← AI Adoption Cycle 🕐 16 min read
AI Adoption Cycle

Rewired Transformation Roadmap Template

Brandon Sneider · April 2026

The roadmap assumes these three decisions have been made before Month 1. If any are unresolved, that resolution is Month 1 and the roadmap below shifts accordingly.

12-month transformation roadmap keyed to Rewired’s six-capability sequencing logic. Designed to hand to a PMO. Each quarter has: focus capability, milestone, resource requirement, primary risk, and how to measure.

Grounded in: Lamarre, Smaje & Levin, Rewired 2nd ed., 2024; NBER (n=5,867, Feb 2026); BCG (n=640, Jan 2026); MIT CISR (n=721); Stanford Enterprise AI Playbook 2026 (n=51 deployments).

Before Starting: Three Preconditions

The roadmap assumes these three decisions have been made before Month 1. If any are unresolved, that resolution is Month 1 and the roadmap below shifts accordingly.

  1. Named domain. One business domain is selected as the primary transformation target for Year 1. Rewired Ch. 3 (p.49): domains, not use cases. The domain must be large enough to affect a P&L line and small enough to own end-to-end. Examples: commercial claims processing, customer acquisition for a specific segment, supplier-risk qualification.

  2. Named domain owner. A business leader — not the CIO or CAIO — owns the domain outcome and has authority to change the workflow, allocate time, and hold vendors accountable. Rewired Ch. 4 (p.63): “have business leaders lead the reimagination.” This is the Air Canada model. If the domain owner is IT, the roadmap produces IT deliverables, not P&L impact.

  3. Production path committed. A 12-month production path is funded and documented before the pilot launches. S&P Global (n=1,006, 2025): 46% of AI POCs are scrapped before production because the production budget requires a separate business case that pilot teams never build. Commit the production path before the pilot starts, or the pilot is a research project.

Quarter 1: Foundation — Strategy, Data, Talent Baseline

Focus Capabilities: Capability 1 (Strategy & Roadmap) + Capability 4 (Data Backbone — assessment only)

Rewired anchor: Ch. 3 (domain selection), Ch. 25 (data architecture precondition), Ch. 9 (talent triage)

Milestone 1.1 — Domain Brief (by end of Week 4)

The domain owner produces a one-page domain brief covering: (a) the P&L line this domain affects and its current baseline; (b) the 3–5 specific workflows within the domain that AI could affect; © the economic case for the #1 priority workflow (current cost per cycle × annual volume); (d) the named production path and budget range.

Why this comes first: BCG AI Radar 2026 (n=640 CEOs): only ~15% of CEOs apply AI end-to-end; the gap between aspiration and execution starts with domain selection. McKinsey Manifesto (n=20 leading companies, Apr 2026): organizations concentrating on 1–3 domains with full workflow reinvention delivered $3 EBITDA per $1 invested vs. scattered use-case portfolios.

Resource requirement: 0.25 FTE domain owner + 0.25 FTE strategy or transformation lead + 1 structured session (3–4 hours) with the CIO, domain owner, and CFO or P&L sponsor.

Risk: The domain selected is driven by technology enthusiasm rather than P&L impact. Mitigation: the economic case must include a counterfactual — what the baseline cost is today, measured, not estimated.

Measure: Domain brief is signed off by both the domain owner and the CFO or P&L sponsor. If either does not sign, the domain is not selected — it is still under discussion.

Milestone 1.2 — Data Architecture Classification (by end of Week 8)

For the #1 priority workflow identified in the Domain Brief: run the data reset decision tree and classify the workflow as Path A (Proceed), Path B (Prepare, 2–4 weeks), or Path C (Reset, 9–24 months).

Why this cannot wait: Gartner (2026): 60% of AI projects abandoned through 2026 due to lack of AI-ready data. Pertama Partners (n=2,400+, 2025–2026): organizations that complete a formal data readiness assessment before approval achieve 47% production success rate vs. 14% without. The data architecture classification determines whether the production timeline on Q2’s deployment is 10 weeks or 18 months. Getting this wrong is the most expensive mistake in the roadmap.

