I like an axiom, one thing simple to recollect, quick to say, and punchy sufficient to stay. With my older boys, I’ve typically mentioned, “If doubtful, don’t.” With my youthful son, who’s autistic, I say, “Keep shut; keep protected.” These are quick phrases with large truths, the sort that helps within the second when time is brief and the stakes are excessive.
That’s in all probability why concepts like Moore’s legislation, Amara’s legislation, and Parkinson’s legislation proceed to resonate with expertise leaders. They assist us maintain on to easy concepts as we make sense of tech adoption, worth, and implementation at scale. They’re a part of how we keep the course and maintain our individuals anchored when the media shrill rises, the consultants’ decks get thicker, and LinkedIn pundits fill our feeds with certainty. And that brings me to Conway’s legislation.
Platform Alternative Isn’t The Beginning Level
I’ve been presenting a number of instances per week to public-sector purchasers a workshop session entitled “Overcoming The Hurdles Of AI Adoption At Scale.” Not too long ago, a shopper mentioned, “Conway’s legislation performs out each single time. We need to implement techniques earlier than addressing our enterprise … and each time we find yourself with the identical outcomes, in that the techniques find yourself simply as tousled as our organizations.” In different phrases: Begin with the working mannequin and organizational construction, then orient platforms to the suitable domains.
One of many nice errors in AI proper now could be the assumption that the reply lies primarily in choosing the proper platform, mannequin, or vendor stack. It doesn’t. If the working mannequin is unclear, fragmented, or constructed for an earlier period of labor, the AI system will inherit these flaws and reproduce them at machine pace. That’s why Conway’s legislation feels so related once more: Programs don’t transcend organizations — they mirror them. And within the age of agentic AI, they amplify the worst of them: the silos, the politics, and extra.
Begin With Your Group And Your Individuals
This level sits on the coronary heart of what we’re doing with our analysis into the cognitive working mannequin, intelligence enterprise, and skills-oriented structure. And the core premise of this analysis is the AI productiveness paradox: Positive aspects dissipate inside working fashions designed for human-only, task-based work. Bolting brokers onto yesterday’s roles, workflows, and resolution rights is expertise deployment with higher advertising from corporations that want to maximise IPO valuation to get the capital wanted to feed the AI money furnace.
That’s additionally why the shift from generative AI to agentic AI issues a lot. GenAI was the warm-up: Agentic AI modifications the sport as a result of we transfer from prompts to plans. These techniques now retrieve, resolve, set off, notify, and act. That shifts the dialog from output high quality to governance, accountability, orchestration, and legitimacy — particularly in authorities, the place explainability, equity, and public belief are nonnegotiable.
The Working Mannequin Shift Issues
In case your working mannequin is siloed, fragmented, overloaded with handoffs, and constructed round a human-only conception of labor, your AI property will replicate that complexity. Brokers will likely be chosen, deployed, and ruled based on those self same fault strains. The outcome? Duplicated capabilities, fragmented context, inconsistent controls, and level options masquerading as transformation.
What Conway’s legislation explains is why the working mannequin shift is so central. At its core, agentic AI is a piece structure downside and an working mannequin shock. If brokers more and more develop into the default executors of routine cognitive work, then the group should be redesigned round that actuality. Roles, workflows, escalation paths, administration assumptions, and accountability fashions all change. In any other case, the expertise will merely automate the archaeology of right now’s enterprises.
The Expertise And Context Matter
This is the reason our work encourages our purchasers to maneuver away from use-case considering and towards abilities because the atomic unit of design. A use case describes an issue to unravel. A talent describes a bounded cognitive functionality that may be reused, ruled, and composed throughout roles and workflows. Manage agentic portfolios round remoted use circumstances, and also you get fragile, siloed deployments that resist scale. Manage round abilities, and also you create the situations for composition, governance, and sturdy operating-model change: dynamic, agile, and versatile.
The opposite half of that is context. Functionality by itself isn’t sufficient. Actual competence will depend on the encompassing semantic layer of coverage, vocabulary, reminiscence, resolution traces, tacit data, and organizational logic. And not using a coherent strategy to floor and govern context, agentic techniques will mirror the enterprise’s lacking data, fragmented coverage interpretation, weak accountability, and rising prices.
Conway’s Legislation Issues
If I needed to flip Conway’s legislation right into a sensible guidelines for leaders within the age of agentic AI, it will be this:
- Begin with the working mannequin. Let the platform comply with the work, the issue domains, and the outcomes the group wants to realize.
- Construct reusable organizational capabilities. Design abilities, roles, workflows, and governance buildings that compound throughout use circumstances.
- Deal with context as organizational intelligence. Make coverage, data, reminiscence, and resolution logic machine-readable, governable, and out there on the level of labor.
- Design brokers across the group you need to develop into. Brokers amplify the system they function inside, together with its strengths, gaps, and accountability mannequin.
For me, that’s the trendy worth of Conway’s legislation. Within the dizzying storm of change we’re in, if we wish agentic AI to create compound worth, we should first redesign the working mannequin that surrounds it. That’s the work, that’s the hurdle, and that’s why our present analysis is so targeted on construction, context, and the redesign of labor itself. In any other case, we aren’t constructing the way forward for work — we’re automating the previous. So bear in mind, youngsters: “Working fashions ship outcomes.”
