Most businesses are reacting to the hype, but answering that first makes everything that follows simpler and cheaper.
Most people are either sold on AI doing everything or worried about what it might do unsupervised, both of which are fair views to have. Pointing it at one defined workflow step is already cheap and reliable, however it's giving it access to the right data that's the tricky bit.
Data can be scattered across systems that were never built to talk to each other. Processes only work because someone re-keys the details by hand. That's just how businesses grow, but pointing AI at it can make even the best tools fall over.
Thinking clearly about what you want to achieve with AI, and connecting the data you already have makes AI decisions smaller, cheaper and safer.
The gap between what a business needs and what its software does usually gets closed by someone re-keying data by hand.
I learned that the long way. Chartered accountancy at Deloitte, then nine years as finance director of a superyacht management company here in Palma, designing and building the systems the fleet's finances ran on. Then six years as Product Director at a San Francisco software practice, architecting custom builds for clients whose workflows don't fit off-the-shelf tools. For the last few years that has meant building with AI tools every working day.
Nabarro Consulting applies both sides at once: the accountant's understanding of how a business actually runs, and the builder's understanding of what software and AI can honestly do.
I don't resell licences and I'm not tied to any vendors, so the advice can be "don't buy anything yet" when that's the truth.
No platform, no licences to resell, no three-year programme. An engagement has two parts taken in a deliberate sequence.
A fixed-price look at how the work actually gets done. I sit with the people who do it, in the real spreadsheets and systems, and map where the time goes and where the value sits.
The output is a written report that stands on its own, with or without me: what to fix, in what order, what it costs, and what it returns.
The report becomes a sequence of fixed-price steps, each one small enough to decide on its own. Built on tools you already own, nothing ripped out, no new system to learn.
Each step pays its way before the next one gets committed.
Already bought the tools and it stalled? That's common, and it's usually the sequence, not the tools. Discovery starts from wherever things actually stand, including there.
Help deciding what AI should do in a business before money gets spent on it. It starts from how the work runs today and where the data sits, then sets out what to automate, what to leave alone, and in what order.
Start by deciding where this should land: full automation at one end, doing nothing at the other, and a lot of sensible ground in between. Then connect the data that is already there. In most businesses the data is the gap, not the AI.
Discovery first: a fixed-price assessment of how the work actually gets done, delivered as a written report with costs and expected returns. Then, if it makes sense, a build in fixed-price steps that you commit to one at a time.
Happy to have a chat, if I can help I'll tell you how, and if I can't I'll say so.
