AI capability transformation
Every business now runs two workforces - the human one and a growing agentic one. I map capability across both, decide what to automate, augment, and protect, and build the frameworks that make each better.
Start a conversationCapability stopped being about people
Organisations have mapped workforce capability for decades: who can do what, where the gaps sit, what training or hiring closes them. It’s a mature discipline with a mature language.
What’s changed is the population being mapped. Every business now runs two workforces, whether it knows it or not - the human one, and a growing agentic one of AI systems doing real work. Both have capabilities. Both have gaps. Both fail in characteristic ways. And most organisations are managing one with a mature discipline and the other with guesswork.
The question is no longer “how do we train people to use AI?” It’s “what does capability mean across a workforce that’s part human, part agentic - and how do we make both better?”
What I actually do
Capability mapping, both workforces - The same discipline, applied across people and agents: what each can do, where the gaps sit, and what closes them. For people that might be development or hiring. For agents it’s usually context, not rebuilding - when an agent fails at a simple task, the instinct is to retrain or re-engineer, but more often it simply didn’t have what it needed to do the job. It’s the same mistake organisations make with people: retrain the person, when the actual failure was the brief.
Capability architecture - Deciding which work should be automated, which should be augmented, and which should be protected for humans - the Automate/Augment/Protect framework, applied with discipline.
Learning design - Building programmes that use AI as a scaffold, not a substitute. A scaffold helps people build capability; a substitute replaces it, and over-reliance makes teams weaker.
Metacognition frameworks - Teaching teams to think about their own thinking in an AI-augmented environment. When do you trust the output? When do you push back? How do you keep critical judgement sharp when the machine does the heavy lifting?
Measurement & evaluation - Most capability metrics measure consumption, not capability. I build evaluation frameworks that tell you whether your people and your agents are actually getting better, not just completing modules or closing tickets.
The framework: Automate, Augment, Protect
This is the core of how I think about capability in an AI-integrated business:
Automate - Work AI can handle reliably. Procedural knowledge, reference lookups, compliance updates, administrative tasks. Free the cognitive bandwidth for higher-value work, and hold the agentic workforce to the same capability standards you’d hold a team to.
Augment - Complex work where AI makes humans better without replacing the human contribution. Research synthesis, draft generation, scenario modelling. The human still owns judgement and decisions.
Protect - Capability-building where the cognitive effort IS the learning. Critical analysis, creative problem-solving, ethical reasoning, interpersonal skills. If AI does this for your people, they don’t learn - and your organisation gets weaker.
Who this is for
- Boards and leadership teams who need a strategic approach to a workforce that’s now part human, part agentic
- CHROs and L&D directors whose capability remit just doubled
- Agency and consultancy leaders building AI-native service delivery without hollowing out their teams
- SME founders who need small teams to be brilliant with AI, not dependent on it
Is this just AI training?
No. AI training teaches people to use tools. This is about mapping and transforming capability across both of your workforces - the human one and the agentic one - and redesigning how each learns, works, and improves. Tools are part of it, but not the point.
What does 'agentic workforce' actually mean?
The AI agents, automations, and model-driven systems that now do real work inside your business - drafting, analysing, monitoring, producing. They have capabilities and gaps exactly as people do, they fail in characteristic ways, and the capability-mapping discipline that took decades to mature on the human side transfers across surprisingly well.
How long does a typical engagement take?
Capability mapping and framework design typically takes 6-8 weeks. Implementation support varies by organisation size. I also run intensive workshops for leadership teams (1-2 days) to align on the strategic approach before committing to a full programme.
What makes this different from other AI capability consultants?
Most focus on tool adoption - getting people to use AI more. I focus on capability architecture across both workforces: keeping people sharp rather than tool-dependent, and giving agents the context they need rather than endlessly rebuilding them. The Automate/Augment/Protect framework gives you a decision-making structure, not a training plan.
