Implementation Roadmap

Where to start, how fast, and what to watch for.

The Sequencing Problem

70–85% of GenAI deployments fail to scale beyond pilots (NTT DATA, 2024). The technology works. The organizational execution doesn't.

The primary reason: organizations digitize existing processes without first redesigning them (HBR, 2025). They layer AI onto old workflows and wonder why nothing changes structurally.

This page provides sequencing guidance for leaders who have read The Business Case, assessed their organization, and are ready to move. It is not a generic playbook. It is a sequence of decisions, each informed by the one before it.

Acceptance Criteria

These criteria define what "done" looks like at each level. Drawn from the Reference Framework.

Level 2 — Achieved when ALL these criteria are met:

AI usage is a documented expectation for every role, not optional
Every department maintains a structured context file loaded before AI tasks
Shared prompt libraries or workflow templates exist and are in use
At least 1 workflow per department has been redesigned around AI (before/after documented)
KPIs include AI output metrics (not just activity)
"How did AI help?" is asked in reviews and retrospectives
If AI disappeared tomorrow, at least some workflows would break

Level 3 — Achieved when ALL these criteria are met:

Roles are defined by judgment and direction, not execution
Agents, pipelines, or decision systems are in production (not prototypes)
Non-trivial tasks have written specifications conforming to the execution standards
Every AI system in production has an assigned Spec Owner, Context Owner, and Evaluation Owner
AI impact is measured by department (time saved, costs reduced, quality improved)
Hiring profiles require Tier 2+ minimum
If AI disappeared tomorrow, the department couldn't function

The Transformation Path

Level 1 → Level 2

Prerequisites:

Leadership commits to AI as an operational standard, not optional
Investment in shared AI infrastructure (tools, templates, training)
Processes audited and redesigned for AI integration
KPIs updated to measure AI output
"How did AI help?" becomes a standard question

Timeline: 3-6 months with committed leadership

Level 2 → Level 3

Level 2 is the operational floor: every department must reach it. Level 3 is the organizational target. Non-engineering departments aim for Level 2 as their first milestone; engineering aims directly for Level 3 via the AI Lab.

Prerequisites:

Leadership is willing to eliminate roles, not just tasks (see the Role Decision Matrix)
Hiring profiles change to require Tier 2+ minimum
Product/service is redesigned assuming AI execution
Organizational structure flattens significantly

Timeline: 6-12 months

For engineering, the AI Lab lifecycle defines the specific phase sequence from Rung 3 to Rung 5.

Which Team First

Not every team is equally ready or equally valuable as a starting point. The research points to two criteria for selecting your first-mover:

Criterion 1: Highest value density

Customer service generates 38% of AI's total business value — more than any other function. Operations accounts for 23%, marketing and sales for 20%, and R&D for 13% (BCG, 2025).

Customer service is the default recommendation for most organizations because:

It has the highest actual AI task coverage (Anthropic, 2026)
The role evolution path is the clearest (see Skill Progression Map — Customer Service)
Results are measurable fast: deflection rate, resolution time, customer satisfaction
The work decomposes naturally into what AI handles and what requires human judgment

Criterion 2: Highest readiness

Value density matters, but readiness matters more for the first team. A team with lower value potential but high readiness will produce a faster, cleaner success that builds credibility for the next team.

Readiness signals (from Assessing Your Organization):

Team is already at Level 1 with visible adoption
Manager is personally at Level 2 minimum
Team culture is open to process change
Clear, measurable workflows exist (not ad hoc work)
Leadership support is explicit, not just implied

If your highest-value team isn't your most ready team, start with the ready one. A clean success is worth more than a messy one in a high-value area.

The sequencing matrix

Scenario	Start with	Why
CS is ready and high-value	Customer service	Default recommendation — highest value, clearest path
CS isn't ready, but marketing is	Marketing	Fastest path to visible results; content workflows decompose well
Engineering is already at Level 2	Engineering	Leverage existing momentum; engineering success de-risks the approach for other teams
No team is ready	One manager's team, any function	Pick the manager most likely to succeed; the goal is a proof point, not max value

Infrastructure Before Transformation

You cannot redesign workflows on top of broken infrastructure. Before launching your first-mover team, verify these prerequisites:

1. Tool access

Every team member has access to AI tools appropriate for their work. Not "can request access" — has access, configured, and working. The most common silent killer of transformation is friction: people revert to old workflows when the AI tool takes two extra clicks.

2. Context systems

Each team maintains a structured context file containing goals, constraints, terminology, quality standards, and relevant documents. See AI Execution Standards — Layer 2. AI tasks load this context before execution.

Without context systems, every AI interaction starts from zero. This is the difference between Level 1 (ad hoc prompting) and Level 2 (systematic integration).

3. Quality standards

Define what "good enough" means for AI output before people start producing it. Without standards, you get either over-trust (publishing AI output without review) or over-caution (editing everything back to manual quality, negating the speed gain).

The AI Execution Standards provide the framework. The minimum: every AI-enabled workflow defines its acceptance criteria before launch.

4. Governance baseline

Who is authorized to deploy AI in customer-facing contexts? What data can and cannot be used? What happens when AI output is wrong? These questions need answers before the first pilot, not after the first incident.

The 30/60/90 for Leaders

This section is original to this framework. The published literature covers tool adoption timelines, not structural transformation timelines. These phases are synthesized from the operational experience behind this framework and the research in The Business Case.

Days 1–30: Foundation

Your job: Prepare the ground. No workflow changes yet.

