The competitive window is open — and it is closing
AI is not a cost story. It is a scale story. The companies that figure this out first will be able to do what took organizations ten times larger — and do it faster. The board's question is whether the CEO is the one leading that charge, or watching it happen.
The default frame in most boardrooms is still efficiency — fewer headcount, lower cost per output, margin expansion through substitution. That frame is not wrong. It is small. And in a growth-stage company, small is the wrong ambition.
The more accurate frame is force multiplication: a small, well-configured team with the right tools and instincts can do what previously required organizations ten times their size. That changes the economics of scale, the calculus of talent density, and the leadership model required to capture the opportunity. It also changes what the board should be asking its portfolio CEOs.
The window is not permanent
The capability curve is not linear. The length of tasks AI agents can complete autonomously has been doubling roughly every seven months (McKinsey, 2025). Companies that are exploring now will be building around AI in 12 months. Companies that are lagging now will be structurally disadvantaged — not because they lack tools, but because the organizational learning, the workflows, and the leadership muscle will not have compounded.
This has a direct implication for the PE board. The question is not whether to bring this to a portfolio CEO. The question is how fast, and whether the CEO you have is capable of leading it. The diagnostic on tab 06 is designed to help answer both.
Most mandates collapse at the same point
McKinsey, BCG, and Bain now distinguish three tiers of value capture: deploy (give people tools), reshape (redesign workflows), and invent (build AI-native products). Most companies are stuck at deploy. The bold CEO rhetoric tends to be a deploy-tier story dressed up as reshape-tier ambition. That gap is where most mandates collapse.
BCG's 10/20/70 rule is consistent across failed AI programs: 10% of the effort to algorithms, 20% to technology and data, 70% to people, process, and change management. The CEO is the 70%. That is not a technology investment decision — it is a leadership assessment question.
The Klarna case is instructive on the other end. A public commitment to AI efficiency ahead of internal readiness, followed by reversal when quality degraded. For PE-backed companies, AI productivity language in LP communications that runs ahead of internal traction creates a credibility gap that is difficult to recover from. The lesson is sequencing: build the capability, then tell the story.
Every layer of the organization — but the leverage is at the top
AI transformation does not fail because employees resist it. It fails because leadership does not model it, and the organization takes its cue from what leaders actually do. The board's question is whether the right people, in the right roles, have the orientation and urgency to drive this.
The CEO
Sets the signal. Without personal practice that is visible and demonstrable, everything else is performance. The organization calibrates its urgency to the CEO's urgency — precisely and quickly.
Senior leadership team
The substrate the CEO's signal has to land in. One or two engaged members produces isolated adoption. All members engaged, each with a function-specific application, is what drives org-wide movement.
The broader organization
Can only move as fast as the infrastructure allows — approved tools, clear data policies, role-specific training. Without this layer, "play and learn" selects for the curious 20% and leaves the rest behind.
AI compresses performance distributions
The most rigorous empirical work on this (Dell'Acqua, Mollick et al., HBS/BCG, 2023 — a field experiment with 758 BCG consultants) found that AI-assisted consultants completed 12% more tasks, 25% faster, at 40% higher quality. Two findings matter more than those headlines.
The skill-leveler effect: Bottom-quartile performers gained 43%; top-quartile gained 17%. AI compresses performance distributions. For a PE portfolio company, this is a talent density argument — the same headcount can produce materially more output if the bottom of the distribution is lifted. The implication for talent strategy is significant.
The jagged frontier: AI is unevenly capable across tasks that look similar from the outside. People who used AI on tasks just outside its capability envelope produced worse results than those without it — they "fell asleep at the wheel." This is the failure mode most leadership teams do not anticipate, and it is a direct argument for the CEO having a genuine working understanding of what AI can and cannot do.
