Study by appliedAI & ODGERS: Leading the AI Transformation

AI capability keeps compounding - enterprise value does not. 95% of generative-AI pilots deliver no measurable P&L impact [1]; only 39% of adopters report EBIT impact at enterprise level [2]; roughly one in four executives says their company has created significant value from AI [4]. The evidence consistently points away from the technology and toward the organisation: AI is a transformation task, not an IT rollout - and a permanent one, not a project.

That makes it a core leadership task. The insight is widely cited; what has been missing is its practical translation into the leadership function - what, concretely, should an executive team do differently on Monday morning, and how would it know it is improving?

appliedAI and Odgers have developed the Leadership AI Maturity Assessment to close that gap: eight dimensions, 35 behaviour-anchored statements, each rated from two perspectives - self-image and the executive team as a collective - mapped to a five-level maturity model. It takes about twelve minutes and turns “leadership matters” into a measurable profile, visible blind spots, and a concrete board agenda. And supports an increased company valuation through a progressing maturity. [7]

By 2026, the capability question is settled. Models write production code, agents execute multi-step processes, and every quarter moves the frontier again. The economics question is not settled at all. MIT's “GenAI Divide” research - based on 150 executive interviews, a 350-employee survey and 300 analysed deployments - found that about 5% of enterprise AI pilots achieve rapid value acceleration; the rest stall with little or no measurable P&L effect [1].

The pattern repeats across studies. McKinsey reports that 64% of organisations say AI enables innovation and many see use-case-level benefits - yet only 39% register EBIT impact at the enterprise level [2]. BCG finds that only about a quarter of executives report significant value created from AI initiatives [4]. Adoption is nearly universal; returns are rare. Something between the demo and the P&L is systematically breaking.

And in our daily project work, we see the same patterns repeating, too.

What breaks is almost never the model. MIT attributes the failure pattern to a “learning gap” - organisations and tools that do not adapt to each other - rather than to model quality, regulation or infrastructure [1]. BCG quantifies the same intuition with its 10–20–70 rule: roughly 10% of the effort behind AI value lies in algorithms, 20% in data and technology, and 70% in people, processes and culture [3].

The single strongest lever in McKinsey's data makes the point concrete: of 25 practices tested, the fundamental redesign of workflows has the largest effect on EBIT impact from AI - and only 21% of adopters actually do it [2]. Most organisations bolt AI onto processes designed for humans and are then surprised that nothing moves. Value comes from changing how work is done; that is organisational change, and organisational change cannot be purchased with a licence key.

The second misunderstanding is treating AI as a programme with an end date. Three forces make it permanent. First, the technology refuses to stabilise: each capability jump - most recently autonomous agents - reopens yesterday's decisions about processes, roles and sourcing. Second, regulation has arrived as a standing obligation, not a one-off compliance exercise: the EU AI Act's high-risk requirements became binding in August 2026, with further categories following in 2027 [6]. Third, the workforce dimension never closes - skills, roles and expectations shift continuously as AI takes over more of the routine layer [4].

An organisation can delegate a project. It cannot delegate a permanent change in how it operates. That is why the accountability lands - structurally, not rhetorically - with the executive team.

“AI transformation is a leadership topic” has become a truism of keynotes and annual reports. The data behind it is real: McKinsey finds CEO oversight of AI governance to be the element most strongly correlated with bottom-line AI impact - present in only 28% of adopting organisations [2]. Odgers’  work with CFOs and board directors shows the same readiness gap from the inside: over 80% of organisations describe themselves as stuck in planning or preliminary stages, and financial executives voice explicit doubts about their leadership's competence to execute an AI strategy [5].

What the truism lacks is operationalisation. Between “leadership matters” and a board calendar lies a set of unanswered questions: Which leadership behaviours actually move AI value? How does an individual executive score on them - and how does the top team as a whole? Where do self-image and collective reality diverge? Without answers, the leadership task stays an applause line. The Leadership AI Maturity Assessment was built to answer them.

Developed jointly by appliedAI, Europe's largest initiative for applied artificial intelligence, and Odgers, a leading global executive search and leadership advisory firm, the assessment translates the transformation evidence into an instrument for executive teams and boards. Underlying is among others 4 years of experience within the joint effort of the “Data & AI Leadership Circle”, a Masterclass program designed and executed by the two partners jointly that aims to develop leadership skills for AI leaders. Evaluating jointly AI Leaders for the Best of AI Awards together with one of Europe’s leading economic magazines “Capital” has added additional insights.

