Knowledge bank · Kilwa Research

Latest “Best” AI Strategies

Nine of the most-cited AI strategy playbooks, from Stanford, McKinsey, BCG, Bain, Anthropic, Microsoft, Google and MIT Technology Review, read, distilled to what an institution can act on, and reconciled into one view. Kilwa builds its thinking and its platform on the best ideas from the best sources. This is where we show the work.

Read this if you sit on an investment committee weighing AI exposure across a portfolio, a board approving a transformation budget, or a ministry or regulator scoping a sovereign AI or public data programme.

One view of nine

Where the best thinking converges.

Nine documents, four consultancies and platform companies, two research labs and a business publisher, published between 2018 and 2026. Read together, they agree far more than they argue. Eight themes recur; three of them appear in every source that addresses the question.

Themes by source
Theme	Stanford	McKinsey QB	Anthropic	McKinsey Q	BCG	Microsoft	Google	Bain	MIT TR	Count
Leadership owns it	yes	no	yes	yes	yes	yes	yes	yes	no	7
Redesign the work	yes	no	yes	yes	yes	yes	no	yes	no	6
Data foundations first	yes	yes	yes	yes	yes	yes	yes	yes	yes	9
Trust is the licence	yes	no	yes	yes	yes	yes	yes	no	yes	7
Adoption over tooling	yes	yes	yes	yes	yes	yes	yes	yes	yes	9
Value, focused and measured	yes	yes	yes	yes	yes	yes	yes	yes	no	8
Built to scale past pilots	yes	yes	yes	yes	yes	yes	yes	yes	yes	9
Agentic AI is next	yes	no	yes	yes	yes	yes	no	no	no	5

Kilwa reading of the nine sources listed in the footnotes. A dot means the source makes the point in its own text; a blank means it does not address it, not that it disagrees. Two of the nine predate the agentic wave (McKinsey's 2018 playbook, Google's 2020 framework), which is why the last row is thinner.

The model is the cheapest part.

Every recent source puts most of the effort, and most of the failure, in people, process and data. For a committee approving an AI budget the question is not which model, but who owns the workflow, what data exists and how the change will be managed.

Trust is a precondition, not a phase.

McKinsey’s no trust, no right to deploy; Microsoft’s governance-first foundation; MIT’s 98% who would forgo being first. The sources agree that controls come before scale, and that the institutions treating governance as the programme move faster later.

Scale is rare, so the gap is the prize.

Four percent creating substantial value (BCG), 76% stuck at one to three use cases (MIT), one deployment in five agentic (Stanford). A disciplined institution in an African market is not behind. The field has barely started.

The evidence, redrawn

The hard part is not the model.

Two findings from two different methods land in the same place: most of the work, and most of the risk, sits in people, process and data. And where the work can be made autonomous safely, the return roughly doubles.

Two different measures, one shape. BCG's rule is where a transformation's effort should go; Stanford's figure is where practitioners said the hardest challenges actually sat across 51 deployments. Sources 5 and 1 in the footnotes; chart drawn by Kilwa.

Median productivity gain
Share of the 51 cases

Human in the loop: A person reviews or approves each output before action
High automation: AI handles more than 80% of the work; humans review exceptions
Agentic: AI takes autonomous, multi-step actions end to end

Stanford Digital Economy Lab, 51 deployments interviewed August 2025 to February 2026 (source 1). The authors note the gap partly reflects the tasks each model was given, and that regulated, high-stakes decisions need human review by design. Chart drawn by Kilwa.

Six numbers, six sources

Most organisations are still stuck before scale.

The sources measure different things, so the tiles are not one series. Read them as six independent sightings of the same gap: nearly everyone is using AI, very few have turned it into measurable value at enterprise scale. That gap is the opportunity a disciplined institution in an African market can take.

of companies are creating substantial value from AI; about a quarter have moved beyond proofs of concept.

BCG, December 2024 · source 5

76%

of surveyed companies had deployed AI in only one to three use cases.

MIT Technology Review Insights, 2024 · source 9

70%

of generative AI pilots stall on data practices and governance, per Gartner.

Cited by Microsoft, May 2025 · source 6

61%

of successful deployments followed at least one failed attempt.

