TL;DR
- More AI spend does not automatically create revenue.
- Returns appear when companies fix the process underneath the tool: ownership, records, approvals, and measurement.
- The better starting question is which workflow would affect the P&L if it ran faster and cleaner this quarter.
A Deloitte survey reported by Khaleej Times this June put numbers on something I see every week: 90% of UAE business leaders say they’re increasing AI spend. Only 16% can point to revenue that came from it.
That gap is not a technology problem. The models are extraordinary. The tools are cheap. The gap is operational.
Why the spend doesn't convert
Most companies buy AI the way they buy software: pick a tool, roll it out, wait for magic. But AI is not a tool you install — it’s a worker you deploy. And like any worker, it produces nothing useful if the process it’s dropped into is undocumented, duplicated, and living in five disconnected apps and one manager’s head.
Automating a mess gives you a faster mess.
The 16% do something different
The businesses seeing returns share a pattern:
- They fixed the base first. One source of truth per record. Named owners per workflow. Data structured so a system can use it.
- They pointed AI at boring, expensive processes — quotes, reports, follow-ups — not at press-release projects.
- They kept humans on approvals. Trust builds adoption; adoption builds returns.
- They measured against a baseline. If you didn’t audit the cost of the process before, you can’t prove the saving after.
None of this is glamorous. All of it is the difference between the 90% and the 16%.
The honest starting question
Not “which AI should we buy?” but: “which process, if it ran itself, would we feel in the P&L within a quarter?”
Answer that first. The tooling decision becomes almost trivial afterwards.
That’s the whole logic behind how Avera engages — a three-day audit before anyone talks about systems. Find the leak, price the leak, then decide if it’s worth fixing. Most of the time, the audit pays for itself before the build starts.
FAQ
Why do AI projects fail to show revenue?
Most fail because the workflow underneath the AI is undocumented, duplicated, or spread across disconnected tools. AI speeds up the existing process, including the mess.
How should a company measure AI ROI?
Measure the baseline first: time spent, errors, delays, handoffs, and revenue leakage. Then compare the same workflow after automation or AI support is live.
What is the safest first AI project?
Pick a boring, expensive, repeatable process with clear human approval points, such as quote drafting, reporting, follow-ups, or document lookup.