Insights
How to measure the ROI of corporate AI training
The ROI of corporate AI training can be measured honestly, but only if the measurement is set up before the training runs, and only for a narrow set of things: time on specific recurring tasks, throughput on a named process, and the rate at which a team still uses what it learned sixty days later. Satisfaction scores, completion rates and self-reported hours saved are not ROI. This guide, by Ahmed Amrousy — Executive Education Instructor at the American University in Cairo and founder of Menova — sets out the four measurements worth taking and the ones worth refusing.
Last updated:
The return on corporate AI training can be measured. Most of the numbers you will be shown cannot be trusted, because they are produced after the fact by the party being paid.
Honest measurement needs four things decided before anyone books a date: which recurring tasks you are watching, how long they take now, who owns the measurement, and when you will look again. Nothing after that is complicated. Nothing before it can be recovered later.
What can honestly be measured
Time on named recurring tasks. Pick two or three things the team does every week — a weekly report, a tender summary, a first-draft proposal, a budget variance commentary. Record how long each takes today and who does it. These are measurable because they are specific, repeated, and owned by someone who can tell you the truth about them.
Throughput on one process. Not “marketing productivity”. One process: campaign briefs produced per month, support tickets resolved without escalation, proposals out of the door. One number, one owner, one definition that does not move.
Sustained use at sixty days. The single most informative measurement in corporate training, and almost nobody takes it. Ask every participant one question two months later: did you use any of this in your actual work last week? The proportion who say yes is the real result of the programme. Enthusiasm at the end of a workshop is free; use at sixty days is not.
Quality against a standard you already have. If you have a review step — a manager who checks proposals, a compliance read, an editor — ask whether the rework rate moved. You already have the instrument. You do not need a new one.
What cannot honestly be measured
A company-wide productivity percentage. There is no clean way to isolate the effect of two days of training from hiring, seasonality, a reorganisation, a new system or a good quarter. Any provider quoting you one is estimating, and the estimate was built to be quoted.
Self-reported hours saved. People overestimate time saved on tasks they now enjoy more. Collected after a good workshop, these numbers are a mood reading.
Satisfaction scores. They measure whether the day was well run. That is worth knowing — it is simply not the same thing. A programme can average 4.8 out of 5 and change nothing, because participants rate the experience before anything has been asked of them. A high score earns you the right to ask for the sixty-day number. It is not a substitute for it.
Completion rates. Completion measures attendance.
Set it up before you book
Four decisions, half an hour, before the agenda is written:
| Decision | Example |
|---|---|
| Which tasks are we watching? | The weekly commercial report and the first draft of client proposals |
| What is the baseline? | 6 hours and 4 hours per week respectively, done by two named people |
| Who owns the measurement? | The department head — not the trainer |
| When do we look? | 30 days, 60 days (adoption question), 90 days |
Hand those tasks to the provider as the design brief. A programme built around the two things you have agreed to measure is a different programme from one built around a generic agenda — and if the provider cannot build around them, you have learned something useful before spending anything.
The arithmetic nobody does
Before comparing fees, price the room. Twenty managers away from their work for two days is a large internal cost in salary alone, and you incur it whether or not the training works. Against that, the difference between two providers’ fees is usually small.
This cuts both ways. It is an argument against haggling over the fee, and it is an argument for being ruthless about whether anything will change — because the expensive part of bad training is never the invoice.
“Why not just buy online course seats?”
A fair question, and for part of the problem the answer is yes. A self-paced library is the right tool for broad individual upskilling, and it costs a fraction as much per head.
It is the wrong tool for three things. It contains nothing about your processes, your constraints or your data. It asks for no decisions, so no decisions get made. And participation is optional, which in practice means it is partial. If the objective is “our managers should know what these tools do”, buy the library. If the objective is “our management team should agree what we are going to change”, that requires facilitated time, and no amount of video substitutes for it.
Most organisations need both and buy only one.
What a good provider will tell you
That the effect is not uniform. Some departments will change how they work within a fortnight; others will not move at all, usually because their work is less document-heavy or because their head of function was not in the room. A provider who promises an even gain across the organisation is selling you a number, not a programme.
The two measurements worth insisting on — baseline task times, and the sixty-day adoption question — cost nothing and settle the argument either way. Ask for them in writing, and keep the measuring yourself.
If the question you are still weighing is what the programme should cost and how it is quoted, that is answered in what corporate AI training costs in Egypt and the Gulf. If you are choosing between providers, start with nine questions to ask before booking.
Ahmed Amrousy is an Egyptian AI-for-Business Expert, Executive Education Instructor at the American University in Cairo’s Onsi Sawiris School of Business, Head of Marketing & PR at Nile Air, and the founder of Menova.