Stanford Digital Economy Lab, 2026 · source 1

of companies believe their generative AI investments have reached full maturity, against 92% planning to invest.

Cited by Anthropic, October 2025 · source 3

20%

of the 51 successful deployments were agentic, the operating model with the highest median gain.

Stanford Digital Economy Lab, 2026 · source 1

The nine playbooks

Each source, distilled.

For every playbook: the thesis in one sentence, the insights that survive contact with a real budget, the number worth remembering, and what it means for an institution operating in African markets.

Stanford Digital Economy Lab

The Enterprise AI Playbook: Lessons from 51 Successful Deployments

Research report, 116 pages · April 2026 · Elisa Pereira, Alvin Wang Graylin and Erik Brynjolfsson

77%

of the hardest challenges in 51 deployments were invisible costs: change management, data quality and process redesign, not the technology.

Read the source

Footnote 1 · Leadership owns it · Redesign the work · Data foundations first · Trust is the licence · Adoption over tooling · Value, focused and measured · Built to scale past pilots · Agentic AI is next

Same models, same use cases, outcomes measured in weeks at one company and years at another. The difference was never the AI. It was the organisation.

Budget for a failed first attempt. 61% of the successful projects followed at least one earlier failure whose cost never appears in the final return. First attempts failed when AI was applied to broken workflows, or led by technical teams without business ownership.
Sponsorship is a verb. Effective sponsors cleared blockers weekly, bridged business and technical teams and tied adoption to corporate objectives. The seven organisation-wide transformations in the sample all reached that level of integration; approving a budget was not enough.
Resistance comes from staff functions first. Legal, HR, risk and compliance were the most frequent source of resistance, in 35% of cases, ahead of end users at 23%. Mandates tied to objectives moved them; persuasion did not.
Autonomy paid where errors were recoverable. Escalation models, where AI handles 80% or more of the work and humans review exceptions, showed 71% median productivity gains against 30% where every output needed approval. The authors note the tasks differed, and that regulated decisions need review by design.
The model is usually a commodity. In 42% of implementations model choice was fully interchangeable; the durable advantage sat in the orchestration layer, the data and the process. Messy data was rarely a blocker: language models unlocked previously unusable data in 88% of cases.

Why projects failed before they succeeded, share of cases

The organisation was not ready to adopt
35%
Critical knowledge was never captured or stored
27%
Legal or compliance blocked the project
18%
Technology broke or was not mature enough
16%
Wrong problem chosen or expectations unrealistic
14%
Talent or sponsorship gap
12%

Stanford Digital Economy Lab, root causes consolidated across the 51 cases (appendix). Percentages are shares of cases citing each cause; chart drawn by Kilwa.

The Kilwa read

For a board or investment committee in Lagos, Nairobi or Cairo this is a budgeting fact, not a caution: the second attempt is the one that works, and the money goes to process documentation, data access and change management before it goes to models. Kilwa's engagements start there, with the decision the intelligence will change, who owns it and what data exists. The platform is built on the same premise: any single model is replaceable, and the provenance-flagged data layer is the asset.

McKinsey & Company, QuantumBlack

The executive’s AI playbook

Interactive playbook · November 2018

industries sized for the annual value of AI and analytics in the Value & Assess section, built on McKinsey Global Institute’s study of more than 400 use cases.

Read the source

Footnote 2 · Data foundations first · Adoption over tooling · Value, focused and measured · Built to scale past pilots

Size the prize in your industry first, then execute in a way that gets past pilot purgatory, and know the red flags before you start.

Value before technology. The first section lets an executive size the annual value of AI and analytics for their own industry and see which techniques and functions carry it. The underlying MGI work put the potential of AI techniques at $3.5 trillion to $5.8 trillion a year across nine business functions in 19 industries.
Execution is three moves. Align on strategy; build the technology, data and people capabilities; complete the last mile, where a model becomes a decision that someone actually makes differently.
Beware the ten red flags. The closing section lists ten warning signs that an AI programme is failing, each paired with a response, on the premise that pilot purgatory is a management failure that can be anticipated rather than a technical one.
It has aged well. Written before generative AI, its sequence, value first, capability second, adoption third, is the same one the 2025 and 2026 sources on this page still use.

