Start here
Not with a tool. Not with a vendor. With one workflow you can map, measure, and prove — and the people who will run it after you. This page routes you to the three or four things that matter for where you actually sit.
Three pieces, in this order
These are the ones that get quoted back to us most often — by practitioners, by boards, and by the internal teams who inherited the work six months later.
-
The roadmap — five cumulative stages
Why it's first: the model is the map. Pilot, Expand, Executize, Transform, Autonomate — and the reason skipping a stage costs more than it saves. Fifteen minutes, and every other page on this hub will make more sense.
-
The pilot method — instrument one workflow
Why it's second: ninety percent of organizations are somewhere in the first two stages. The Pilot method is the artifact that makes the rest provable — data in, transforms, integrations, outputs, and the HITL rate.
-
Why it's third: the second workflow is when risk catches up with you. Tiered review — sized to the risk of each use case — is what makes scale possible instead of fragile. Read before you need it.
You know who you are in the organization
Each path below gives you the first three moves, the weekly habit that makes them stick, and what "good" looks like in 90 days. Different roles need different first moves — the model is the same, the entry point isn't.
Executives & boards
You're accountable for direction, investment, and risk appetite — and you're going to be asked by your board what you're doing about AI.
Pick the one workflow your board will ask about in Q3. Write down the number it has to move. Fund that workflow — not a platform.
- 01The roadmapThe model your leadership team will keep coming back to.
- 02Strategy playbookConnecting AI investment to the goals your board already tracks.
- 03The pilot methodWhat a legitimate pilot needs before you fund it.
One funded pilot with a documented baseline, a board narrative that isn't about tools, and a decision log that would survive an audit. Fewer dashboards, more numbers you can defend in a budget review.
Transformation & strategy leads
You own the portfolio, the operating model, and the business case. You're the one making the pattern library real — even before it has a name.
Before your second workflow goes into production, outline the pattern library. The pattern is the deliverable — the workflow is only proof.
- 01Expand — the pattern libraryHow the second workflow becomes cheaper than the first.
- 02The roadmapThe substrate argument, stage by stage.
- 03Use-case scorecardRank candidate workflows by value, feasibility, and risk.
Two more workflows running on the same pattern as the first, at measurably lower cost to launch. A written operating model that names operators and strategists as roles — even if the seats are still empty.
Function leaders
You own the workflows. You know which ones are slow, information-heavy, or repetitive — and you're the one whose numbers will move when this works.
Walk your slowest workflow. Map the data in (start and mid-stage), the transforms, the integrations, the outputs. That map is worth more than any vendor comparison.
- 01Pilot — the full methodWorkflow map, baseline, automation posture, HITL rate.
- 02Pilot Plan WorksheetScope, baseline, success metrics, go/no-go — one page.
- 03Use cases by functionWhat other teams in your function have already done.
One workflow mapped, baselined, and running with a measured HITL rate you can show your CFO — not a subjective "it feels faster." A go/no-go decision you can defend, in writing.
IT, data & engineering
You're asked to make AI work with the systems you already have. The integrations — in and out — are where it succeeds or fails.
Pick the integration that will bite first. Make it reusable — the second workflow will need it too, and so will the coordination layer after that.
- 01Expand — reusable integrationsWhy the second workflow shouldn't touch the same systems twice.
- 02Executize — the coordination layerRegistry, controller, decision log. What to build, in what order.
- 03Governance playbookTiered review that doesn't become an approval gate.
Reusable integrations for the first two workflows, a registry a coordination agent could actually read, and an access model that survives a security review without becoming a workaround.
Risk, legal & compliance
You're asked to enable AI without enabling the next incident. The answer is tiered review — not a blanket ban, and not a rubber stamp.
Write the tiered review model before the second workflow is attempted. Retrofitting risk after the fact is the expensive way — and the way that fails audit.
- 01AI governance playbookTiered review sized to the risk of each use case.
- 02Autonomate — the tiersL0 through L4, with the guardrails required at each.
- 03Executize — the decision logWhat an auditable coordination layer requires.
Tiered review live, an audit trail in place, and a documented veto pattern that doesn't stall teams. Risk seen as a design partner, not a phase gate.
HR & people leaders
You're asked to prepare skills, roles, and change management — and to name the roles that didn't exist a year ago.
Name the operator and strategist roles for the first workflow. Even if the seats are unnamed hires, name the roles themselves — the org chart follows the language.
- 01Expand — operators and strategistsWhere the new roles actually get defined.
- 02AI adoption playbookFrom power users to operators of the workflow.
- 03Transform — new functionsWhen role charters become function definitions.
Two new role charters approved, a training path attached to each, and an internal communication that names the change plainly. No euphemisms, no "digital fluency" language — concrete roles, concrete skills, concrete owners.
You know where you are
Each stage is a specific piece of substrate. Pick your stage for the shortest route forward — and the advance condition that tells you when you're ready for the next one. If you can't tell which stage you're in, take the diagnostic.
- Stage 01 Pilot One workflow, mapped end to end. Baseline set. Automation posture chosen. HITL rate measured. The artifact the next four stages never stop using.
- Stage 02 Expand One pattern, many workflows. Operators and strategists named. Launch velocity falling. This is where the operating model gets built, not just the tech stack.
- Stage 03 Executize A registry, a coordination agent, and a decision log. The layer most organizations never build — and the one that caps Transform and Autonomate if missing.
- Stage 04 Transform New workflows, roles, and functions — producing value that isn't attributable to an old workflow running faster. The hardest stage to do, and the easiest to fake.
- Stage 05 Autonomate Deep autonomy, tiered. L0 through L4. Kill switches, evaluation ceilings, quarterly review discipline. The frontier — deliberately reached last.
Not sure which stage you're in?
The diagnostic on the roadmap page takes 60 seconds. Five questions, no signup. It tells you the stage where every answer was yes — and the specific next move.
How to use this page
If I only have one hour, what should I read?
Read the roadmap first — it takes about 15 minutes and gives you the mental model everything else depends on. Then read the Pilot method for the stage most organizations start in. If you have a further 30 minutes, skim the governance playbook, because the second workflow is where most programs get caught, and knowing the shape of tiered review early saves the retrofit later.
Should we hire an AI team before we start?
No. Hire for one workflow first. Most organizations need one AI engineer embedded with a function leader, not a central team — until the pattern library exists in Expand. A central team built before the first pilot tends to become an approval gate, not an accelerator.
Where does governance fit — is it a stage?
Governance is not a stage. It runs across all five. At Pilot it's a named reviewer with a documented approval step. At Expand it's tiered review sized to risk. By Autonomate it's per-tier guardrails, kill switches, and quarterly reviews. It grows with the model; it doesn't wait for one.
What if our leadership wants to skip Pilots and go straight to transformation?
Show them the contracts table on the roadmap page. Transform requires new-workflow attribution — a business case that isolates value from new work, not old workflows running faster. You cannot attribute new-workflow value without a baseline. Without a Pilot, Transform has no accountant.
What is the difference between this page and the homepage finder?
The homepage finder asks two questions and gives you three links in 20 seconds. This page is the destination — each role gets a specific weekly tactic, three sequenced links, and a 90-day expectation. The finder is the doorway; this page is the room.