Methodology

1. Intent and audience

AIRI is built for boards and executive teams that want a fast, structured and confidential read of their capacity to capture value from AI. The diagnostic is not a substitute for an audit or for professional, legal, financial or investment advice: it is intended to support strategic reflection.

2. The six dimensions

AI maturity is read across six complementary dimensions. Each carries four calibrated questions, for 24 questions in total. The choice of dimensions draws on recognized frameworks (McKinsey, MIT CISR, Gartner, Oxford Insights, World Bank / AfDB / ITU work) and has been adapted to African and emerging-market contexts.

3. Scales and scoring

Each question uses a 0–3 anchored scale (0 = absent, 1 = nascent, 2 = in place, 3 = mature). A dimension score is the mean of its four items, rebased to 100. The overall score is the mean of the six dimensions. The maturity level (1–5) follows fixed 20-point bands: 0–19, 20–39, 40–59, 60–79, 80–100.

4. Configuration coherence engine

Beyond scores, AIRI classifies the organisation’s configuration into one of six archetypes (balanced, exposed base, tools without adoption, scaling without control, governed but inert, ambition ahead of capability). The classification is deterministic, based on dimension spread and the weakest-link dimension.

• Library v1.1.1, validated by invariant tests on 106,496 profiles.
• The engine never selects a tool, vendor or investment: it diagnoses organisational configuration.

5. Regional comparisons (visibility threshold)

Regional comparisons are shown only when a sufficient self-reported sample exists (n ≥ 30 per sector, geography or size segment). Below the threshold, no number is displayed; a neutral message invites the user to return later. Benchmarks do not rely on any external dataset and are labelled as such.

6. Low-effort response detection

A simple heuristic detects very low-engagement response patterns (abnormally short duration, uniform answers). Affected profiles are not removed from the report, but are excluded from comparison cohorts to preserve statistical quality.

7. Privacy and data residence

Responses are stored encrypted on a database hosted in the EU (west region). Generation functions run at the edge. No personally identifiable information is transmitted to the language model used for narration: only numeric scores and dimension labels are sent.

8. Legal notices and non-affiliation

The methodology is the work of Stredge Partners and draws on the published academic literature on complementarities and configurational fit. The frameworks cited (McKinsey, MIT CISR, Gartner, Oxford Insights, World Bank, African Development Bank, International Telecommunication Union) are referenced for context only; Stredge Partners is neither affiliated with nor endorsed by these organisations.

“Stredge Partners” and the SP mark are trademarks of Stredge Partners. © Stredge Partners.

9. References

Milgrom and Roberts (1995), “Complementarities and Fit”, Journal of Accounting and Economics.

Kremer (1993), “The O-Ring Theory of Economic Development”, Quarterly Journal of Economics.

Brynjolfsson, Hitt and Yang (2002), “Intangible Assets: Computers and Organizational Capital”, Brookings Papers.

Siggelkow (2001), “Change in the Presence of Fit”, Academy of Management Journal.

Teece (2007), “Explicating Dynamic Capabilities”, Strategic Management Journal.

Dawes (1979), “The Robust Beauty of Improper Linear Models in Decision Making”, American Psychologist.

Smith and Kendall (1963), “Retranslation of Expectations”, Journal of Applied Psychology.