The playbook’s three parts

Value & Assess
Size the annual value for your industry; see which techniques and functions carry it.
Execute
Align strategy, build technology, data and people capability, finish the last mile.
Beware
Ten red flags, each with a response.

Structure as published on mckinsey.com; McKinsey Global Institute, Notes from the AI frontier, April 2018, for the value sizing.

The Kilwa read

Sizing is the step most African deployments skip, because the market data to do it does not exist off the shelf. That gap is what Kilwa’s country and sector intelligence is for: the annual value of a use case in Kenyan insurance or Moroccan manufacturing has to be built from primary evidence, not read off a global table. And the last mile is longer here, because the users and the data often sit in different languages and different systems.

Anthropic

The Enterprise AI Transformation Guide

Guide, 19 pages · October 2025

8–12 weeks

is how long most successful pilots run. If there are no results by week twelve, change the use case, not the timeline.

Read the source

Footnote 3 · Leadership owns it · Redesign the work · Data foundations first · Trust is the licence · Adoption over tooling · Value, focused and measured · Built to scale past pilots · Agentic AI is next

Lay the foundation, launch a pilot, scale the impact: a three-step method drawn from Anthropic’s customers and its own teams, with the numbers needed to run it.

Alignment is a committee, not a memo. A steering group with a C-suite sponsor, functional leaders, technology, finance and legal, plus champions at every level, including the sceptics whose questions surface the real implementation risks.
Agree the metrics before the pilot. Four dimensions: adoption (daily active users, feature use), efficiency (time saved, throughput), quality (accuracy thresholds, error rates) and satisfaction. Expect onboarding in weeks two to three, measurable efficiency gains by weeks four to six, and clear adoption patterns by weeks eight to ten.
Pick pilots where failure is cheap. One or two projects across functions, judged on return and feasibility, never customer-facing or mission-critical first. Most show meaningful results within 30 to 60 days.
Governance early, not late. Access controls, usage guidelines, quality standards and compliance protocols from the start. The guide opens with the finding that 92% of companies plan to invest in generative AI over three years while 1% believe their investments have reached full maturity.
Scale through people. Role-specific training, certification that counts in promotion decisions, and centres of excellence that rotate business experts through for three to six months.

Anatomy of a successful pilot

Wk 1–3Onboarding and adjustment as users learn the tool
Wk 4–6Measurable efficiency gains emerge
Wk 8–10Quality improvements and adoption patterns become clear
Wk 12Decide: scale, change the use case, or stop

Anthropic, The Enterprise AI Transformation Guide, step one. Timings are the guide’s typical ranges; figure drawn by Kilwa.

The Kilwa read

This is the closest thing to an operating manual for a ministry or a bank starting from zero, and it maps onto Kilwa Sovereign AI almost line for line: a readiness assessment, one narrow pilot with metrics agreed up front, then capability transfer so the partner can run and extend the system without us. The guide’s eight-dimension readiness matrix, from executive commitment to budget, is a useful self-test before any public-sector pilot is scoped.

McKinsey & Company

The AI transformation manifesto

McKinsey Quarterly, twelve themes · April 2026 · Alex Singla, Alexander Sukharevsky, Eric Lamarre, Kate Smaje and Robert Levin

$3 : $1

incremental EBITDA for every dollar invested, reported for a group of twenty leading companies, with breakeven in one to two years.

Read the source

Footnote 4 · Leadership owns it · Redesign the work · Data foundations first · Trust is the licence · Adoption over tooling · Value, focused and measured · Built to scale past pilots · Agentic AI is next

Technology alone does not create advantage; enduring capabilities do. Twelve themes separate the companies rewired for AI from those still layering it on.

Advantage is capability, not tooling. Everyone can buy the same models. The durable edge is repeatable capability in six areas: strategic road-mapping, talent, operating model, technology, data, and adoption and scaling.
Go where the economic leverage is. Focus deeply on a small number of leverage points. If the value does not move the business, treat the programme as wrong, not the metric.
Senior business leaders in the driver’s seat. Building the technology and AI fluency of the top team is a first-order priority, because every AI transformation is ultimately a people transformation.
Speed is the organisational advantage. The metabolic rate of the company, how fast it redeploys people and money and how short the path is from insight to action, is a measurable property, and the leaders treat platforms and data as products with owners and budgets.
No trust, no right to deploy. Risk, testing and controls are prerequisites. The next capability to master is agentic engineering: automating the guardrails and controls as deliberately as the work itself.

