From AI Skills to Organisational Capability
In many organisations, the conversation around AI capability begins with skills.
Leaders invest in training programmes, workshops, and upskilling initiatives; teams are encouraged to experiment, a small number of people quickly become very good at using AI tools, sharing prompts, and finding clever ways to speed up their work. From the outside, this looks like progress, and in some ways, it is.
Over time, however, a different question starts to surface in leadership conversations. It’s not “Are people learning?” but “Are we seeing productivity gains we can actually repeat?”
This distinction matters. Many organisations can point to impressive AI use cases or pockets of efficiency. Far fewer can reproduce those gains across teams, roles, and time.
This pattern is clearly reflected in the data. According to McKinsey’s most recent State of AI research, while nearly nine out of ten organisations now report using AI in at least one business function, only a much smaller share have managed to translate that usage into consistent, enterprise-wide productivity impact.
In other words, AI exposure is becoming common, but repeatable productivity gains remain rare. The gap leaders experience is not imagined, and it shows up in how few organisations have moved from isolated improvements to sustained performance gains.(1)
What sits beneath this concern is not doubt about the technology itself. It is a question of
What sits beneath that question is not doubt about the technology itself, but concern whether AI-supported ways of working actually change how work gets done at scale.
Skills and capability are not the same thing
Skills create potential productivity.
Capability turns that potential into repeatable productivity at scale.
Skills describe what individuals can do when conditions are favourable. Capability describes what an organisation can consistently deliver, even when conditions change.
At first glance, this may sound like a subtle distinction. In practice, it determines whether AI becomes a source of dependable performance or remains a collection of promising experiments.
An organisation can have highly skilled people and still lack AI capability.
In those situations, productivity gains depend on individual initiative rather than on shared ways of working. Results vary across teams doing similar work, and when priorities shift or key individuals move on, the gains quietly disappear.
OpenAI’s analysis of enterprise AI usage reinforces this pattern. Even in organisations with broad access to advanced models, usage tends to cluster among a subset of employees and functions, while deeper integration into everyday workflows remains uneven.
High individual usage, in other words, does not automatically translate into repeatable productivity. Without shared baselines and embedded practices, performance remains fragile.(2)
From a leadership perspective, this creates a practical challenge. You cannot steer, scale, or govern AI when productivity gains are inconsistent and dependent on individual effort. Without capability, AI remains visible and promising, but operationally thin.
Capability begins to take shape when AI-supported productivity becomes expected rather than exceptional, embedded into everyday work rather than reliant on personal motivation.
Why a shared baseline matters more than elite expertise
A common assumption is that AI capability comes from cultivating experts.
Though, in practice, we continuously see that what enables organisations to realise repeatable productivity gains is something else entirely: a shared baseline of proficiency.
This doesn’t mean everyone needs to become an AI specialist, it means enough people across the organisation understand how to work with AI confidently and responsibly, so that AI-supported productivity no longer feels exceptional, risky, or dependent on permission.
When a baseline exists:
Employees know when and how AI can support their work
Managers recognise what good AI-supported productivity looks like
Expectations are broadly consistent across teams
When it doesn’t, even strong individual skills fail to scale, leaders hesitate to encourage wider adoption, and employees hesitate to apply AI to meaningful tasks. Activity is visible, but productivity remains uneven.
Deloitte’s recent research on enterprise AI adoption helps explain why this gap persists. While organisations continue to invest in AI and express growing confidence in the technology, far fewer have redesigned roles, workflows, or decision rights to reflect AI-supported work.
As a result, skills rise faster than organisational capability. Productivity potential increases, but the organisation is not yet structured to repeat it.(3)
This dynamic is often reinforced by a gap in perception. Research from McKinsey suggests that employees frequently experiment with and adopt AI tools faster than leadership realises. Skills accumulate informally, while shared expectations and organisational alignment lag behind.(4)
Over time, AI use spreads, but without a common baseline, leaders struggle to guide it deliberately toward repeatable outcomes.
From scattered skills to a Skill Stack
The missing link between AI skills and organisational capability is not more experimentation, but a way to structure and compound those skills.
Individual AI skills translate into repeatable productivity only when they build on one another and are widely shared.
When skills remain scattered, progress stays fragile.
When they stack, capability compounds.
At a high level, the Skill Stack includes:
Prompting - the ability to clearly direct AI toward useful outcomes
Tool selection - knowing which AI tools fit which tasks
Responsible use - understanding limits, risks, and accountability
Co-intelligence - combining human judgment with AI strengths
AI mastery - orchestrating all of the above deliberately
Each layer strengthens the next, and together, they turn individual effectiveness into organisational capability. When one layer is missing, productivity gains still occur, but they remain uneven and difficult to reproduce.
This is why many organisations appear busy with AI yet struggle to see sustained impact. Skills are present, enthusiasm is real, but repeatable productivity has not yet taken shape.
Why this matters for leaders
For leaders, AI capability answers a simple but demanding question:
Are productivity gains from AI repeatable across teams, roles, and time, or do they depend on individual initiative?
When capability is present, AI-supported productivity becomes part of roles rather than personal effort: outcomes are predictable enough to inform strategy, and governance enables speed rather than slowing it down.
When capability is missing, AI adoption feels uncontrolled: leaders oscillate between enthusiasm and caution, and progress depends on heroes rather than systems.
Building AI capability is therefore not about accelerating experimentation, but about creating the conditions for productivity gains to scale safely and intentionally.
A reflection to end the week
Before adding the next AI initiative, it’s worth pausing.
Where do AI-driven productivity gains currently live in your organisation?
With a few individuals, or embedded into roles and routines the organisation can repeat?
The answer reveals far more about readiness than any skills inventory ever could.
This framework is part of a broader playbook I’m developing for leaders navigating AI, the AI Strategy Playbook: From Zero to 100. If you’d like early access to the book and related materials as they become available, you can join the waitlist here.
_
McKinsey & Company — The State of AI
https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-aiOpenAI — The State of Enterprise AI 2025
https://cdn.openai.com/pdf/7ef17d82-96bf-4dd1-9df2-228f7f377a29/the-state-of-enterprise-ai_2025-report.pdfDeloitte — State of AI in the Enterprise / enterprise AI adoption research
https://www2.deloitte.com/global/en/pages/consulting/articles/state-of-ai-in-the-enterprise.htmlMcKinsey analysis (summarised in Innovation Leader) — Employees adopting AI faster than leaders expect
https://www.innovationleader.com/report-tldr/mckinsey-report-finds-employees-moving-faster-with-ai-than-leaders/