Human value concentrates in different places
As AI takes on coordination, execution, and increasingly analysis and synthesis, the nature of what leadership is changes — not just what leaders use. Human value concentrates in three places:
Problem definition. AI agents are highly effective when goals are precisely defined and nearly useless when they are vague. The leader's most important capability shifts to defining problems with precision — and knowing which problems are worth solving at all.
Judgment over execution. When AI handles execution, human value concentrates in the calls machines cannot reliably make: ethical decisions, relationship management, creative direction, navigating high-stakes ambiguity where data is thin and context is everything.
Building the operating model around the capability. The CEO's job becomes identifying who the right people are to harness this capacity, creating the environment in which they can operate at full effectiveness, and migrating the company's output toward that model. This is not an IT project. It is an operating model transformation — and it is the CEO's project.
The challenge ahead is not scaling from $1B to $3B. It is building something capable of reaching $100B. Those challenges are not the same. The first is an execution problem. The second is a transformation problem — one that requires fundamentally new thinking about what leadership means when the nature of work itself is changing.
TalentScape Partners · CEO Leadership DebriefTwo layers. Both required. Neither sufficient alone.
The companies that successfully transform have two things moving together: visible leadership behavior at the top, and the infrastructure that lets that signal reach the organization. Most companies build one and wonder why the other isn't working.
Layer 1 — Leadership behavior
What the CEO and senior leadership team personally do. Their visible practice, the questions they ask in every standing meeting, the lab leaders they identify and elevate, the function-specific ambitions they set and own. This is the urgency and direction signal. Without it, the organization has no calibration point.
Layer 2 — Enablement infrastructure
What the company builds for the average employee — structured training, role-specific prompt libraries, workflow templates, clear governance and approved tooling. This is the substrate the leadership signal is supposed to activate. Without it, only the curious 20% adopt, and the signal never translates into org-wide traction.
Behavioral, not rhetorical
Two CEO archetypes dominate the current literature. Tobi Lütke at Shopify declared "reflexive AI usage" a baseline expectation and built it into performance review. That model is often cited, seldom replicable — it is SaaS-native, rhetoric-heavy, and lands differently in operational or international cultures.
Nicolai Tangen at Norges Bank Investment Management is the more usable reference. He mandated universal ChatGPT Enterprise access and asks in every standing meeting how staff are using AI. Reported result: approximately 50% of office employees now write code despite only approximately 20% being trained coders. The mechanism is behavioral, not technological. A single meeting question, asked consistently, without exception.
This matters for the board because it is observable. The question — "What is the most interesting AI use you or your team had this week?" — costs nothing and re-frames every meeting within 60 days. A CEO who is not doing this is not leading the transformation, regardless of what the strategy deck says.
Uniform ambition destroys credibility
Knowledge-work functions — legal, HR, finance, investor relations, marketing, internal customer-facing operations — have meaningfully higher near-term ceilings than operations-heavy or field functions. Applying a single AI productivity target across all functions alienates the operational leaders most likely to tune out generic rhetoric, and gives the high-leverage functions permission to coast.
The right frame is two tiers: high-leverage functions with aggressive, specific ambition; operations-heavy functions with bounded but real targets. Each function head should have signed onto their target and be able to explain the logic. A CEO who is talking about AI as a single number across the company does not yet have a strategy — they have a message.
Governance as an adoption enabler
Absent guardrails are a brake on adoption, not a sign of openness. The rational employee response to ambiguity about what tools are approved and what data can go into them is to avoid AI entirely. The absence of a clear, plain-English policy is itself a strategic failure — it produces either non-use or unmanaged risk, sometimes both simultaneously.
Approved tool lists with enterprise licensing, a plain-English data classification policy, and clear escalation paths framed as enabling safe experimentation are prerequisites for adoption at scale. A CEO who has not commissioned this is not serious about the transformation, regardless of the public language.
The diagnostic framework
Eight dimensions, each with three tiers and a specific diagnostic question. The tiers are not a ladder — a CEO can be cutting edge in personal practice and lagging on enablement simultaneously. The diagnostic question for each dimension is what a board member or operating partner should ask directly.