The twelve themes, in Kilwa’s words

Enduring capabilities beat tooling
Focus on economic leverage points
Value must move the business
Senior business leaders drive it
Every transformation is a people transformation
Speed is the defining advantage
Platforms are strategic assets
Make data easy to consume and enrich
Design for adoption, build for scale
No trust, no right to deploy
Agentic engineering is the next capability
Keep learning as the end goal

Kilwa paraphrase of the twelve theme headings. McKinsey Quarterly, 7 April 2026, an excerpt from Rewired, second edition.

The Kilwa read

Theme ten is Kilwa’s operating rule: every published score ships with its inputs, provenance flags and a model card, because trust is the licence to be used at all. Theme eight, data as a product that is easy to consume and enrich, is the design brief for the Kilwa Factbook. For an allocator the useful test is the third theme: if an AI programme in a portfolio company cannot show value that moves the business, it is a cost line, whatever the demo looked like.

Boston Consulting Group

The Leader’s Guide to Transforming with AI

Guide with function-by-function playbooks · December 2024

10-20-70

of the effort in an AI transformation: 10% algorithms, 20% technology and data, 70% people and processes.

Read the source

Footnote 5 · Leadership owns it · Redesign the work · Data foundations first · Trust is the licence · Adoption over tooling · Value, focused and measured · Built to scale past pilots · Agentic AI is next

Deploy, reshape, invent: three plays in rising order of ambition, and a 10-20-70 rule for where the effort goes.

Three plays, rising ambition. Deploy the technology for immediate productivity; reshape critical functions such as finance, operations, HR and IT, where the guide cites efficiency gains of up to 50%; invent new products, services and revenue streams that competitors cannot copy.
Few win yet. BCG’s research found only about a quarter of companies had moved beyond proofs of concept to generate real value, and only 4% were creating substantial value.
Leaders pull away. The companies BCG classes as AI leaders reported around 50% higher revenue growth and 60% higher total shareholder return than their peers.
Concentrate, then transform the core. Fewer, higher-priority use cases; AI applied to the functions that define the business, research in pharma, underwriting in insurance, not only the support functions; predictive and generative AI combined.
People and guardrails carry the weight. Behaviour change, training and organisational alignment take most of the effort, and responsible-AI guardrails are what earn the trust to scale.

Three plays

Deploy
Immediate productivity from tools and automation.
Reshape
Rewire a critical function end to end.
Invent
New products, services and revenue streams.

BCG, The Leader’s Guide to Transforming with AI, December 2024; figures as published there.

The Kilwa read

10-20-70 is the budget shape Kilwa puts in front of every institution building AI capability, and it is why our sovereign AI work funds capacity transfer rather than licences. The three plays give an African corporate a sequence it can finance: deploy for productivity now, reshape one function, then invent the products that incumbents in London or Dubai cannot, because the data and the customer are local.

Microsoft

Rewriting the IT Playbook: Empowering CIOs to Lead with Confidence in the AI Era

Microsoft 365 Copilot blog introducing The Strategic CIO’s Generative AI Playbook · May 2025

70%

of generative AI pilots do not move past pilot because of inadequate data practices and governance, a Gartner finding Microsoft puts at the front of the playbook.

Read the source

Footnote 6 · Leadership owns it · Redesign the work · Data foundations first · Trust is the licence · Adoption over tooling · Value, focused and measured · Built to scale past pilots · Agentic AI is next

The CIO has moved from running infrastructure to leading transformation: secure the foundation, partner across the business, empower employees, and measure.

