AI is only worth it where it replaces real work
Most AI projects that fail did not fail technically. They failed because nobody could say which task was being replaced, or how the gain would be measured. "We want AI in our product" is not a use case; "our support team spends two hours a day sorting incoming email" is.
So the first conversation is not about models. It is about which task is repetitive enough, high-volume enough and stable enough in its rules to be worth automating — and what would count as success once it is.
The use cases that hold up in production
Four families cover the large majority of what I deliver, and they share a property: the rules are stable and the output is checkable.
- Document processing — extracting structured data from invoices, contracts, forms or CVs, in a format your systems can consume directly.
- Internal documentation search — a RAG assistant that answers from your own procedures and product sheets, citing its sources, instead of from general knowledge.
- First-line customer support — an agent that genuinely resolves part of the requests and escalates explicitly as soon as it leaves its scope.
- Automation workflows — connecting the tools you already use so information stops being copied by hand between them.
Measure before generalising
A generative AI feature is not deterministic: the same input can produce ten different outputs. Conventional tests don't catch that, which is why the first thing built alongside any AI feature is a set of evaluations — a body of real cases with quality criteria, replayed whenever the model or the prompt changes.
Without that, quality drift is invisible until a customer sends you a screenshot of your assistant inventing a policy that doesn't exist. With it, you know before they do — and you can change model without crossing your fingers.
Your data, and where it goes
Every AI project raises the same question, and it deserves a direct answer rather than reassurance: what leaves your infrastructure, to which provider, under what retention terms. These are settled at scoping, in writing, before any development.
Self-hosting an open-weights model is a legitimate option, but it trades a usage bill for GPU infrastructure to run, secure and keep updated. For the vast majority of products the cloud API wins on cost and quality. Where a genuine constraint blocks it, the self-hosted route is the right call — and I'll say which case you're in.
What I don't take on
- AI projects without an identified use case. If we can't name the task being replaced, there is nothing to build.
- Automating processes that aren't stable. Automating a process nobody agrees on hard-codes the disagreement.
- Training models from scratch. I adapt and integrate existing models; I don't train foundation models.
- Guaranteeing a percentage gain before measurement. Anyone promising a figure before seeing your data is selling, not estimating.
Frequently asked questions
What can AI actually do in my processes?
It reliably handles tasks that are repetitive, high-volume and stable in their rules: sorting incoming email, extracting data from documents, answering from internal documentation, handling first-line support. Expect two to four weeks for a first well-built workflow, and a 30 to 70 % time saving on the targeted task.
Does my data go to OpenAI?
That depends on choices made at scoping, and they are written down before development: which data leaves your infrastructure, to which provider, and under what retention terms. Where a genuine constraint blocks the cloud, a self-hosted open-weights model is an option — at the cost of GPU infrastructure to run and secure.
What does AI cost to run?
The recurring API bill is the item almost everyone underestimates, and it grows with usage while development is paid once. It is controlled by choosing the model per task rather than defaulting to the most capable one, and by limiting the context sent with each request.
How do I know the system is working well?
Through evaluations: a body of real cases with quality criteria, replayed whenever the model or the prompt changes. A generative AI feature is not deterministic, so conventional tests aren't enough — without evaluations, quality drift stays invisible until a customer reports it.
Do we have to replace our current tools?
No, and that is usually the wrong move. The workflows plug into what you already run — email, CRM, document storage, business tools. Replacing a working tool to accommodate an automation adds a migration project to an automation project.