CEO personal practice
The foundational dimension. Without this, the rest is performance. Must be demonstrable — not describable in categories, but visible in outputs the senior leadership team can see.
Daily use across multiple contexts — board prep, reading synthesis, drafting, decision pre-mortems. Demonstrates without preparation. Senior leadership team can see AI was in the loop in the CEO's work. Has an explicit, updated view on what AI is and is not good at in their own practice.Ask: What did the CEO do with AI yesterday and last Tuesday — specifics, not categories. Without a real answer here, the rest of the agenda is performance.
Uses AI occasionally and intentionally, but not yet a pattern. Can describe one or two recent uses. Would need to prepare to demonstrate. Treats AI as a tool to respect rather than one to rely on.
Treats AI as something the technology team handles. Cannot name a recent use unprompted. May still be using AI as a search engine rather than a thought partner.
Senior leadership team fluency and signal
The CEO is the signal. The senior leadership team is the substrate. Uneven engagement means the signal does not reach the organization below the direct-report level.
Every member can describe their own AI use pattern, has named a function-specific high-leverage application, and is visibly modeling for their teams. AI shows up in planning, reviews, and hiring decisions by default.Ask individually: "What's your team doing with AI that's working, and what have you personally tried in the last two weeks?" — who answers crisply, who deflects, who fakes it.
Two or three members are genuinely engaged. The rest are compliant but not pulling. Direct reports see uneven signal across functions — which functions move depends on personality, not strategy.
Team collectively defers AI questions to the technology function. "We're exploring" is the default answer. Direct reports look up and see no signal.
Function-stratified ambition
Explicit, differentiated targets by function. High-leverage functions (legal, HR, finance, marketing) have aggressive ambition. Operations-heavy functions have credible, bounded targets. Each function head has signed onto theirs and can explain the logic.Ask: Has anyone done a function-by-function ceiling analysis, or is "AI productivity" talked about as a single number across the company?
Awareness that functions differ, but no formal framework. Risk of uniform overshooting (alienating operational leaders) or implicit undershooting (letting high-leverage functions coast).
A single uniform ambition applied across the company. Lands as performative in some functions and unattainable in others. Or no stated ambition at all.
Lab function and visible exemplars
Three to five named individuals across functions with explicit lab status, a light charter, and a regular cadence of sharing. CEO can name them and what each is working on. Their work surfaces in internal communications as concrete stories, not abstractions.Ask: Can the CEO name three people in different functions doing genuinely interesting AI work today? If not, the lab function does not yet exist.
Tinkerers exist and do interesting work, but are invisible or isolated. CEO knows of one or two but not as a coordinated set with any shared visibility.
No visible exemplars. AI work happens quietly or not at all. CEO cannot name anyone doing interesting AI work in different functions.
Enablement infrastructure
Pull-based resources for the curious (prompt libraries, sandboxes, forums) and push-based programs for the rest (role-specific training, workflow templates, hands-on workshops, AI office hours). Funded as core infrastructure, not an innovation experiment.Ask: Can a typical mid-career employee in a non-technical function point to a specific resource that taught them how to use AI in their actual work? If not, "play and learn" is the de facto strategy.
Some training exists but is generic and one-time. Adoption visibly tracks personality rather than role need. The curious have figured it out; the rest have not been given the tools to.
Pure "play and learn." Those who would benefit most are least likely to engage without structured support. Adoption is concentrated in early-adopter personalities and stalls there.
Governance and guardrails
Approved tool list with enterprise licensing. Plain-English data classification policy. Clear escalation paths. Framed as enabling safe experimentation, not restricting it. Employees can confidently answer what they can and cannot put into AI tools.Ask: If a mid-career employee were asked whether they can put a customer-confidential document into a public AI tool, would they give a confident, correct answer?
Some policies exist but are scattered, dated, or not reaching employees. Mixed messages from IT, security, and legal. Most default to caution and underuse.