Adoption is fast; impact is not automatic. 24% of leaders say their organisation has deployed AI company-wide and only 12% are still in pilot mode, per Microsoft’s 2025 Work Trend Index. Rapid adoption alone does not deliver lasting business impact.
Governance first. Data quality and compliance, oversharing and insider risk, and a culture of responsible use are the foundation. Without them, pilots fail.
The business must be in the room. 27% of chief data and analytics officers name a lack of business-stakeholder involvement as their biggest challenge, per Gartner. The answer is anchoring every initiative to a business goal owned jointly with HR, finance, marketing, legal, sales and operations.
Tools are not skills. 69% of CIOs plan to upskill employees on AI but only 15% feel their workforce is fully prepared, per Gartner. Targeted skilling paths and safe places to experiment close the gap.
Measure in three dimensions. Readiness, adoption and impact, then scale with agents that act autonomously inside workflows, the pattern Microsoft calls the Frontier Firm.

The gaps the playbook targets

Generative AI pilots that stall on data and governance
70%
CIOs planning to upskill employees on AI
69%
Leaders reporting company-wide AI deployment
24%
CIOs who feel their workforce is fully prepared
15%
Organisations still in pilot mode
12%

Figures as cited by Microsoft: Gartner CIO Report (70%, 69%, 15%) and Microsoft Work Trend Index 2025 (24%, 12%). Different surveys and bases; read as separate sightings, not one series. Chart drawn by Kilwa.

The Kilwa read

Readiness, adoption, impact is how Kilwa reports the results of an engagement, and governance-first is how it designs public data systems: residency, access and audit are day-one decisions, because the failure Gartner counts is a governance failure. For a CIO in a Nairobi bank or a Kigali ministry the implication is blunt: the data-governance work is the AI programme, and the tools come after.

Google Cloud

Google Cloud’s AI Adoption Framework

Whitepaper, 37 pages · June 2020

6 × 3

the AI Maturity Scale: six themes, Learn, Lead, Access, Scale, Secure and Automate, each scored on a tactical, strategic or transformational phase.

Read the source

Footnote 7 · Leadership owns it · Data foundations first · Trust is the licence · Adoption over tooling · Value, focused and measured · Built to scale past pilots

Six themes across people, process, technology and data, each placed on three phases, tactical, strategic and transformational, give an organisation a map of where it is and what comes next.

Maturity is measurable. Tactical means narrow, short-term use cases with no coherent plan; strategic means several machine-learning systems sustained in production under a broader vision; transformational means AI diffused across lines of business with a mechanism to keep scaling it.
Learn and Lead come first. Skills, hiring and partners on one side. On the other, a leadership mandate, executive sponsorship, dedicated budget and teams organised around business goals rather than heroic individual projects.
Access is the data question. Whether datasets are curated, owned, discoverable and reusable, and whether trained models and features are shared across the organisation rather than rebuilt.
Secure and Automate make it repeatable. Data protection, responsible and explainable AI, and pipelines that deploy, monitor and retrain reliably.
The evidence it cites. A Google Cloud study with MIT Technology Review found organisations adopting machine learning made twice as many data-driven decisions, decided five times faster and executed three times faster.

The AI Maturity Scale, condensed

Phase	What it looks like	The leap to the next phase
Tactical	Narrow, short-term use cases; complex problems outsourced; no plan to scale.	A leadership mandate, a data lake and the first production ML systems.
Strategic	Several ML systems in production; a central advanced-analytics team; a shared data model.	Data science embedded in each line of business; feature stores and shared assets.
Transformational	AI diffused across the business; continuous experimentation; research and innovation teams.	Keep alive the mechanism that scales and promotes ML capability.

Kilwa condensation of the whitepaper’s phase descriptions and maturity scale.

The Kilwa read

A maturity scale is a map, and Kilwa uses the same three-phase logic to place a ministry or a bank before scoping work. Most African institutions are tactical, and the leap to strategic is nearly always a data-access problem, not a talent problem. The framework’s age is a feature for this reader: it describes the machine-learning foundations, governed data, pipelines and shared assets, that generative AI now sits on and that cannot be skipped.

Bain & Company

Transforming Your Business with AI: Five Questions for Every CEO

Interactive · January 2025, updated October 2025

questions Bain says every CEO should be able to answer, from “am I moving fast enough?” to “how should I lead the organisation on this journey?”