Design principles

• Behaviour-anchored and verifiable. Items describe observable behaviour or checkable artefacts (“a written target picture exists”), not attitudes - limiting social-desirability inflation.
• Two perspectives per statement. Every item is rated twice: for oneself (own area of responsibility) and for the executive team as a collective. The delta between the two is diagnostic gold - it surfaces blind spots and misalignment inside the top team.
• Executive-grade length. Eight dimensions, 35 statements, roughly twelve minutes - deliberately compact so complete, honest answers are realistic.
• A five-level maturity model. Results map to levels from Awareness (1) through Building, Established and Scaling to Excellence (5), per dimension and overall - making progress measurable across repeat assessments.
• Current as of 2026. The item set covers what boards face now: autonomous agents, AI-specific security threats, shadow-AI usage, EU AI Act high-risk obligations, and workflow redesign rather than 1:1 automation.

The dimensions span the full arc of the leadership task: from personal understanding (1) through organisational and cultural enablement (2–4), execution and value capture (5–7), to strategic anchoring (8).

6.1  AI & Data Strategic Literacy

Leaders cannot steer what they do not understand. This dimension measures whether executives grasp core AI concepts and their failure modes, use AI hands-on in their own work - delegating entire tasks, not just asking questions - demand credible ROI logic instead of vague promises, and treat data as a strategic product. MIT's research identifies a “learning gap”, not model quality, as the root cause of failed initiatives [1]; that gap starts at the top.

Questions the assessment raises: Do I understand hallucination and bias well enough to weigh risk against value? Do I use AI for whole workflows myself? Would I recognise a weak AI business case?

6.2 Ambidextrous Transformation Leadership

AI forces organisations to run today's business efficiently while rethinking it radically - at the same time. This dimension measures whether leadership protects experiments from bureaucracy and the corporate immune system, keeps critical operations stable while moving fast, and is willing to ask the uncomfortable founding question: what would we look like if we started this company AI-first today?

Questions the assessment raises: Do we shield innovation teams? Do we accept failed experiments as learning? Have we seriously stress-tested our own business model?

6.3  Change & Culture Orchestration

In BCG's 10–20–70 rule, 70% of AI value sits in people, processes and culture [3]. This dimension measures the leadership behaviours that unlock that 70%: a credible shared narrative for why the organisation adopts AI  including an honest account of what changes for roles - visible personal use (“walk the talk”), taking fears of job loss seriously, involving affected employees in solution design, and building networks of AI champions across hierarchy levels.

Questions the assessment raises: Is our story about AI consistent and credible? Do employees shape solutions or have them imposed? Who are our champions below the executive level?

6.4  AI Governance & Responsible AI

McKinsey finds that CEO oversight of AI governance is the single element most strongly correlated with bottom-line impact from AI - yet only 28% of organisations have it [2]. This dimension measures whether governance is owned by leadership rather than delegated to legal, whether the organisation knows which of its systems fall under the EU AI Act's high-risk obligations (binding since August 2026 [6]), where ethical red lines lie, who is accountable for AI-produced outcomes - and, new in 2026, whether autonomous agents are risk-classified, monitored and given escalation paths, with AI-specific security threats anchored in risk management.

Questions the assessment raises: Which of our systems are high-risk under the AI Act? When must a human take over? Who answers for an agent's mistake?

6.5  Use Case Portfolio & Scaling Excellence

This is where the value gap becomes visible: 95% of pilots never reach the P&L [1], and fewer than one in four organisations get AI agents into production [5]. The dimension measures whether large initiatives are prioritised by value while teams keep the freedom to build small solutions themselves, whether successful pilots are industrialised instead of celebrated, whether usage and impact are measured against hard business KPIs - and whether processes are redesigned around what AI can do rather than automated as-is. McKinsey identifies workflow redesign as the practice with the biggest EBIT effect of all attributes tested, practised by only 21% of adopters [2].

Questions the assessment raises: How many pilots reached production in the last year? Do we measure adoption and impact? Have we redesigned a single end-to-end process?

6.6  Ecosystem & Technology Alliances

No organisation masters AI alone. MIT's data shows externally sourced solutions and partnerships succeed roughly twice as often as internal builds [1]. This dimension measures whether leadership maintains an active partner network to spot trends early, has a clear line on what to build in-house (because it is competitively differentiating) versus buy (because it is commodity), and manages dependency - avoiding lock-in to single providers and making sure supplier knowledge transfers into the organisation.