Complete your organizational assessment — maturity map by team
Select your first-mover team using the sequencing criteria above
Verify infrastructure prerequisites (tool access, context systems, quality standards, governance)
Brief the first-mover manager using the Leading the Transformation framework — Layer 1 (Personal Competence) and Layer 2 (Team Context Mapping)
Define 2–3 specific workflows to redesign (not "use AI more" — specific, named workflows with current and target states)
Set baseline metrics for those workflows: current time, cost, quality, throughput

What success looks like at Day 30: You have a map, a team, a manager who's done the context mapping, and 2–3 workflows selected with baselines measured. No one has changed how they work yet.

Days 31–60: Pilot

Your job: Redesign and execute the selected workflows with the first-mover team.

Manager completes Layers 3–4 of Leading the Transformation (Intent Definition + Transition Specification) for each selected workflow
Team members begin their own Transforming Your Role process (Layers 1–2: AI Literacy + Work Mapping)
Deploy the redesigned workflows — AI handles execution, humans handle judgment
Measure weekly: time, cost, quality vs. baseline
Run weekly team sessions: what worked, what broke, what needs adjustment
Document what you learn — this becomes the playbook for the next team

What success looks like at Day 60: 2–3 workflows are running in the new mode. Measurable improvements exist (even if modest). The team can articulate what changed and why. You have a documented playbook.

Days 61–90: Validate and Plan Scale

Your job: Confirm the model works, then plan the next teams.

Review results against baselines — are the gains real and sustainable?
Identify what worked and what didn't (be honest; partial success is still data)
Determine which role evolution patterns are emerging in the first-mover team
Select the next 2–3 teams for transformation based on readiness and value
Begin Layer 1–2 preparation with those teams' managers
Present results to leadership: what changed, what it cost, what's next

What success looks like at Day 90: Validated results from one team. An honest assessment of what worked. A plan to scale to 2–3 more teams. Leadership buy-in based on evidence, not promises.

Decision Checkpoints

Transformation is not a project you launch and forget. It requires structured moments to evaluate whether to accelerate, adjust, or pause.

Checkpoint 1: After Assessment (Day 30)

Question: Is the infrastructure ready and is the first-mover team viable?

If yes → proceed to pilot
If no → fix prerequisites before starting. Launching a pilot on broken infrastructure wastes the team's goodwill
If uncertain → run a 2-week mini-pilot with one workflow to test readiness

Checkpoint 2: After Pilot (Day 60)

Question: Are the redesigned workflows producing measurable improvement?

If yes → validate and plan scale
If improvement is modest but real → continue; early results compound. The J-curve of adoption means a dip before the rise
If no measurable improvement → diagnose. Was the workflow a good candidate? Was the specification clear enough? Did the team have the right tools and training?
If the team resists → this is a management problem, not a technology problem. See the "To your team" section in The Business Case

Checkpoint 3: Before Scaling (Day 90)

Question: Should we scale to more teams?

If pilot succeeded and next teams are ready → scale
If pilot succeeded but next teams aren't ready → invest in readiness (tools, training, manager preparation) before launching
If pilot produced mixed results → run a second pilot with a different team before committing to scale. One data point is not a pattern

Ongoing: Quarterly Review

Once scaling begins, review quarterly:

Which teams have progressed? Which are stuck?
Are the role evolution patterns emerging as expected?
Has the organization's overall maturity level shifted?
What new infrastructure or governance needs have emerged?
Is the pace sustainable, or are teams burning out?

77% of enterprises have scaled fewer than 40% of their GenAI pilots (Concentrix/Everest, 2025). The most common reason: experimentation happens faster than governance. Quarterly checkpoints prevent this drift.

Scaling from Pilot to Organization

The transition from one successful team to organization-wide transformation is where most efforts fail. MIT CISR identifies four challenges at this stage (MIT CISR, 2025):

1. Strategy

The transformation must be connected to business outcomes, not positioned as a technology initiative. The Business Case provides the framing. Each new team that joins should understand why they're transforming, not just how.

2. Systems

The infrastructure that worked for one team may not scale. Context systems, quality standards, and governance need to evolve as more teams onboard. What started as one team's context file becomes an organizational knowledge system.

3. Synchronization

People and roles need to evolve together with the systems. This is where the role evolution patterns become critical at scale. Different teams will experience different patterns — engineering may undergo Elevation while customer service undergoes Convergence. The organization needs vocabulary and process to handle this diversity.

4. Stewardship

Someone must own the transformation at the organizational level — not as a project manager, but as a system designer. This is an emerging role. It requires specification engineering skills, comfort with ambiguity, and organizational authority.

The scaling sequence

Phase	Teams	Focus
Pilot	1 team	Prove the model works. Document everything.
Expansion	2–3 more teams	Prove the model transfers. Refine the playbook.
Integration	All willing teams	Build organizational infrastructure. Establish governance.
Native operations	Organization-wide	The transformation becomes the operating model.

The timeline from pilot to native operations is typically 12–24 months (Promethium, 2025). Organizations with mature data infrastructure move faster. Those requiring foundational work should add 6–9 months of preparation.

Rushing this timeline is the single most common scaling mistake. A team that reaches Level 2 in a quarter can sustain it. A team that's pushed to Level 2 in a month reverts the moment attention shifts.

What This Does Not Cover

This roadmap addresses the organizational transformation — redesigning how work gets done so that humans specify and systems execute. It does not cover:

AI product strategy — building AI into your products for customers
AI infrastructure engineering — the technical stack decisions for AI deployment
Regulatory compliance — AI governance and compliance are prerequisites (see the governance baseline above) but are organization-specific

These are important topics. They are not this framework's scope.

← Back to home · The business case · Assessing your organization · Leading the transformation