Ambiguous or absent guardrails — employees using personal accounts, sensitive data in consumer-tier tools. Or blanket prohibitions everyone routes around. Both states create risk and suppress legitimate adoption simultaneously.
Operating cadence and ritual
AI is a default input in the operating rhythm. The Tangen question is in every standing meeting. Performance reviews include AI leadership criteria. Headcount conversations explicitly require demonstrating AI alternatives were explored.Ask: What existing meeting, review, or decision-gate would look meaningfully different in 90 days if the agenda lands? If the answer is "nothing yet," it hasn't landed.
AI is a topic that comes up when someone raises it. Reviews and decision gates are unchanged. The agenda exists as a document, not a rhythm.
AI lives entirely outside the operating rhythm. Treated as an innovation initiative or IT project, not a way of working.
Internal narrative and alignment
Consistent internal vocabulary that fits the company's culture — not borrowed from an investor's deck or another CEO's memo. Senior leadership team uses it consistently. External narrative sequenced behind internal traction, not ahead of it.Ask: If five senior leadership team members were asked separately what the company is doing on AI internally, how aligned would the answers be?
Multiple vocabularies in use across the team. The "what do we say externally" question is unresolved. Risk of importing investor or peer-CEO language that does not fit the culture.
No vocabulary at all, or external language that does not land internally. AI is talked about in board decks but not in hallways.
What the board should expect to see, and when
These are the observable signals of a CEO who is leading with pace and urgency — not a to-do list, but a view of what good looks like from the board's vantage point at 30, 90, and 180 days. A CEO who cannot account for these signals is not moving fast enough.
Personal posture is established. The CEO is doing this — visibly — before asking anyone else to.
CEO can describe — without preparation — two or three specific things they did with AI in the last week. Not categories. Specifics.
The Tangen question has been introduced in the CEO's standing one-on-ones. It is asked. Every time.
Each senior leadership team member has had a direct conversation with the CEO about their current AI use and where it is going. The CEO has a baseline read on the team.
Three to five potential lab leaders have been identified — by name, not by category. At least two are outside of technology or product.
The CEO knows what training currently exists, what the data policy says, and whether either is actually reaching employees.
The infrastructure is funded and moving. The agenda has a skeleton the senior leadership team has signed onto.
Enablement infrastructure is commissioned and resourced — not planned, resourced. Role-specific training, prompt libraries, office hours, or an AI guild. Treated as core infrastructure budget, not an innovation experiment.
Approved tool list and plain-English data policy are published. Employees can find them. They are not ambiguous.
Lab leaders have a light charter and a meeting cadence with the CEO. Their work is starting to surface in internal settings — specific stories, not abstractions.
Function-stratified ambition exists in a short document — two tiers, signed by function heads. Not a single number applied across the company.
AI leadership criteria are drafted into the next performance review cycle. Not a separate instrument — one behaviorally anchored dimension in the existing process.
The board now has data, not impressions. The CEO has evidence — or does not.
First measurement is available: percentage of employees using AI weekly to improve productivity, reported by function. The CEO can tell the board which functions are moving and which are lagging — and why.
First performance review cycle has reflected AI leadership criteria. The CEO has data on which senior leadership team members are moving and which are not — data, not impressions.
Lab leader stories are in town halls and internal communications. Concrete, function-specific, named. Not "we are investing in AI."
The board conversation about external narrative — LP communications, customer-facing language — is now informed by internal traction, not aspirational. If the CEO is pushing for external language before this evidence exists, the sequencing is wrong.
Is my CEO leading this transformation?
Rate each item based on observable evidence — not the CEO's self-report, not what the strategy deck says. The scorecard updates live. The notes fields are where the real analytical value lives.
Rate based on observable behavioral evidence only — not self-report, not reputation. Where formal assessment data (Hogan or equivalent) is available, use it to inform these ratings and note it in the observations field.