Related question
How long should an AI assessment take?
Short answer
Weeks, not months. A single department or process can be examined in about a week. A full organization takes roughly four weeks. Anything much longer risks findings that are stale on arrival, and anything much shorter has not talked to enough people to see how the operation actually runs.
The timeline is a design decision, not a scheduling one, and it is a fair thing to ask any firm to defend.
Too short is the more common failure. A two-day engagement can only talk to the people leadership selects, which means it collects the official version of how the operation works. The gap between the official version and the actual version is precisely what a diagnostic exists to find. If nobody spoke to the person who does the work every day, the findings will confirm what the executive team already believed, which is a comfortable outcome and a useless one.
Too long has a different problem. In a market moving this fast, a six-month assessment produces conclusions built on conditions that have changed by the time anyone reads them. It also consumes exactly the attention and momentum that should be going into the first project.
Roughly a week for a single department and roughly a month for a whole organization is the range where the work is deep enough to be honest and fast enough to still be actionable. Scope, not speed, is what moves the number: how much of the operation is being examined, how many locations, and how complicated the data picture turns out to be.
Cost varies with the same factors, which is why it is worth scoping the question before pricing it. Any firm that quotes a number before understanding what it is examining is selling a package rather than a diagnosis.