--- title: "AI: From What's True to What to Do" source: https://steadman.ai/newsletters/david/ai-from-whats-true-to-what-to-do.html published: 2026-04-05 summary: A structured argument for leaders that builds from three facts about what language models can do today, follows their implications to a decision only a leader can make, and maps the consequences that follow regardless of which route you choose. --- # AI: From What's True to What to Do *Steadman. A Framework for Leaders. 5th April 2026.* This is a structured argument, not a briefing. It starts from three facts about what language models can do today, follows their implications to a decision only a leader can make, and maps the consequences that follow regardless of which path you choose. Each claim is tagged with its epistemic role — axiom, implication, your call, consequence, or recommendation — so you can see where the logic depends on evidence and where it depends on you. **Epistemic labels used throughout:** - **Axiom**: a statement accepted as true, serving as a starting point for reasoning. - **Implication**: a conclusion that follows logically from something already accepted. - **Your Call**: a decision, especially one that rests with a single person. - **Consequence**: a result or effect of an action or condition. - **Recommendation**: a suggestion or proposal as to the best course of action. --- ## I. The Case — what's true today The opportunity builds from this. ### [Axiom 01] Language models read, think, and write well. *axiom, noun. a statement accepted as true, serving as a starting point for reasoning.* Three core skills: **comprehension** (reading text and images), **synthesis** (reasoning, analysis, connecting ideas), and **writing** (including code, and therefore tool use). "Well" matters. The output is already useful, often good, sometimes excellent, today. ### [Axiom 02] They loop to handle complex work. A model reads, thinks, writes. Then reads its own output, thinks further, writes again. Research, apps, analytics, computer control: all are read-think-write loops applied to different domains. *This looping behaviour is what defines the [Generation 2 tools](https://steadman.ai/newsletters/david/gen1-vs-gen2.html) like Claude Code, as distinct from the Generation 1 chatbots most people first encountered.* ### [Implication 01] Good first drafts of almost anything, quickly and cheaply. *implication, noun. a conclusion that follows logically from something already accepted.* "Almost anything" because the loop covers most knowledge work. "Quickly" because compute is fast. "Cheaply" because the marginal cost is negligible versus a human doing the same task. "First draft" because human judgement still owns the final output. The CEO principle: Check, Edit, Own. ### [Implication 02] This changes everything. Not a slogan. A logical consequence. If you can get a good first draft of almost any knowledge task, at near-zero marginal cost, then the economics of every knowledge-work function are being rewritten. This is true even if capability froze tomorrow. --- ## II. The Question — only the leader can set this direction. ### [Your Call 01] What do you want to do with AI? *call, noun. a decision, especially one that rests with a single person.* Everything downstream depends on it: which roles change, which processes to redesign, what to build. And the decision requires genuine knowledge of what AI makes possible, which is why we began by laying out the facts. Two routes (or a blend): - **Extraction — Same things, done better, quicker and happier.** Become more efficient and effective at what you already do. Do you get smaller, stay the same size at higher quality, or grow within existing markets? - **Expansion — Different things, now viable.** AI makes some products, services, and markets economically feasible that weren't before. Too hard, too expensive, too complex. A growth path into new territory. --- ## III. What You'll Encounter When You Act — four consequences, regardless of the route ### [Consequence 01] Roles shift: from doing, to setting up and judging. *consequence, noun. a result or effect of an action or condition.* If AI produces good first drafts, the bottleneck moves from production to direction and evaluation. Everyone becomes an orchestrator of AI output: senior leaders, the middle, junior recruits in their first week. We are all managers now. Managers of AI agents. ### [Consequence 02] Judgement is still learned, but learned differently. Traditionally, judgement came slowly, from years of doing the work yourself. In an AI-augmented world, [it is learned](https://steadman.ai/newsletters/david/weve-had-this-conversation-before.html) more quickly through setting up and evaluating, not through doing. A different acquisition path, not a loss. ### [Consequence 03] Roles, teams and processes must change to capture the benefits. Without structural adjustment, productivity gains get absorbed. People do the same work to a higher standard, but the organisation does not capture the value. This is an organisational design problem, not a technology problem. If you don't reorganise roles, teams, and processes, the bottleneck just shifts. Speed up one step and the constraint moves elsewhere: to review queues, to approval chains, to the people making decisions, to the hand-offs between teams. ### [Consequence 04] Getting organisational results is hard and complex. The honest counterweight. AI touches every function, requires new ways of working, and demands change across multiple dimensions simultaneously. Realism, not pessimism. --- ## IV. The Urgency — and it is getting harder ### [Axiom 03] Capabilities are expanding. The frontier keeps moving, rapidly. This is not about current AI being inadequate. It is already good enough to change everything. The point is that models become more capable on multiple fronts every month, and costs continue to fall. The two compound. **Task → Process.