written into the agreement so it is never a surprise.

That is the normal starting point, and it is what the free call is for. Describe the problem in plain English and I will tell you what category it falls into, roughly what it costs, and whether an existing product would do the job better. I turn down work regularly on that last point.

Yes. Most of my work is remote and asynchronous, with a scheduled call each week. I overlap a few hours with most timezones and have run projects across Europe, North America and the Middle East.