Five Things Boards Need to Get Right with AI: BCG Names the Director-Level Playbook

See also (wiki): board-ai-strategy, ai-vendor-contracts

Executive Summary

• BCG’s February 24, 2026 article by Nicolas de Bellefonds and Vladimir Lukic opens with the single most-cited data point about boardroom AI readiness today: at a recent gathering of corporate directors, when asked who felt ready to oversee AI, no one raised a hand. The piece is BCG’s first dedicated board-of-directors framework in the 2026 AI corpus.
• The framework is five tight director-level obligations — drive pace and priorities, protect strategic freedom, shape the AI investment portfolio, be a growth-oriented voice of reason, and keep communications measured — plus a sixth behavioral prompt: get in the sandbox and play.
• The most actionable concept is one-way doors: a practical board test for irreversible technology decisions (partner ecosystem lock-in, proprietary architecture, narrow talent model) that is simple enough to use in a 60-minute board session this quarter.
• BCG explicitly calls out the remuneration committee as the governance body that converts AI ambition into delivery. If compensation and evaluation criteria still reward the legacy model, the transformation stalls. This is the specific governance lever most mid-market boards underuse.
• BCG references a new SEC disclosure rule requiring companies to identify which annual report statements are generated by AI. Boards that haven’t reviewed communications discipline around AI statements — “aggressive rollout” comments read as earnings signals, efficiency claims read as pricing concessions — are carrying disclosure risk they may not have priced.
• The piece is a prescriptive framework, not RCT evidence. BCG’s Board Governance and Corporate Governance practice benefits from board-advisory engagements. Read it as the board-level companion to McKinsey’s AI Transformation Manifesto (Apr 7, 2026) and BCG’s own AI Radar 2026 (Jan 15, 2026) CEO-lens data — not as a new quantitative benchmark.

The Admission That Frames the Piece

De Bellefonds and Lukic open with a sentence mid-market directors will recognize: “At a recent gathering of corporate directors, when asked who felt ready to oversee AI, no one raised a hand.” That is where the piece lands its thesis — awareness is high, stakes are obvious, and the judgment muscle is not yet built.

This tracks against the quantitative baseline already in the research corpus: 79% of board members report limited-to-no AI knowledge (Deloitte, n=468, May–July 2024) against 88% of organizations using AI in at least one function (McKinsey State of AI, November 2025, n=1,993). The Deloitte number predates current model capabilities by a generation and should be read as historical trend context, not a current operational claim. The direction of the gap, however, is consistent across the 2025–2026 evidence.

BCG’s January 2026 AI Radar (n=2,360 executives including 640 CEOs) adds the complementary data point: 72% of CEOs now describe themselves as their company’s main decision-maker on AI — nearly double the prior year. The practical effect is that AI decisions are being made at the CEO level without the board assurance layer that governs comparable fiduciary domains. The Five Things framework is BCG’s attempt to give directors a way to engage without pretending to be practitioners.

The Five Things, in Plain Language

1. Drive Pace, Purpose, and Priorities

BCG’s prescription is concentration: “A few big bets are better than a dozen pilots, but those bets need to matter.” The AI agenda should target capabilities and value pools that shape competitive position — differentiation, customer economics, speed to market — not a long tail of use cases justified by the easiest business case.

The board move BCG recommends is devil’s advocate questioning at the strategy-intersection level: “What would a serious rival or an AI-first entrant try to change over the next few years? How could they use AI to reset customer expectations, cost structures, or speed to market?”

The triangulation here matters. McKinsey’s April 2026 AI Transformation Manifesto (n=20 AI-leading companies) finds concentration on 1–3 business domains — not broad use-case portfolios — is the pattern behind the average 20% EBITDA uplift and $3-per-$1 return across leading transformations. BCG and McKinsey are converging on the same prescription from different angles: the 5% captures value because they concentrate; the 95% accumulates pilots because no one at the board level is forcing the concentration decision.

2. Protect Strategic Freedom — The One-Way Doors Test

This is the single most executable idea in the piece. BCG frames certain AI choices as one-way doors — decisions that effectively lock a company into a partner ecosystem, a proprietary architecture, or a narrow talent model, and that are difficult to unwind once committed. The director-level test is to ask, for any major AI choice: what alternatives remain, and what would it take to reopen them?

The practical design responses BCG names are modular design, diversified partnerships, and a deliberate balance between internal and external capabilities. This is the board-level articulation of vendor-lock-in risk that MIT CISR’s Minimum Viable Governance briefing (Mar 2026, n=17 FinCo leaders) captures at the operational level: five governance domains across two axes with explicit tradeoffs between speed and optionality.

