--- title: Atoms of Enterprise AI Deployment source: https://steadman.ai/newsletters/david/atoms.html published: 2026-05-05 summary: A reader's guide to the interactive Atoms diagram. Twenty-four atoms, one hundred and forty-five bonds, and a seven-link chain. --- # Atoms of Enterprise AI Deployment A reader's guide to the interactive Atoms diagram. The diagram is a working map of how an enterprise AI rollout actually behaves. This companion explains what's in it and why each piece sits where it does. ## What the map is Most enterprise AI rollouts treat themselves as a list of activities. Senior coaching done. Training delivered. Tools rolled out. Activities tick along while nothing changes. The map below treats deployment as a chain instead. Twenty-four atoms. One hundred and forty-five bonds. Seven links along a spine: a map of the dependencies between adoption and value capture, not a fixed order. Every link is only as strong as the audience it actually reaches. Business outcome, baseline, owner, selected workflows, capacity captured: that strip sits above the diagram. The map covers adoption and the changes needed to capture value in the existing business. New products and revenue are a separate kind of work, shown in the [Value Map](https://steadman.ai/newsletters/david/ai-value-map.html). Name the business outcome first, and [name the decision-makers and owners](https://steadman.ai/newsletters/david/five-questions.html); then inspect the workflow and role changes needed to achieve it, and the dependencies that might prevent them. It isn't a model that predicts outcomes. It's a structured way to argue about them. Every bond is author judgement, not a regression coefficient. Hover any atom to read its definition, its failure mode, and the diagnostic question that tells you whether it's broken. ## How to read it Three kinds of atoms make up most of the diagram. Brown rounded squares are AI transformation initiatives: senior coaching, team training, the strategy review, embedded transformation support, system builds, the weekly email cadence, the skill-building day, peer-to-peer learning, and the reverse apprenticeship pairing. Deep navy rectangles across the middle are the seven chain links. Blue circles below the spine are the audiences each link must reach. Arrows are colour-coded by source band. Brown arrows leave the AI transformation initiatives. Blue arrows leave the audiences. Dark grey, dashed arrows leave the external inputs along the top. The dark navy arrows along the spine are the chain itself: each link gates the next. The dashed boxes along the top are external pressures: model releases, vendor pricing, competitor moves, regulation. They feed into the weekly email, which translates them into the rollout. ## What's in the chain L1 Executive Sponsorship is the start. Senior leaders use the tools personally on real work. They name the shift instead of euphemising it. If the top three people in the firm aren't using AI weekly, link one is broken and nothing downstream holds. L2 Application Approvals is where most rollouts run aground first. Tools have to be capable, secure, and approved fast enough that they don't go obsolete during procurement. The shadow AI economy, where everyone uses personal accounts because the firm hasn't approved anything good, is the visible signature of a broken link two. L3 Foundational Skills is training that earns the next step. People are taught to use the tools competently. Confidence becomes curiosity and curiosity becomes investment. Most users never set custom instructions, never look at settings, never learn what a context window is. Five workshops without reinforcement and the skills fade by Thursday. L4 Standardised Tasks turns private craft into shared assets. Custom instructions, prompt libraries, reusable skills, evaluation patterns. Done well it compounds permanently. Done badly it produces a hundred and fifteen custom GPTs nobody uses. L5 Process re-engineering is the most-broken link in practice. Whole workflows run end-to-end with AI in the loop. Humans shift from doing to specifying, reviewing, intervening. Most firms bolt AI onto existing steps and call that orchestration, then wonder why the gains are twenty per cent rather than seventy. Process reinvention earns the larger numbers. It also breaks five layers of management, which is why most firms don't get there. L6 Sustained Reinforcement is monthly sharing meetings, Friday stand-ups, always-on training, custom instructions that compound. Without it, link five quietly reverts. Three weeks after senior team training, many people drop back to where they started. L7 Roles and teams is the link the chain ends on. Role and team changes are planned alongside the redesigned workflow, so that created capacity reaches the intended outcome; [explore a team design](https://steadman.ai/newsletters/david/ai-native-team.html). The org chart finally moves to fit what AI now does. Layers flatten. Junior pipelines get rebuilt, not abandoned. Pay and promotion track the new shape. Most firms haven't started this and won't admit it. ## What's not in the map Earlier drafts carried more layers of synthesis that were useful in working out the model but added clutter on the page. The live version simplifies to the chain plus the audiences who unlock each link, plus the AI transformation initiatives that drive each link, plus the four external pressures that feed into the weekly email and from there into the rollout. If you want a capability profile across the five assessment areas, the Organisational AI Diagnostic does that job on its own page. ## Try it Hover any atom to see the bonds it sits inside. Click any atom to pin it. The inspector below the chart shows what it is, what it connects to, and the notes that didn't fit the spine. Hover any bond to see its source and type. A useful starting move: name the business outcome first. Then click L5 Process re-engineering to read why it breaks for most organisations, and L7 Roles and teams, the link the chain ends on. Hover the bonds running into each to see what's true about every connection. If you want to use this in a workshop, a board paper, or a discovery call, the page exports a clean PNG of the map. The schema lives in two JSON files alongside this page. Write back if you find an atom missing, an edge that should exist, or a bond that's typed wrong. The next version is one synthesis pass away.