Resource requirement: Data engineer or architect (0.5 FTE, 3 weeks) + workflow owner (0.25 FTE for data mapping sessions) + IT infrastructure lead for access and lineage questions.

Risk: The classification reveals a Path C requirement that delays the 12-month roadmap. Mitigation: if Path C is the result for the priority workflow, immediately evaluate the scope-reduction alternative (Can a single-domain version of this workflow deliver 80% of the value and qualify for Path A?). See data reset decision tree — “scope reduction” section.

Measure: Decision tree output is documented: Path A/B/C, domain count, primary data structure finding, timeline estimate, and a named decision on scope-reduction evaluation.

Milestone 1.3 — Talent Triage (by end of Week 10)

Identify the roles required for the priority workflow deployment: ML engineer, data engineer, workflow designer (business-side), eval engineer or equivalent. Classify each as: In-house available / Contract within 60 days / Gap requiring 3+ month hire or SI engagement. Rewired Ch. 9 (p.141): triage, not headcount reduction.

Resource requirement: CHRO or talent lead (0.5 FTE, 2 weeks) + CIO for technical role calibration.

Risk: The talent gap is larger than expected and creates a 3–6 month dependency that delays Q2. Mitigation: for mid-market organizations, the SI + platform path replaces platform-build roles; the talent gap is a build-vs-buy decision, not a hire-everything mandate. See Tension 3 in synthesis.

Measure: Role inventory is complete. Every role in the deployment org chart is classified. No role is listed as “TBD.”

Q1 Checkpoint (end of Month 3):

  • [ ] Domain brief signed by domain owner + P&L sponsor
  • [ ] Data architecture classification complete and documented
  • [ ] Talent gaps identified and resolution path committed
  • [ ] Production path is documented and funded (not just verbal commitment)

If any Q1 checkpoint item is not complete, do not start Q2 deployment work. The NBER (n=5,867, Feb 2026) finding that 89% of firms report zero productivity impact is substantially explained by organizations that skip the Q1 foundation and proceed directly to deployment.

Quarter 2: Workflow Redesign + Pilot Deployment

Focus Capabilities: Capability 5 (Adoption) + Capability 3 (Operating Model) — workflow redesign runs parallel, not after

Rewired anchor: Ch. 5 (agentic workflow reimagination), Ch. 12 (pod operating model), Ch. 30 (make adoption stick)

Milestone 2.1 — Workflow Redesign (Weeks 11–16)

The domain owner leads a structured workflow redesign for the priority workflow. Required outputs: (a) the current-state workflow map at the task level, with step-by-step documentation; (b) the future-state map specifying what AI handles, what humans handle for exceptions, what is eliminated; © the decision architecture — in writing, how a reviewer verifies correct AI output in under 5 minutes; (d) escalation criteria — system-enforced rules for when the workflow routes to a human.

Why workflow redesign precedes deployment: McKinsey (n=1,993, Nov 2025): 55% of high performers fundamentally redesigned workflows vs. 18% of others — the single most predictive variable across 25 tested organizational attributes. MIT CISR (n=721): Stage 1 (tools without redesign) = −12.6 pp vs. industry average growth; Stage 3 (with redesign) = +11.3 pp. The redesign is not a Phase 3 activity. It is a Phase 2 prerequisite.

Resource requirement: Domain owner (0.5 FTE, 6 weeks — this is not delegatable to IT) + business-process analyst or workflow designer (1.0 FTE) + IT architect for system integration mapping (0.25 FTE) + 2–3 frontline employees from the workflow for documentation sessions.

Risk: Business leaders resist documentation because it surfaces how little the process is actually defined. Mitigation: this is expected — treat undocumented process as the signal, not a failure. Stanford Enterprise AI Playbook 2026 (n=51): 27% of prior failures attributed to “critical knowledge never captured.” Undocumented process is the finding; the document is the fix.

Measure: Workflow redesign document is signed by the domain owner. It includes: what AI handles, what humans handle, what is eliminated, decision criteria in writing, escalation rules. If the “what is eliminated” column is blank, the redesign is incomplete. See Red Flag #1 in failure checklist.