The director audience BCG is writing for owns a specific lever here — the board approves capital that commits multi-year architecture direction. A 200–2,000 person company making a 3-year Microsoft Copilot-plus-Azure-OpenAI commitment or a 5-year Salesforce Einstein-plus-Data-Cloud commitment is walking through a one-way door whether or not that is named. The board should price the cost of reversal.

3. Shape the AI Investment Portfolio — The Deploy–Reshape–Invent Mix

BCG reframes AI investment not as a project list but as a three-bucket portfolio:

• Deploy — near-term performance gains, ROI discipline where outcomes should be clear
• Reshape — transforming cost structure or operating leverage over time
• Invent — less-certain growth avenues that are strategically important

The board’s value-add is insisting on ROI discipline in the Deploy bucket without crowding out the longer-horizon Reshape and Invent bets. BCG’s warning: “Without that balance, AI programs default either to safe incrementalism or scattered experimentation. Neither changes the company’s competitive trajectory.”

For a mid-market board, the practical translation is a question for the CFO and CIO jointly: what percentage of this year’s AI capex sits in each bucket, and how does that mix compare to last year? If 100% sits in Deploy, the company is harvesting today’s productivity at the expense of tomorrow’s advantage. If 60%+ sits in Invent, the company is funding science projects while competitors monetize.

This maps cleanly to BCG AI Radar 2026’s “Trailblazer” vs. “Follower” archetypes: Trailblazers upskill 69% of the workforce vs. 35% for Followers, and the Trailblazers’ capital allocation is noticeably more balanced across the three buckets.

4. Be a Growth-Oriented Voice of Reason — The Remuneration-Committee Lever

This is where BCG names the specific governance body that most mid-market boards underuse for AI transformation: the remuneration committee.

The logic: AI transformation changes how work gets done, how decisions are made, and what performance looks like. If compensation, evaluation criteria, and governance still reward the legacy model — revenue per sales rep, cost per case, tickets closed per agent — the transformation stalls, regardless of budget. BCG’s prescription is to align leadership incentives with AI-readiness outcomes: data readiness, adoption of new workflows, measurable business impact of AI-enabled processes.

This is the single hardest recommendation to execute. Comp committees are legally conservative, shareholder-sensitive, and institutionally biased toward metric continuity. Adding a new AI-readiness metric to the FY27 executive scorecard is a real fiduciary decision, not a workshop exercise. BCG is explicit that boards also have to decide “how much disruption the company is willing to absorb” — leaders rewiring parts of the business need clear backing from the board to push through short-term disruption.

The practitioner-voice corroboration comes from Deloitte’s 2026 Global Human Capital Trends (n=9,000+, Mar 2026): organizations using human-centric AI strategies are 1.6x more likely to exceed investment return expectations versus tech-first approaches — yet 59% of organizations still default to tech-first. Boards that only approve tool budgets without touching the incentive system are funding the pattern that underperforms.

5. Keep Communications Measured and Intentional — The SEC Disclosure Note

BCG’s fifth item introduces what may be news to many mid-market directors: SEC rules now require companies to identify which annual report statements are generated by AI. The piece does not specify the rule or effective date, but frames the broader stakes: even casual AI-related comments create expectations. A reference to an “aggressive rollout” reads as a cost target. Efficiency claims invite customer pricing pressure.

The board move BCG recommends is discipline around the before: directors should press management on what is real today, what remains experimental, and what is still uncertain before those messages leave the room.

This item connects directly to the corpus’s AI-washing and enforcement thread. The SEC created its Cyber and Emerging Technologies Unit (CETU, Feb 2025) with AI-washing as a top priority. The FTC, Federal Trade Commission, and multiple state AGs have taken enforcement positions on overstated AI capabilities. The practical governance consequence: a board that has not reviewed AI communications discipline in the past four quarters is carrying disclosure risk that may not be priced into its D&O posture.

The Sixth Item: Get in the Sandbox and Play

BCG closes with a behavioral prompt rather than a governance action. Boards do not need to become fluent in AI — but they cannot govern what they only hear about secondhand. The most effective boards, BCG argues, proactively commit to upskilling: they engage not only with the tools their businesses use but also incorporate AI tools into their own flow of work.

The payoff BCG names is specifically political: firsthand experience spread across multiple board members shifts the dynamic in the boardroom. It keeps the conversation from leaning too heavily on one or two tech-savvy voices — the director-specific version of the broader “AI literacy is an adoption architecture” finding from the training-effectiveness corpus (144% higher trust from hands-on vs. observational training, Deloitte 2024; 2.6x higher usage consistency when workers trust the program, multiple sources).