Read the source

Footnote 8 · Leadership owns it · Redesign the work · Data foundations first · Adoption over tooling · Value, focused and measured · Built to scale past pilots

AI is a business transformation, not a technology deployment, and it starts and stops with the CEO. Five questions frame the agenda.

Am I moving fast enough? Bain opens with the pace of the technology itself. Its chart of successive frontier models in late 2024 shows accuracy on competition mathematics rising from about 13% to 83% and on competition coding from 11% to 89%, with PhD-level science questions passing the human-expert benchmark. Global venture funding for generative AI on the same chart climbs from under half a billion dollars in 2018 to about $62 billion in 2024, on Bain’s compilation of Crunchbase and VC Lookback data, alongside a July 2024 survey of what AI leaders do differently.
How might AI change the future of my industry? The worked example is marketing services: Bain maps how the profit pool moves between creative origination, content, production, media and technology providers as AI absorbs the work.
How can AI strengthen our competitive advantage? Two routes, productivity and innovation, and the discipline to choose where the advantage is defensible rather than where the demonstration is easiest.
How is this different, and how do I enable the tech foundation? Generative AI changes the architecture, data and vendor decisions a CEO must sponsor rather than delegate; Bain frames it as a business transformation, not a technology deployment.
How should I lead the organisation on this journey? The closing section is about orchestrating the transformation across people, process and technology, with the CEO as the owner of the agenda.

Bain’s five questions

Am I moving fast enough?
Speed matters: the technology and the leaders are both accelerating.
How might AI change the future of my industry?
Profit pools are being reshaped, function by function.
How can AI strengthen our competitive advantage?
Productivity and innovation, where the advantage is defensible.
How is this different, and how do I enable the tech foundation?
A business transformation, not simply a technology deployment.
How should I lead the organisation on this journey?
Orchestrating the transformation is the CEO’s job.

Question headings as published in Bain’s interactive; notes are Kilwa’s paraphrase of each section’s framing.

The Kilwa read

Two of the five are questions Kilwa’s country intelligence is built to answer with evidence rather than assertion: how AI changes an industry in a specific market, and where an advantage is defensible there. The first question, speed, cuts the other way for African institutions. The leaders will not be the fastest adopters but the ones whose data foundations let them move the moment a use case is proven, which is the argument for building the foundation now.

MIT Technology Review Insights

A playbook for crafting AI strategy

Research report and executive survey, produced in partnership with Boomi · August 2024

76%

of surveyed companies had deployed AI in only one to three use cases, while half expected to deploy it across all business functions within two years.

Read the source

Footnote 9 · Data foundations first · Trust is the licence · Adoption over tooling · Built to scale past pilots

Ambition is universal and scale is rare. Moving from pilots to enterprise-wide AI depends on data liquidity, governance and infrastructure, and the foundations have to be laid now.

Everyone is in; few have scaled. 95% of the companies surveyed already use AI and 99% expect to, yet only 5.4% of US businesses were using AI to produce a product or service in 2024.
Spending shifts to readiness. Nine in ten respondents expected to raise spending on data readiness, platform modernisation, cloud migration and data quality, and on strategy, culture and business-model change.
Data liquidity is the attribute that matters. The ability to access, combine and analyse data from many sources, already curated for the task, is what lets AI be applied to a specific business scenario.
Data quality is the binding constraint. Half of respondents named it the most limiting data issue, and it rises with size: companies above $10 billion in revenue were the most likely to cite both quality and infrastructure.
Safety over speed. 98% would forgo being first if it meant deploying safely and securely. Governance, security and privacy were the biggest brake, cited by 45% overall and 65% at the largest companies.

What the survey found, share of respondents

Would forgo being first in order to deploy safely
98%
Already using AI
95%
Have deployed AI in only one to three use cases
76%
Name data quality the most limiting data issue
50%
Name governance, security and privacy the biggest brake
45%

MIT Technology Review Insights survey, 2024, produced in partnership with Boomi. Chart drawn by Kilwa.

The Kilwa read

Data liquidity is the constraint everywhere. In African markets it is the constraint squared, because the data sits fragmented across regulators, registries and the press, in several languages, and partly in the informal economy. Kilwa’s platform exists to make market data liquid: ingest it, flag its provenance, score it and explain the score. For an allocator the survey’s other number matters as much: half of respondents plan full deployment within two years, so the window in which foundations are a differentiator is short.

