For much of the past year, I have described our direction as a human-agent operating model. The phrase remains useful. It signals that agents will increasingly participate in work and that humans must continue to lead, govern and own the outcomes.
I no longer think it describes the destination.
The destination, if there is one, is a continuously learning organisation. One that can understand how its work actually flows, measure whether it is producing the intended outcomes, identify where the system is constrained and redesign itself without waiting for a crisis or a major technology programme.
A human-agent operating model is a waypoint because agents can help create that learning loop. Once processes, decisions, controls, and outcomes are sufficiently codified, AI may observe patterns across the operating model that no individual can see. It may identify duplicated work, predict emerging bottlenecks, simulate alternatives and recommend where agents could carry more of the execution load.
That is not a self-managing organisation. The accountable human remains central. A process owner may approve a contained change within clear guardrails. A Value Stream Council may consider a change with broader consequences. A Value Stream Owner may decide where interests conflict. AI improves the visibility and the options. It does not inherit the authority.
The foundations build on one another. Psychological safety allows people to expose what is really happening. Business architecture turns that truth into connected knowledge. Operating-model design clarifies outcomes, governance and decision rights. Specification-led development makes intent executable. Agentic architecture distributes work across specialised systems and human judgement. Scaffolding keeps those systems safe, observable, and controlled.
Then the cycle begins again. Evidence from operation reveals the next opportunity. People challenge the recommendation. The accountable owner decides. The system learns from the result.
When I began this journey, I was focused on safely scaling AI. I still am. But the more I learn, the more I am convinced that there are opportunities for exponential increases in productivity and quality of outcomes, while the challenge to realising that value sits in our culture and operating model.
The competitive advantage may not come from deploying the most agents. It may come from building an organisation capable of telling itself the truth, learning from what it sees, and redesigning itself responsibly.
The human-agent operating model was never the destination. It was how we began learning to become that organisation.