How This Piece Fits the Board-Level Evidence Stack

Read as one piece in a four-voice 2026 board-governance cluster:

Mid-market boards that want to wire up AI oversight without buying a full board-advisory engagement can combine these four into a 90-minute board working session: BCG’s Five Things as the agenda, AI Radar for the peer data, the McKinsey Manifesto as the benchmark, and the IBM IBV Playbook as the agent-specific control list.

Source Credibility — What to Weight and What to Caveat

Credibility: MEDIUM. BCG’s Board Governance and Corporate Governance practice has a direct commercial interest in board-advisory engagements. The piece is prescriptive framing, not primary-survey data or RCT evidence; there is no sample size, no methodology section, and no quantitative benchmark offered in the body. The “no one raised a hand” anecdote is illustrative, not measured. The framework is useful because it distills a consensus pattern from BCG’s board-advisory practice, not because it is empirically new.

The piece is strongest as a discussion structure for a board meeting; it is weakest if read as evidence that the BCG framework outperforms alternatives. There is no counterfactual in the article.

Key Data Points

What This Means for Your Organization

If your board has not yet set up a structured AI-oversight cadence, BCG’s five items are a better agenda than most of the templates being sold in the current advisory market. The framework is compact enough to work as a 60-minute board session this quarter. It does three things a longer framework cannot: it names the one-way doors test in a form directors can actually apply; it points at the remuneration committee as the specific governance body that converts ambition into delivery; and it gives directors an explicit behavioral prompt — use the tools, do not just hear about them — that avoids the trap of trying to turn non-technical directors into practitioners.

The gap most mid-market boards are sitting on is not the agenda. It is the assurance layer behind it. Seven items to consider for the FY27 board calendar: (1) a quarterly AI-oversight agenda item with minimum 30 minutes of full-board time; (2) a board-level inventory of one-way doors already walked through and planned; (3) a deploy–reshape–invent portfolio summary from the CFO and CIO jointly, with year-over-year movement; (4) a remuneration-committee review of how executive scorecards reward AI-readiness outcomes such as workflow adoption and data readiness, not just tool budgets; (5) an AI communications-discipline review covering earnings-call language, investor-day claims, and the SEC AI-disclosure note on the annual report; (6) a board-member AI-literacy plan with personal-use components, not just briefings; (7) a named executive sponsor with sustained CEO-level backing — projects with this structure achieve 68% success rates versus 11% for those that lose it within 6 months (Pertama Partners, 2026).

If you are standing up or restructuring an AI governance architecture at the board level this year and want a second set of eyes on how the five items map against your committee charters, reporting cadence, and comp-committee scorecard — I would welcome the conversation. brandon@brandonsneider.com.

Sources

• Nicolas de Bellefonds and Vladimir Lukic, “Five Things Boards Need to Get Right with AI,” Boston Consulting Group, February 24, 2026. https://www.bcg.com/publications/2026/five-things-boards-need-to-get-right-with-ai — MEDIUM credibility: prescriptive framework from BCG Board Governance practice with direct commercial interest in board-advisory engagements; no primary-survey data; useful as discussion structure.
• BCG AI Radar 2026, “As AI Investments Surge, CEOs Take the Lead,” Jan 15, 2026, n=2,360 executives including 640 CEOs. https://www.bcg.com/publications/2026/as-ai-investments-surge-ceos-take-the-lead — MEDIUM-HIGH credibility: BCG primary CEO survey; companion CEO-lens data to the Five Things framework.
• McKinsey & Company, “The AI Transformation Manifesto: 12 Themes Driving Growth,” QuantumBlack, April 7, 2026, n=20 AI-leading companies. https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/the-ai-transformation-manifesto — MEDIUM-HIGH credibility: small-n prescriptive benchmark with explicit portfolio-discipline finding; triangulates with BCG’s Portfolio and Priorities items.
• Deloitte, “2026 Global Human Capital Trends: From Tensions to Tipping Points,” March 4, 2026, n=9,000+ leaders across 89 countries, Oxford Economics fieldwork. https://www.deloitte.com/us/en/insights/topics/talent/human-capital-trends.html — MEDIUM-HIGH credibility: large-sample survey evidence for the remuneration-committee recommendation.
• MIT CISR, “Minimum Viable Governance for Generative AI,” van der Meulen, Jewer, Levallet, March 19, 2026, n=17 leader interviews plus FinCo case study. https://cisr.mit.edu/publication/2026_0301_GenAIGovernance_VanderMeulenJewerLevallet — HIGH credibility (academic): the governance-architecture operational companion to BCG’s board-level framing.
• Deloitte, “On the Board’s Agenda: Generative AI,” n=468 board members, May–July 2024. Cited as historical trend context (Tier 3); predates current model generation.

Brandon Sneider | brandon@brandonsneider.com April 2026