ExplainerAIStrategy

Growth-First AI Strategy: Five Rules That Move the Numbers

Most organisations aim AI at cost. The ones pulling ahead aim it at growth, and hold it to three numbers.

OMBIRO4 min read
Many small pilotsstage gatesFew capabilities at scaleFund AI as a portfolio: test broadly, scale what works, stop the rest

Bill Gates framed the strategic stake in 2023: “businesses will distinguish themselves by how well they use it.”1 Adoption is no longer the differentiator. In McKinsey’s 2025 survey, 88 percent of organisations used AI in at least one function, yet only 39 percent reported any enterprise-level profit effect.2 Almost everyone is using AI. Few are using it well.

The problem is aim, not adoption

Ask a room of executives what AI is for, and the answer is usually efficiency. A June 2026 Harvard Business Review analysis calls that reflex “badly misguided”: for businesses valued on sustained organic growth, growth lifts company value far more than cost cuts do.3 BCG’s 2025 study of 1,250 firms points the same way. The 5 percent creating AI value at scale achieved 1.7 times the revenue growth and 1.6 times the EBIT margins of laggards, while 60 percent saw little material value.4

The aim matters because, in many institutions, the binding constraint has moved. It is no longer the money or capacity to act. It is how quickly good intentions become delivered results.

“The constraint is no longer capital—it is intermediation.” Samaila Zubairu, President and CEO, Africa Finance Corporation, 20265

The African Development Bank’s president, Sidi Ould Tah, made a similar diagnosis at his first annual meetings in May 2026. The shortfall, he argued, is not projects or ideas but bankable structures, and only about one project in ten reaches financial close.6 The pattern is not African or financial. Any organisation with surplus capital, talent or data can ask the same question: why does intent not become value?

The question sharpens as capability rises. Anthropic’s chief executive, Dario Amodei, has summarised the powerful AI he expects as “a country of geniuses in a datacenter”.7 That is a forecast, not a fact. But if intelligence does become abundant, the scarce resource is direction: knowing which problem to point it at.

1. Find where intent stalls

Every organisation has a point where decisions slow down, queue or fail to become delivered results. Diagnose three things before choosing any tool: how long decisions take, what share of them is delivered, and where the loss concentrates. Compare commitments with delivery over several years, allowing for normal timing differences, and trace the lag stage by stage.

Exhibit 1From intent to value
FROM INTENT TO VALUE PIPELINE opportunities identified DECISION decisions committed DELIVERY decisions executed VALUE value realised START HERE Where does intent stall? TARGET 1 Time to decision TARGET 2 Conversion to delivery TARGET 3 Value captured Illustrative funnel: each target sits where value leaks. Author's framework.

Author's framework. An illustrative funnel; the three targets sit where value leaks.

2. Set three targets a board can read

Time to decision. Conversion of decisions into delivery. Value captured. Everything else is a means to these. Targets of this kind are easy to explain and easy to audit, which helps them survive a change of leadership.

3. Give every function one lever

Each executive names the one number their function moves. AI work that cannot attach to a lever does not start. That is how an AI strategy stops being a list of tools and becomes part of the operating plan.

Exhibit 2One lever per function
FUNCTION THE ONE LEVER IT MOVES THE AI USE ATTACHED TO IT Research and strategy Quality of the pipeline Sector and market scans Origination Qualified opportunities Prospect and account research Assessment and approval Time to decision Reading and checking long documents Delivery Conversion to delivery Conditions and milestone tracking Risk Quality of the portfolio Early-warning signals and stress tests Funding and reporting Cost and volume of funding Forecasts and structured outcome data No lever, no use case. Functions and levers are generic. Author's framework.

Author's framework. Functions and levers are generic and should be adapted to the organisation.

4. Sequence four kinds of AI work

Automate routine tasks now. Lead with growth work. Run risk and evidence work alongside. Transform core systems once training, change and data groundwork are ready. Hours released by productivity tools are an input to growth, not a result, so they only count once they are pointed at a lever.

Exhibit 3Four kinds of AI work
SEQUENCE OF FOUR KINDS OF AI WORK NOW ENABLERS IN PLACE 1 Routine-task automation START NOW Releases hours; easy to check 2 Growth work LEAD WITH THIS Acts on income: pipeline, decisions, delivery, funding 3 Risk and evidence BUILD ALONGSIDE Protects value and trust: early warning, evidence 4 System transformation ENABLERS FIRST Pays once training, governance and data are in place Hours released are an input to growth, not a result. The order is a judgement about value and readiness, not a measured ranking. Author's framework.

Author's framework. The order is a judgement about value and readiness, not a measured ranking.

5. Review quarterly, on the scorecard

A small steering group, a living roadmap and a quarterly executive review keep the strategy honest. The scorecard carries the three targets and little else. A strategy that cannot be read from the growth numbers is a technology list.

What the evidence does not show

Efficiency still matters, because savings can fund growth investment. Where capital is genuinely scarce, as with concessional funds limited by donor contributions, AI will not remove the constraint.

The HBR value claim rests on the authors' own modelling. The BCG and McKinsey figures are self-reported surveys. The Gates and Amodei statements are opinions and forecasts by people with a stake in AI, used here as signals of direction, not as proof.

Sources

  1. Bill Gates, "The Age of AI Has Begun", GatesNotes, 21 March 2023. gatesnotes.com
  2. McKinsey & Company, "The state of AI in 2025: Agents, innovation, and transformation", November 2025. mckinsey.com
  3. Shlomo Benartzi, Randall Long and Stefano Puntoni, "Companies Are Using AI for Efficiency. They Should Use It to Grow.", Harvard Business Review, 1 June 2026.
  4. Boston Consulting Group, "The Widening AI Value Gap: Build for the Future 2025", September 2025 (survey of 1,250 firms). bcg.com
  5. The Independent (Uganda), "Africa has the money... now it must learn how to use it", June 2026, reporting remarks by Samaila Zubairu at the Africa Finance Corporation's Africa We Build Summit and the AFC State of Africa's Infrastructure Report 2026. independent.co.ug
  6. Ecofin Agency, "Africa's New Banker-in-Chief Bets on Plumbing Over Promises", May 2026, reporting Sidi Ould Tah's opening address to the African Development Bank's 2026 Annual Meetings. ecofinagency.com
  7. Dario Amodei, "Machines of Loving Grace", October 2024. darioamodei.com

Read more on AI