Contents
KPMG published its 2026 Africa CEO Outlook on 22 October. Most of the coverage has fastened on the readiness gap: 72 per cent of African chief executives call artificial intelligence a top investment priority, against 79 per cent globally, but only 14 per cent believe their organisations are keeping pace with or ahead of AI developments, against 66 per cent globally.
That gap is real, and it is not the most actionable number in the report. This one is: 42 per cent of the African CEOs surveyed expect their AI investments to pay back beyond three years, against 12 per cent globally.
A payback horizon is not a measure of enthusiasm. It is a financing decision with a price — and these are the longest horizons in the survey, held in the markets where long horizons cost the most.
What is happening?
The 2026 Africa CEO Outlook is the third edition of KPMG’s Africa report and part of the twelfth edition of its Global CEO Outlook: 1,400 CEOs globally, including 176 African CEOs, with fieldwork between 11 August and 21 September 2026. The Africa findings were released in Lagos on 22 October 2026.
| Measure | Africa | Global |
|---|---|---|
| AI a top investment priority | 72% | 79% |
| Keeping pace with or ahead of AI | 14% | 66% |
| Payback expected within 1–3 years | 48% | 71% |
| Payback expected beyond 3 years | 42% | 12% |
The regional cuts diverge sharply. On the first two measures, West Africa (n=30) reports 63 and 10 per cent, East Africa (n=50) 74 and 12, Southern Africa (n=61, the South Africa survey cut) 79 and 20, and North Africa (n=18) 56 and zero: none of the CEOs surveyed there considered their organisation to be keeping pace.
Three limits should be stated. These are expectations and self-assessments, not measured returns: KPMG reports no realised return on AI investment anywhere in the Africa report, and “keeping pace” is scored against a benchmark the survey does not define. The regional samples are small, and KPMG’s own note says they should be read directionally — its Technology & AI extract records that the North Africa segment is labelled “French Africa” in the source data, 16 of its 20 respondents being North African and four Francophone sub-Saharan. And these are chief executives of larger organisations, not the median business in any market.
Why it matters economically
Set the payback finding against what money costs, using figures the central banks publish. In Kenya, the Central Bank held the Central Bank Rate at 8.75 per cent on 7 October 2026 and reported average commercial bank lending rates of 14.4 per cent in September, against 17.2 per cent in November 2024. In Ghana, the Bank of Ghana held the policy rate at 14.0 per cent on 24 September 2026 and reported the banking sector’s average lending rate down to 15.9 per cent in August, from 24.2 per cent a year earlier. Both are falling fast — and both remain several times the cost of capital where most of the global sample sits.
Take an illustrative project costing 100 today and returning an equal amount each year, discounted at 15 per cent — between the two averages above. Over four years it must return about 140 in total, undiscounted, simply to break even; at 5 per cent, about 113. The project has not changed; the money has. The assumptions — level annual benefits, discounted at the cost of debt, no tax or inflation adjustment — are yours to replace. A four-year horizon is not a slower two-year horizon; it is a materially larger commitment.
KPMG offers a fair alternative reading: longer payback expectations may reflect the time required to modernise systems, prepare data and build governance rather than weak conviction. That is probably right about the cause, and it does not change the consequence. Work that genuinely takes four years needs funding that tolerates four years — retained earnings, equity or a committed programme budget — not a facility priced at 15 per cent and repayable on demand.
The combination to watch is a small budget with a long horizon. That is what gets cancelled at the first cash squeeze, because it produces no interim number with which to defend itself.
What it means for African businesses
The useful thing in the regional data is not the ranking, but that the first blocking item differs by region.
In West Africa the constraint is data: data readiness and quality is cited by 73 per cent, against 55 across Africa. The region also reports the highest operational impact — 63 per cent say agentic AI is already having a significant or transformational impact, against 48 across Africa — while only 10 per cent are confident of keeping pace. Reported impact and shaky foundations are coexisting.
In North Africa it is governance: ethical considerations and regulatory disparities lead the challenges at 69 and 63 per cent, and 75 per cent say AI and data sovereignty are sharpening their focus on internal data control. Spending intent is nonetheless concentrated — 22 per cent plan to put more than a fifth of budget into AI, against 8 per cent globally.
In Southern Africa it is the operating model: 67 per cent report redeploying employees into AI-enabled roles and 57 per cent are redesigning roles and career paths.
