AI spend without a visible return, rivals pulling ahead and POPIA risk from staff tools: the unease owners feel is reasonable, and fixable
By Francois van der Merwe, founder of Otinga.io
You have probably sat through an AI talk or course, and little changed afterwards.
Much AI advice is written for a bank, not for forty people and a full diary.
In my view incentives are part of the reason: advisers are often paid by the hour, the seat or the follow-on project, and none of those pays them more when you need them less.
Many do excellent work. Ask any adviser, me included, two questions: is the advice built on your own numbers, and what happens after the invoice?
The return
You are paying for three AI subscriptions and could not say what any of them has returned.
I think AI earns its keep in unglamorous places, like taking cost out of work you already do.
BCG’s 2025 study of 1,250 senior executives worldwide found 60 percent report minimal gains from AI.
The 5 percent getting value at scale generate 1.7 times the revenue growth of that 60 percent (maturity is self-reported).
What to consider
Put each subscription through the four tests from my piece on AI and casino gambling: who gets the value, does it arrive every time, does it scale, and what is left after the AI bill.
Note the one number each tool should move.
The freed hours
Your people say AI saves them an hour a day, and you cannot find that hour anywhere.
BCG’s 2026 global survey of 11,749 workers found 66 percent of white-collar non-managerial staff using AI regularly report limited or no guidance on what to do with the time saved.
The fix is to measure it first, then decide whether it changes how you deliver or price your work.
What to consider
Ask three regular users what time it saves and what they did with it.
“Not sure” is the leak. If you bill by time, write down your pricing position before a client asks who keeps the saving.
Control and POPIA
You know your people are using ChatGPT, but could not say who, for what, or with whose information.
Unchecked AI use is one of the reasons owners come to us.
Bowmans partners Melissa Cogger and Talita Laubscher say staff uploading personal or customer data into public AI tools could create a POPIA compliance problem for the employer, which remains responsible for that information.
What to consider
If someone mentions pasting client figures into a public tool, leave blame aside: ask everyone which tools they use and for what, rule that client information stays out of public tools, and name an approved alternative.
Name who checks AI-assisted figures and citations before they reach a client. South Africa’s draft National AI Policy was withdrawn over fictitious sources.
Rivals and pace
You sense a rival’s proposals have become sharper and quicker.
Owners tell us they want clarity on using AI to deliver their services.
Thomson Reuters’ 2026 global survey reports 78 percent of corporate clients consider AI-enabled quality improvements very important or essential, yet just 6 percent say most of their providers deliver it.
That gap is an opening, not a verdict.
What to consider
Ask three of your best clients what they would like from you that AI makes possible.
A self-check
Underneath all four worries sits what I call the AI action gap: the course attended, the demo seen, and no job done differently on Monday.
Deloitte’s global survey of 3,235 leaders found most focus on educating employees, far fewer on re-architecting roles and workflows.
Training matters, but closing the gap starts with three lines, written in ten minutes before asking anyone.
- Where AI is used in your firm, by whom, with what data.
- The uses that pass the four tests, and the one you would cut.
- Who owns that list, and what the freed-up time is for.
If all three come easily, stop reading here.
A stalled line is not a failing; it is where to look first, and your people hold the answers.
Then pick one workflow, one owner and one number, and run it for 30 days.
If you want help gathering those answers, we run the AI Accelerator Workshop, built on our work with large South African banks and insurers and sized for a firm of forty.
After a short staff survey, an owner and one senior colleague spend a day at Henley Business School Africa in Johannesburg and leave with a map of where AI can pay, a draft AI policy and their top three business cases.
For one client, the same plan put more than 10 business cases live within 90 days and tripled the share of staff using AI.
The first step is a 30-minute call with me to see whether it suits your firm.
Start with the list of tools. The rest follows.