Questions the assessment raises: Where do we deliberately build, where do we buy? What happens if our main model provider changes terms tomorrow?

6.7  Talent & Future Skills

BCG's “future-built” companies plan to upskill more than half of their workforce on AI - laggards, 20% - and are four times more likely to run structured AI learning with protected time [4]. This dimension measures whether budget and working time for AI upskilling are explicitly committed, whether leadership knows which key competences and roles are missing and fills those gaps deliberately, whether people are trained to collaborate with AI - delegate, verify, stay accountable - and whether effective AI use has become a stated expectation in role profiles and performance reviews.

Questions the assessment raises: Is learning time protected or aspirational? Which three AI-critical roles are unfilled today? Is AI fluency part of how we evaluate performance?

6.8  AI Strategy & Organization

Ambition without architecture does not scale. This dimension measures whether a written target picture exists for how AI will have changed the business model in three years, whether roles such as AI Lead and Data Owner carry real mandates within scalable structures, whether the AI roadmap is budgeted with milestones and named accountability, whether strategic assumptions are revisited as the technology jumps - and whether AI shows up in products and services for customers, not only in internal productivity.

Questions the assessment raises: Does the target picture exist in writing? Who owns the roadmap? Where is AI in our offering, not just our cost base?

A maturity score is a means, not an end. The assessment is designed to feed a specific working sequence in the executive team:

• Individual profile. Each executive receives a radar across the eight dimensions, with self-image and collective view overlaid, plus an overall maturity level.
• Team aggregation. Aggregated anonymously across the board, the profiles show where the team is aligned, where perceptions diverge, and which dimensions lag.
• Blind-spot dialogue. The largest self–collective deltas become the agenda of a facilitated board session - they are, in our experience, the most productive conversations the instrument triggers.
• Development and re-measurement. Findings translate into concrete moves - governance ownership, workflow-redesign mandates, upskilling commitments - and the assessment is repeated after six to twelve months to make progress visible.

Used this way, the instrument does for the leadership dimension what dashboards already do for the use-case portfolio: it replaces assertion with measurement.

Scope: 8 dimensions · 35 behaviour-anchored statements · ~12 minutes.  

Rating: five-point agreement scale (1 = does not apply … 5 = fully applies) plus “not assessable”, excluded from scoring.  

Perspectives: each statement rated for self (own remit) and for the executive team as a collective.  

Output: maturity level 1–5 overall and per dimension; radar profile; self–collective delta and blind-spot analysis.  

Formats: interactive web version with automated scoring and JSON export; print/PDF version for workshop settings.

appliedAI Initiative is Europe's largest initiative for the application of trustworthy AI, working with enterprises, startups and the public sector to lift AI capabilities from strategy to implementation.

Odgers is a leading global executive search and leadership advisory firm, supporting boards and executive teams in appointing and developing the leadership their transformation requires.

[1]  MIT NANDA (2025): The GenAI Divide — State of AI in Business 2025. Reported in Fortune, August 2025. fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo/

[2]  McKinsey & Company (2025): The State of AI — How organizations are rewiring to capture value (March 2025) and The State of AI in 2025: Agents, innovation, and transformation (November 2025). mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai

[3]  BCG (2025): From Potential to Profit — Closing the AI Impact Gap (AI Radar). bcg.com/publications/2025/closing-the-ai-impact-gap

[4]  BCG (2026): AI Talk Is Cheap. Value Creation Is Rare. / AI Transformation Is a Workforce Transformation. bcg.com/publications/2026

[5]  Odgers Berndtson (2024–25): CFO Leadership Survey on Artificial Intelligence readiness. odgers.com; Deloitte (2026): Tech Trends — Agentic AI Strategy. deloitte.com

[6]  European Commission: EU AI Act — high-risk obligations applicable from 2 August 2026; selected Annex III categories deferred to December 2027 (Digital Omnibus). digital-strategy.ec.europa.eu

[7] McKinsey study: Beyond productivity: How AI creates value in private equity from 2026

Ready to accelerate your AI transformation? Learn more about your company's leadership ai maturity!

Would you like to learn more about your company's AI maturity and identify recommendations and next steps? Contact us and participate in our Leadership AI Maturity Assessment! Find out more about the new assessment from Odgers and appliedAI here.