In East Africa it is cost, and the horizon problem is sharpest there: 78 per cent expect to allocate no more than 20 per cent of budget to AI over the next 12 months, while 42 per cent anticipate returns within three to five years, against 30 per cent across Africa. Cost and data readiness lead the challenges at 60 per cent each, and the expected benefits read as operational rather than revenue-led — efficiency and productivity at 64 per cent, threat detection at 56.
A single continental AI plan will therefore be wrong in most of the places it is applied, which is the point we make about operating across African markets generally. Nor does the finding apply to every business: a company that cannot state what one of its processes costs is not choosing between horizons at all, but is at the stage we described when the World Bank measured the gap between adopting AI and integrating it.
Technology and AI implications
Be clear about what technology does not fix, because the payback horizon is mostly made of things no model shortens. Cleaning data, agreeing how a process actually runs, defining who approves an exception and training the people who will live with the result are calendar-and-people costs. Nor does technology turn released time into cash: that happens only if the hours are re-sold, more volume is handled by the same team, or a hiring plan changes. And a vendor’s return-on-investment calculator is a sales document; it assumes full utilisation of released time and a baseline the buyer has never measured.
What technology does earn here is metering. Instrument the process — enquiry logged with a timestamp, quote issued with a timestamp, invoice matched by the system rather than by eye — and the baseline and the post-deployment figure are both machine-recorded, so a payback claim can be tested rather than argued. If you cannot record the baseline today, the first increment is the recording, not the model: unglamorous, cheap, and what makes everything after it fundable.
On agentic AI, the West African figures are a caution: high reported impact alongside the weakest data readiness scores suggests most of it is task-level. Real, but not a payback claim until it has been metered.
Practical business implications
Six questions turn this finding into a decision.
- What does money actually cost you? Not the policy rate — the rate on your own overdraft or term facility, or the return you forgo by spending cash. That is the hurdle rate for every AI proposal.
- Can the programme be split into increments that each pay back inside one budget cycle? If not, you are funding a platform rather than buying AI, and it should be approved as one.
- Is the platform work paid for out of the increments? Attach data cleanup, integrations and the system of record to the increments that need them, so they accrete as a by-product instead of a three-year bet.
- Which single number tells you it worked, and what is it today? Quote turnaround, enquiries handled per person, days sales outstanding, reconciliation exceptions per week. Record the baseline before deployment; afterwards you cannot.
- What happens to the released time? Decide in advance and in writing — unassigned released time is absorbed, not banked.
- What horizon is the vendor quoting, and will they put it in the contract? A supplier confident of a twelve-month payback will agree to measure it.
RevenueStack perspective
Our reading, through Strategic Economics × Technology × AI, is that this is a financing mismatch rather than an ambition gap — and a largely imported one.
The economics first. The programme shapes being sold — multi-year platform builds, enterprise data foundations laid before any use case earns anything — were designed where capital costs a fraction of what it costs in Accra or Nairobi and where patient equity funds the gap. Transplanted to a 15 per cent hurdle rate, the same shape does not merely cost more; it often fails outright. The 42 per cent figure is what that transplant looks like in a survey: an investment horizon built for someone else’s balance sheet.
The technology next. The design response is to change the shape of the project, not the ambition: sequence the work so each increment clears your own hurdle rate on its own, and let the platform — records, integrations, governance — accrete underneath as the increments require it. That is slower to describe and faster to fund, and it is the logic behind how we sequence engagements and scope AI and automation work.
And on the 14 per cent: a benchmark nobody defines is not a target. The one that matters is internal — what your processes cost today, and what your money costs today. A business that can answer neither will keep buying horizons it cannot finance, and conclude wrongly that the technology was the problem.
Conclusion
The most quoted number from KPMG’s 2026 Africa CEO Outlook is the readiness gap. The more consequential one is the payback gap: 42 per cent of African CEOs expect AI to pay back beyond three years, against 12 per cent globally, in markets where the central banks of Kenya and Ghana put average bank lending at 14.4 and 15.9 per cent. Long horizons are a legitimate choice when the funding matches them and an expensive error when it does not. The useful move is not to abandon the ambition but to restate it as increments, each with a measured baseline and each paying for itself inside the budget year.
Sources
- KPMG 2026 Africa CEO Outlook — KPMG, 2026-10-22
- 2026 Africa CEO Outlook — Technology & AI — KPMG, 2026-10-22
- KPMG Africa CEO Outlook 2026 (press release) — KPMG, 2026-10-22
- Press Release — Monetary Policy Committee Meeting of October 7, 2026 — Central Bank of Kenya, 2026-10-07
- Monetary Policy Committee Press Release, September 2026 — Bank of Ghana, 2026-09-24