Milestone 2.2 — Pilot Deployment on Production Data (Weeks 13–22)

Deploy the AI solution against production data — not curated sample data. Rewired Ch. 5 (p.92): evaluation discipline requires task success rate, F1/precision/recall, retrieval accuracy, and hallucination rate. These metrics require production data to measure.

Pre-conditions required before this milestone starts:

  • Workflow redesign complete (Milestone 2.1)
  • Data architecture Path A or B confirmed (Milestone 1.2)
  • Human-in-the-loop review design built (not a click-to-approve)
  • Kill criteria defined: adoption >25% at 90 days, unit cost declining at 6 months, documentable P&L impact at 12 months

Resource requirement: ML engineer or SI equivalent (1.0 FTE) + data engineer (0.5 FTE) + domain workflow owner (0.25 FTE ongoing) + 2–4 employees from the workflow for training and feedback.

Risk: Pilot deployed on curated data despite Milestone 1.2 requirement. Pertama Partners (n=2,400+, 2025–2026): data quality issues appear in 71% of AI project failures; 38% of formally abandoned projects cite insurmountable data quality as primary reason. If the pilot must run on curated data, escalate immediately — the production timeline and budget will need revision.

Measure: At 30 days: adoption rate (target >25%), task success rate vs. baseline, error rate tracked and within pre-defined threshold. At 60 days: unit cost per transaction trending down, human review time is measured (not rubber-stamped — see Red Flag #16 in checklist).

Milestone 2.3 — Pod Operating Model Activated (Weeks 13–24)

The deployment moves from an IT-led project team to a stable, business-aligned pod that owns the workflow outcome end-to-end. Rewired Ch. 12 (p.199): pod-based operating model with embedded engineering. This is not a rename — it requires the domain owner to have accountability for outcomes, the pod to have authority to change the workflow without a committee, and engineering to be embedded, not handed off.

Resource requirement: Organizational design decision by CIO + domain business leader (1 structured session). No net new headcount required for this milestone — it is a governance and authority change.

Risk: IT retains de facto control of the project while the domain owner has nominal accountability. Mitigation: test the accountability by asking the domain owner to name the change they would make to the workflow today without asking IT. If they cannot, the pod structure is not real.

Measure: Pod charter is documented: named pod members, named domain owner, decision rights, escalation path to a named executive. Domain owner has made at least one workflow-scope decision without routing through IT.

Q2 Checkpoint (end of Month 6):

  • [ ] Workflow redesign document complete and signed
  • [ ] Pilot running on production data
  • [ ] 30-day adoption rate ≥25%
  • [ ] Unit cost per transaction is being tracked (not hours saved)
  • [ ] Pod operating model chartered and active

Quarter 3: Scale Architecture + Governance Hardening

Focus Capabilities: Capability 4 (Data Products for Reuse) + Capability 6 (Governance)

Rewired anchor: Ch. 26 (data products as reusable building blocks), Ch. 31 (design for scale and reuse from day one), Ch. 34 (right to deploy)

Milestone 3.1 — Data Product Build (Weeks 25–32)

Convert the data work from Q2’s pilot into a reusable data product: a governed, documented, versioned data asset that the next 2–3 AI workflows in this domain can consume without rebuilding the integration. Rewired Ch. 26 (p.401): each high-value data product (e.g., customer 360, supplier-risk profile) is built once and reused. Palantir’s 139% net dollar retention (Q4 2025 audited): after foundational data work, each new use case costs a fraction of the first — the economic mechanism behind the “best use case is the reuse case” principle (Rewired Ch. 31, p.469).

Resource requirement: Data engineer (1.0 FTE, 8 weeks) + domain data owner for business-rule documentation (0.25 FTE) + data governance lead to apply ownership, quality standards, and access controls.

Risk: The data product is built as a point solution for this workflow only — no reuse architecture. This is Red Flag #18 in the failure checklist. Mitigation: require the data engineer to document two additional workflows in the domain that could consume this data product before the build is approved.

Measure: Data product is documented in a data catalog with: owner, source systems, update cadence, quality SLA, access path, and a list of the next 2–3 workflows that could consume it.

