How to choose an AI automation agency

What will be running in production four weeks after signature? A builder answers with a process name. What is the ROI arithmetic on that process — hours, loaded cost, payback month? A builder has it on one page. Who owns the automation after handover, and what does a change cost in six months? Silence here means lock-in.

Then: which of your case studies is closest to my process, and what number moved? And finally — when should I not hire you? A vendor with no honest answer to the last one has not thought about fit, only about closing.

A paid discovery phase before anything is built. Case studies with logos but no numbers. A proposal that leads with model names rather than with your process. Any refusal to quote a first project — that means the scope does not exist yet. And demos on the vendor's data rather than on yours, which prove only that the demo works.

None of these mean the vendor is bad. They mean the vendor is selling something other than working automation, and you should know which product you are buying.

Most bad outcomes are mismatches, not incompetence. An enterprise programme vendor engaged to automate one process will deliver governance you did not need at a price you did not want. A boutique engaged to run an organisation-wide transformation will be underneath it.

Size the problem honestly first: one process, a product build, or a programme. Then pick the shape of vendor that matches. We keep a side-by-side of the named vendors in this category — including where we are the wrong answer — on our comparison page.