The business case for an AI initiative is almost always built the same way: number of licences times price times twelve months. That is a clean, verifiable figure, and it routinely describes less than half of the actual cost.
This is not an argument against the investment. It is an argument for putting the full sum on the table before anyone sets an expectation about the return.
The three invisible items
Preparing the foundations. Cleaning up permissions, organising storage, weeding out outdated documents, documenting data sources. At mid-size that is twenty to sixty person-days, and the work is unrewarding because it produces no visible progress. It is nonetheless the precondition for results being usable.
Enablement. Not the one-off training course, but the guided working sessions, the internal multipliers and the period during which people are slower because they are learning something new. Budget a few hours per person over several months, plus the time of the multipliers.
Operation and upkeep. Use cases go stale, agents need adjustment, permissions drift apart again, someone has to sample the quality. That is a permanent task, not a project end date.
Count only the licence and you have planned a third of the initiative, then wonder about the other two thirds.
The benefit side is just as incomplete
On the other side, the calculation is usually made in minutes saved: twenty minutes per person per day, times the number of people, times an hourly rate. That calculation convinces nobody who owns a budget, because saved minutes only turn into money if the freed-up time is genuinely used differently.
More robust are the measures the business already tracks. Turnaround time for a quotation. Handling time for a customer service case. Share of enquiries resolved at first contact. Time to onboard a new employee. Those numbers already exist, and improvements in them cannot be argued away.
A realistic frame
For a first serious initiative in a mid-sized company, plan on a ratio of roughly one to two: for every franc of licence cost, about two francs for preparation, enablement and operation in the first year. From the second year the ratio shifts markedly in favour of the licence.
That number is a rule of thumb, not a calculation. Its purpose is to move the executive conversation from what does the licence cost to what does it cost to make it work.
The advantage of an honest calculation
Initiatives that start with a complete cost picture are cancelled less often. Not because they are cheaper, but because nobody is surprised by unexpected effort after six months. Surprise is the most common reason initiatives lose their support in the leadership team.