Human capability in the age of agentic AI

How AI changes the way we think, learn and develop.

I’m an independent AI specialist and adviser. I help organisations decide how to apply AI and prepare for changes in professional practice, drawing on a career in professional learning and practical expertise in AI agents.

Read my writing How I help

About

How I help

Thinking

Ideas and practical perspectives on human capability, judgement, learning and development in the age of AI.

Showing 26 of 26 articles

More than skills

As AI takes on separable tasks, human development may need to place greater emphasis on integration, judgement, meaning and understanding the whole.

AI agents and the rebundling of knowledge work

As agents take on execution, organisations need to reconsider how people govern work, develop expertise and sustain the relationships on which it depends.

Proportionality in agentic AI

How much analytical separation does a task warrant? A practical framework for choosing between focused work, multiple perspectives and separate agent contexts.

PACED and the expertise paradox

When AI performs the work through which expertise has traditionally developed, how do people learn to judge the result? PACED offers a framework for practice, reflection and delegation.

What is an AI agent harness?

An agent harness structures the work around an AI model, making stages, tools, memory, checks and stopping rules part of a controlled process.

Autonomy expands accountability

AI can expand our responsibilities beyond our existing roles. Professional development must prepare people for the wider judgement and accountability that follow.

AI and meaning

Wittgenstein and language games

When intelligence outpaces understanding

Even AI that works as intended can outpace individual understanding. Human collective intelligence offers a way to question and build on what others know.

What does it take to work well with AI?

Domain knowledge, AI skills and metacognition offer a useful way to think about human agency and collaboration with AI.

What is Context Engineering?

Context engineering is deciding what information an AI system needs and managing it as the work progresses, so outputs are grounded, auditable and less prone to guesswork.

Are neural networks easier to understand than we thought?

New research challenges how we think AI represents information.

Delegating authority to AI agents: what motivation science tells us

Delegating to AI changes how people experience autonomy, competence and connection. Motivation science offers a way to examine the design of agent-supported work.

The Right Amount of Doubt

Aristotle and Mill offer complementary ways to judge when an AI answer needs further challenge, and when it is reasonable to decide and move on.

What happens when AI becomes better at improving AI?

AI is increasingly involved in developing AI itself. Successive improvements bring questions about evaluation, human authority and whose interests future systems serve.

Can AI discover new knowledge?

AI inherits our knowledge and assumptions. Can the connections it finds lead to knowledge we do not yet have?

Where does intelligence happen?

Intelligence may emerge partly through a system’s behaviour and partly through another system’s interpretation of it. What might that mean for how we understand AI?

AI and the Attention Schema

Attention schema theory offers a way to examine the distinction between computational attention and a system’s ongoing model of its own attention.

Does AI have to be conscious to act conscious?

As AI becomes better at monitoring and describing its own activity, can we distinguish self-aware behaviour from subjective experience?

Channelling Intelligence: The Primacy of Process in AI Agents

Process philosophy suggests a way to understand an AI agent as an organised flow of context, instructions, tools and feedback, rather than a fixed entity.

Thinking Fast and Slow

Jev is designed for fast, structured decisions. How could it work alongside LLMs in agentic systems?

System 0 Thinking

AI systems are moving upstream into the layer that filters and structures information before it reaches conscious attention, shaping what becomes thinkable.

When AI Keeps a Playbook

As AI assistants gain memory and optimisation, personalisation can shift from helpful recall to a tailored playbook for influencing behaviour.

Untangling Agentic AI Protocols

A growing set of protocols defines how agents connect, coordinate, transact and pay, with different roles for tools, communication, commerce and payment authorisation.

Nested Learning - A New Paradigm For Continual Learning

Google Research’s “Nested Learning” work proposes treating a model as a set of nested optimisation problems, each with its own context flow and update rate, to address catastrophic forgetting.

Agentic Context Engineering (ACE)

ACE is a method for making agents learn from experience by improving their context and playbooks, rather than retraining the model.

AI and the power to contribute

AI can help people turn useful ideas into things they can try, examine and improve, opening more routes into contribution and personal development.