Milestone 3.2 — Governance Framework Activation (Weeks 25–34)

Implement the right-to-deploy framework (Rewired Ch. 34, p.511): legal, risk, compliance, and security sign off as a parallel stream to engineering, not a post-deployment review. Rewired’s explicit position: governance is a “right to deploy” gate, not a review meeting. Name the governance owner — McKinsey RAI benchmark (n=~500, 2026): named owner = 2.6/4.0 maturity score vs. 1.8 without.

Resource requirement: Legal + risk + compliance representatives (0.25 FTE each for 4 weeks) + CIO and domain owner to define scope of deployment subject to governance review.

Risk: Governance framework becomes the FinCo failure pattern — comprehensive design, no review cycle, governance stalls deployment. MIT CISR FinCo case: governance without a living review cycle matched to technology pace produces more shadow AI, not less. Mitigation: governance framework must include an explicit review cycle (quarterly at minimum) and a fast path for low-risk deployments.

Measure: Named governance owner documented. Right-to-deploy framework has been applied to the Q2 pilot retroactively (as a dry run) and the output signed by legal, risk, and compliance. Review cycle is on the calendar.

Milestone 3.3 — Training Program for Scale (Weeks 27–34)

Build and deploy the training program for employees who will use the AI tool — covering how to use the tool, when not to use it, quality standards, escalation judgment, and known failure modes. BCG AI at Work 2025 (n=10,600): employees with 5+ hours of AI training become regular users at 79% vs. 67% with less. Workday/Hanover Research (n=3,200, Jan 2026): 40% of AI time savings consumed by rework among under-trained users.

Resource requirement: Learning and development lead (0.5 FTE, 6 weeks) + domain workflow owner for content validation + IT for tool-mechanics content.

Risk: Training covers tool mechanics only, not when not to use the tool. The failure mode is producing “workslop” — AI-generated output that appears polished but lacks substance. BetterUp/Stanford (n=1,150, Sep 2025): 40% of desk workers received workslop from a colleague monthly; $9M annual cleanup cost for a 10,000-person organization.

Measure: Training completion tracked (target ≥80% of users before full rollout). Post-training assessment covers at least one “when not to use” scenario. Review time per AI output is being measured and is not declining toward zero.

Q3 Checkpoint (end of Month 9):

  • [ ] Data product built and in catalog with reuse documentation
  • [ ] Governance framework active with named owner and review cycle
  • [ ] Training completion ≥80% of users
  • [ ] Unit cost per transaction declining vs. Q2 baseline
  • [ ] 6-month kill criteria met: unit cost declining, adoption sustained above 25%

Quarter 4: Production Validation + Second Workflow

Focus Capabilities: Capability 5 (Scale and Reuse) + Capability 1 (Domain Expansion Planning)

Rewired anchor: Ch. 32 (track what matters), Ch. 33 (plan for midstream adjustments), Ch. 31 (the best use case is the reuse case)

Milestone 4.1 — 12-Month P&L Validation (Weeks 37–44)

Document the P&L impact of the Year 1 workflow deployment against the economic case written in Q1. Required: baseline cost per cycle from Q1, current cost per cycle, volume, financial impact calculation, and a qualified sign-off from the CFO or P&L sponsor. Rewired Ch. 32 (p.483): KPI discipline tied to P&L line items, not activity counts.

This is the business case for Year 2. Without a validated 12-month P&L number, the second domain and second workflow will require a separate budget battle. With a validated number, the next deployment self-funds from the first.

Resource requirement: Finance analyst (0.5 FTE, 4 weeks) + domain owner for operational data + CFO or P&L sponsor for sign-off.

Risk: The P&L impact is real but the measurement methodology is disputed. Mitigation: the baseline measurement methodology was established in Q1 (Milestone 1.1). If it was not, the CFO will reject the calculation. This is not a Q4 problem — it is a Q1 omission that must be corrected here.

Measure: P&L validation document is signed by CFO or P&L sponsor. It includes: baseline, current state, delta, methodology, and a recommendation on whether to expand the deployment (scale), adjust scope (adjust), or stop (stop). Rewired Ch. 33 (p.501): midstream adjustment is a planned discipline, not a sign of failure.