From reading to practice

How the nine shape what Kilwa builds.

Curating the best strategies is only useful if they change the work. Four commitments run from these sources straight into Kilwa's research, platform and engagements.

Stanford · BCG · McKinsey

Evidence before tooling.

Kilwa publishes the model, the inputs and the robustness testing with every score, and scopes engagements around the decision the intelligence will change, not the technology.

How we score

McKinsey · Microsoft · MIT Technology Review

Trust is the licence to be used.

Provenance flags on every input, attribution and a model card on every score. The controls the sources call prerequisites are the platform’s transparency surface.

Explainable AI

McKinsey QuantumBlack · Bain · Anthropic

Each service answers one decision.

Sizing the opportunity market by market, and answering how AI changes an industry in a specific country, is what Kilwa’s five services exist to do with primary evidence.

Our services

Google · Anthropic · BCG

Capability transferred, not rented.

Readiness assessment, one narrow pilot, then capability transfer: Kilwa Sovereign AI follows the sequence the playbooks describe, so the partner can run the system without us.

Kilwa Sovereign AI

Sources

The nine, in full.

1.Stanford Digital Economy Lab. The Enterprise AI Playbook: Lessons from 51 Successful Deployments. Elisa Pereira, Alvin Wang Graylin and Erik Brynjolfsson. April 2026. digitaleconomy.stanford.edu/publication/enterprise-ai-playbook
2.McKinsey & Company, QuantumBlack. The executive’s AI playbook. November 2018. www.mckinsey.com/capabilities/quantumblack/our-insights/the-executives-ai-playbook
3.Anthropic. The Enterprise AI Transformation Guide. October 2025. resources.anthropic.com/hubfs/The%20Enterprise%20AI%20Transformation%20Guide%20101425%20(1).pdf?hsLang=en
4.McKinsey & Company. The AI transformation manifesto. Alex Singla, Alexander Sukharevsky, Eric Lamarre, Kate Smaje and Robert Levin. April 2026. www.mckinsey.com/capabilities/tech-and-ai/our-insights/the-ai-transformation-manifesto
5.Boston Consulting Group. The Leader’s Guide to Transforming with AI. December 2024. www.bcg.com/featured-insights/the-leaders-guide-to-transforming-with-ai
6.Microsoft. Rewriting the IT Playbook: Empowering CIOs to Lead with Confidence in the AI Era. May 2025. techcommunity.microsoft.com/blog/microsoft365copilotblog/rewriting-the-it-playbook-empowering-cios-to-lead-with-confidence-in-the-ai-era/4411734
7.Google Cloud. Google Cloud’s AI Adoption Framework. June 2020. cloud.google.com/resources/cloud-ai-adoption-framework-whitepaper
8.Bain & Company. Transforming Your Business with AI: Five Questions for Every CEO. January 2025, updated October 2025. www.bain.com/insights/transform-business-with-ai-five-questions-for-every-ceo
9.MIT Technology Review Insights. A playbook for crafting AI strategy. August 2024. www.technologyreview.com/2024/08/05/1095447/a-playbook-for-crafting-ai-strategy

Every summary on this page is Kilwa’s paraphrase of the cited source, checked against the original text. Figures are quoted as the source states them, with the source’s own caveats where it gives them. No third-party chart, exhibit or extended passage is reproduced; the figures are drawn by Kilwa from published numbers and say so in their captions.

Source names and titles are used to identify the works. Kilwa is not affiliated with, sponsored by or endorsed by any of the publishers. Source 9 was produced by MIT Technology Review Insights in partnership with Boomi, as the publisher states. Source 2 sits behind McKinsey’s registration wall; its structure is summarised from the published description.

This page is analysis and education, not investment, legal or tax advice, and does not recommend any security or transaction. Sources last checked 7 September 2026.

Put the best thinking to work

Bring a decision. We will bring the evidence.

Kilwa's services turn this reading into country, sector and readiness intelligence for the market you are actually in. Scope an engagement, or start with the research.

Scope an engagement Read Kilwa research