** The viable unit of work is shifting. What once required a single prompt-and-check can now run as an extended, multi-step workflow. This shift will continue. - The opportunity is growing. What is possible tomorrow exceeds what is possible today. - The competitive pressure is growing. Competitors, clients, and markets are adapting. - Plans must be adaptive. A static view of AI capability is already wrong by the time you act. - The window for advantage (or not falling behind!) narrows the longer you wait. --- ## V. The How — everyone on the journey, in sequence We could write an entire guide on how to approach AI transformation, and we are happy to do so if that would be useful. [Ask us for it.](mailto:david@steadman.ai?subject=I'd%20love%20to%20see%20the%20AI%20transformation%20guide) For now, we want to point out two high-level implications we see people commonly missing. ### [Recommendation 01] Everyone needs to come on the journey. *recommendation, noun. a suggestion or proposal as to the best course of action.* Not just a technology team. Not just early adopters. AI touches every knowledge-work role, so adoption must be organisation-wide. This is a people challenge as much as a technology one. Once you have given people the chance, the encouragement, training, and the support, tough conversations are necessary for people who are not adopting AI. ### [Recommendation 02] Sequence: individual productivity, team standards, process orchestration, role and team redesign, new revenue. Build foundational capability in individuals first. Then encode what works so teams share it. Then wire AI into how the business runs. Then change the shape of the organisation to capture the value. Only then reach for the genuinely new. Each step depends on the ones before it. 1. **Individual Productivity** — Foundational skills for day-to-day work. 2. **Team Standards** — Encoding what works so the team shares it. 3. **Process Orchestration** — Wiring AI into how the business runs, end to end. 4. **Role And Team Redesign** — Changing the shape of the organisation to capture the value. 5. **New Revenue** — Products and services that only become viable once the economics shift. Use the [diagnostic](https://steadman.ai/newsletters/david/ai-maturity-diagnostic.html) to find which phase your organisation has reached, and the [value map](https://steadman.ai/newsletters/david/ai-value-map.html) to see where the prize sits. --- ## Infographic summary A condensed version of the full argument, as presented in the downloadable one-page infographic. **I. The Case** - Axiom 01: Models read, think, write well. Comprehension, synthesis, writing. Already at useful levels. - Axiom 02: They loop to handle complex work. Read-think-write cycles span research, analytics, apps, computer use. The signature of Gen 2 tools like Claude Code. - Implication 01: Good first drafts of almost anything. Quickly, and cheaply versus a human doing the same task. We still Check, Edit, Own. - Implication 02: This changes everything. Knowledge-work economics are being rewritten, even if capability froze tomorrow. **II. The Question** - Your Call 01: What do you want to do with AI? Extraction (same things, better, quicker, happier) or Expansion (different things, now viable)? Only the leader decides. **III. What You'll Encounter** - Consequence 01: Roles shift: doing to setting up and judging. Everyone orchestrates AI output, from juniors to the C-suite. We are all managers now. Managers of AI agents. - Consequence 02: Judgement still learned, but learned differently. More quickly through setting up and evaluating, not through doing. - Consequence 03: Roles, teams, processes must change. Without redesign, productivity gains are absorbed. Bottlenecks just shift. - Consequence 04: Getting organisational results is hard and complex. Touches every function. Realism, not pessimism. **IV. The Urgency** - Axiom 03: Capabilities expanding. Frontier moving, rapidly. Task to Process. Opportunity grows, pressure grows. The window for advantage (or not falling behind) narrows. **V. The How** - Recommendation 01: Everyone on the journey. Organisation-wide. After support and training, tough conversations for non-adopters. - Recommendation 02: Individual Productivity → Team Standards → Process Orchestration → Role And Team Redesign → New Revenue. Individuals build skills. Teams encode standards. Processes get wired end to end. The organisation reshapes around what AI now does. Then new revenue. --- *Feedback, ideas or questions? Email [david@steadman.ai](mailto:david@steadman.ai)* *Steadman helps organisations use language models to work better, quicker and happier.* *Interactive walkthrough: [steadman.ai/newsletters/david/ai-from-whats-true-to-what-to-do.html](https://steadman.ai/newsletters/david/ai-from-whats-true-to-what-to-do.html)*