Milestone 4.2 — Second Workflow Deployment (Weeks 37–48)

Using the data product and infrastructure built in Q3, deploy a second AI workflow in the same domain. Rewired Ch. 31 (p.469): “the best use case is the reuse case.” This deployment should be materially faster and cheaper than the first because the data foundation is in place.

Target economics: If the first deployment took 20 weeks and $400K, the second should take 10–12 weeks and $180–250K by reusing the data product, governance framework, training materials, and pod operating model.

Pre-condition: Second workflow has been identified and scored against the workflow readiness assessment during Q3. Do not start a second deployment that scores below 16/40 on the readiness assessment.

Resource requirement: Same pod as Q2 (deployed on second workflow). Incremental resource requirement should be 30–50% lower than Q2 because the infrastructure exists.

Measure: Second workflow in production by end of Month 12. Time-to-production and cost-per-workflow documented for the Year 2 roadmap business case.

Milestone 4.3 — Year 2 Roadmap Draft (Weeks 41–48)

The domain owner produces a Year 2 roadmap brief: (a) the 12-month P&L validation from Milestone 4.1; (b) two additional workflows in this domain recommended for Year 2; © whether a second domain should be activated in parallel; (d) the data product expansion required; (e) the governance gaps that surfaced in Year 1.

Resource requirement: Domain owner (0.25 FTE, 4 weeks) + CIO for infrastructure review + strategy or transformation lead.

Measure: Year 2 roadmap brief is reviewed and approved by the same stakeholder group that approved the Q1 Domain Brief. The second domain, if recommended, has a named business owner before Year 2 starts.

Q4 Checkpoint (end of Month 12):

  • [ ] 12-month P&L impact documented and signed by CFO or sponsor
  • [ ] Second workflow in production
  • [ ] Second workflow’s cost and timeline are ≤60% of first workflow’s (reuse dividend quantified)
  • [ ] Year 2 roadmap brief complete and approved
  • [ ] Named owner for second domain (if Year 2 expands)

Resource Summary by Quarter

Quarter Domain Owner FTE Engineering FTE Data FTE Other
Q1 (Foundation) 0.5 0 0.5 (assessment) Talent lead 0.25, Strategy 0.25
Q2 (Deploy) 0.5 1.5 (ML+SI) 0.5 Process analyst 1.0, Training 0.25
Q3 (Scale) 0.25 0.5 1.0 (data product) Legal/Risk/Compliance 0.25 each
Q4 (Validate) 0.5 1.0 (second workflow) 0.5 Finance analyst 0.5, Strategy 0.25

Mid-market note: Engineering FTE for Q2–Q4 is typically filled by an SI (systems integrator) + platform licensing, not internal hires. The economics for mid-market are: buy the platform, invest in workflow redesign and data products on top, reserve build investment for the one or two workflows that are competitively differentiating. See synthesis Tension 3 for the PwC and Grant Thornton evidence on build vs. buy economics.

Common Sequencing Errors

Starting Q2 before Q1 is complete. The most common failure mode. Organizations launch pilots without a signed domain brief, classified data architecture, or committed production path. The pilot succeeds technically and fails organizationally: no production budget was committed, no workflow was redesigned, no business owner has authority to act on the results.

Treating governance (Q3) as a Q1 activity. Building a comprehensive governance framework before the first deployment is the FinCo failure pattern. Governance designed before the first workflow is deployed is designed for a deployment that does not exist yet. Build governance in Q3, when you have a real deployment to govern.

Skipping the data product build (Q3 Milestone 3.1). Organizations that deploy AI as point solutions — one workflow, one integration, one bespoke data pipeline — pay full infrastructure cost on every subsequent deployment. The reuse dividend (Milestone 4.2’s 30–50% cost reduction) only exists if the data product was built in Q3.

P&L validation without a baseline. If Q1 did not include a 30–60 day baseline measurement of the current workflow’s cost per cycle, the Q4 P&L validation has no denominator. The CFO will not sign it. Establish the baseline before the pilot starts.

Brandon Sneider | brandon@brandonsneider.com April 2026