The three questions it answers
A company considering artificial intelligence usually has one question in mind: which tool to buy. It is the wrong question, and the three right ones come first.
Which repetitive activities consume the most hours? Not in the abstract: in your company, this week. In most businesses the list is surprisingly similar — sorting incoming documents, answering the same questions, rekeying data from one system into another, hunting for scattered information, preparing documents that always start from the same template.
Which of those have data already usable? This is the question that separates projects delivering a result in weeks from ones that take months. A perfect use case built on information that lives on paper, or in folders with no logic, is not an AI project: it is a tidying-up project with the wrong label on it.
What has to be sorted out first? Documents, permissions, rules of use. It is the part nobody talks about and it determines success more than the technology does.
How we run it
We talk to the people doing the work, not only to those deciding. This is the step that produces the surprises: management’s perception of where time is lost is often different from that of the people executing, and the gap between the two is already a finding.
We look at where the data actually sits. Not where it should sit according to the organisation chart: where it sits. The spreadsheet somebody keeps updated, the network disk nobody has inventoried, the mailbox serving as an archive.
We assess the risks alongside the opportunity. Confidentiality, personal data, dependence on one supplier. Not at the end as a tick: at the same moment, because sometimes it is the risk that changes the priority.
We order by impact and feasibility. Two axes, not one. The case with the highest impact but the worst data is not the first one to do.
What we deliver
- The map of the processes examined, which in many companies is the first written description of how things actually work;
- The possible use cases, ordered by impact and feasibility, with a realistic estimate of timescales;
- The assessment of the data available for each, which is the part that shifts the priorities;
- The risks on confidentiality and compliance, with the measures that reduce them;
- A recommendation: which case to try first, with a starting measurement and a date for looking at the result;
- The list of what is worth putting in order regardless of AI, which is almost always useful work in itself.
Why one case, and measured
The most useful thing a company can do after the assessment is take one process to a verifiable result. Not out of caution: because it is the only way to know whether it works with your data and your people.
A well-set-up pilot has three elements: a clear boundary, a starting measurement — how long it takes today — and a date by which it gets looked at. If the number does not move, you have learned something cheaply. If it does move, you extend with experience already gained.
The alternative, buying a platform for the whole company before verifying a single result, is the fastest way of turning AI into a cost with nothing on the other side.
Sometimes the answer is “not yet”
And we say so. In some companies the processes consuming the most hours are better solved with conventional automation, with an integration between two systems that do not talk, or simply by configuring what is already there properly.
It is as useful an outcome as the other: it avoids spending that would have produced nothing, and it points to where the real work lies.
The rest
See also AI consulting for companies, AI for small business for the Italian adoption context, the company AI policy and process automation for the stage that follows.
The first step
The assessment itself begins with a conversation, without obligation, about which activities cost you the most time. The first analysis is free.