ChatGPT
Google AI
Claude
Perplexity
Grok
Executive Summary
An AI roadmap translates strategy into a time-bound plan with specific milestones, dependencies, and resource allocations. The 18-month horizon balances long enough for meaningful progress with short enough to maintain accountability.
Effective roadmaps include three phases: Foundation (months 1-6), Build (months 7-12), and Scale (months 13-18). Each phase has defined objectives, deliverables, success criteria, and decision gates. Roadmaps must be living documents, reviewed quarterly and adjusted based on learning.
The best roadmaps balance ambition with realism. They are achievable but stretch capability. Roadmap creation is a team exercise requiring input from business, technology, and operations leadership.
Why This Matters Now
Strategy provides direction. A roadmap provides a path.
Many organizations have AI strategies that never translate to action. The gap isn't commitment. It's specificity. "We will become an AI-enabled organization" is a vision. A roadmap specifies what happens in Q1, what capabilities must be in place by month 6, and what success looks like at month 18.
Without a roadmap, teams pursue initiatives without coordination, resources aren't allocated to the right priorities, progress isn't measurable, and leadership can't track execution against commitments. The result is scattered effort that produces activity without outcomes.
The 18-month horizon is deliberate. Shorter horizons don't allow for meaningful AI capability building. Longer horizons become speculative. Eighteen months provides enough time to demonstrate value while maintaining accountability.
AI Roadmap vs. AI Strategy
| Element | AI Strategy | AI Roadmap |
|---|---|---|
| Focus | Direction and priorities | Execution plan |
| Time horizon | 2-3 years | 12-18 months |
| Detail level | What and why | How and when |
| Update frequency | Annually | Quarterly |
| Primary audience | Leadership and board | Execution teams |
Strategy answers "What should we do?" Roadmap answers "When and how will we do it?"
If you haven't developed your [AI strategy], start there. Roadmapping without strategy produces activity without direction.
The 18-Month Roadmap Structure
Phase 1: Foundation (Months 1-6)
Objective: Establish the capabilities necessary for sustainable AI deployment.
Foundation isn't glamorous, but it determines everything that follows. Organizations that rush past foundation work face repeated remediation later.
Key Workstreams:
| Workstream | Deliverables | Success Criteria |
|---|---|---|
| Data Foundation | Data inventory, quality assessment, governance framework | Key data sources cataloged, quality baseline established |
| Infrastructure | Cloud environment, security controls, integration architecture | Environment provisioned, security approved |
| Governance | [AI policy], risk framework, decision-making structure | Policy approved, committee operational |
| Skills | Training program launch, key hires initiated | 50% of target audience trained on AI fundamentals |
| Pilot Selection | Use cases prioritized, pilot plans developed | 2-3 pilots approved and resourced |
Foundation Phase Milestones:
The foundation phase follows a progressive timeline. By month 2, the AI policy draft should be completed. Month 3 targets data governance framework approval. The first pilot begins in month 4. Month 6 concludes with a foundation phase review and decision gate.
Decision Gate Questions (Month 6):
At the month 6 gate, leadership must assess whether the data foundation is sufficient for planned pilots, whether the governance structure can manage AI risk effectively, whether pilots are on track and what has been learned, and whether the organization should proceed to the Build phase as planned or extend the Foundation phase.
Phase 2: Build (Months 7-12)
Objective: Demonstrate AI value through successful pilots and initial production deployments.
Build phase is where theory becomes practice. The focus is on learning through doing and understanding what works in your specific context.
Key Workstreams:
| Workstream | Deliverables | Success Criteria |
|---|---|---|
| Pilot Execution | Pilots completed, results measured, learnings documented | 2+ pilots deliver measurable value |
| Production Deployment | First AI system in production use | System live, users trained, monitoring active |
| Process Development | AI development lifecycle documented, repeatable process | Second pilot follows documented process |
| Capability Building | Additional training, specialized hires, vendor partnerships | Technical capability sufficient for planned scale |
| Governance Maturation | [risk register] populated, board reporting established | [governance committee] reviews production systems |
Build Phase Milestones:
First pilot results should be available by month 8. The first production deployment decision comes at month 9. A second pilot begins in month 10. Month 12 closes with the build phase review and decision gate.
Decision Gate Questions (Month 12):
The month 12 review addresses whether pilots have demonstrated expected value, whether the organization is ready to scale successful approaches, what capability gaps remain, and what adjustments are needed for the Scale phase.
Phase 3: Scale (Months 13-18)
Objective: Expand proven AI approaches across the organization and establish a sustainable operating model.
Scale phase multiplies success. It's not about starting new experiments. It's about extending what works.
Key Workstreams:
| Workstream | Deliverables | Success Criteria |
|---|---|---|
| Deployment Expansion | Successful pilots scaled to broader use | 3+ AI systems in production |
| Operating Model | AI Center of Excellence or embedded model operational | Clear ownership, processes, and accountability |
| ROI Realization | Business value measured and reported | Documented ROI meets or exceeds projections |
| Next Horizon Planning | Year 2 roadmap developed | Strategy refresh and next 18-month plan approved |
| Continuous Improvement | Optimization of deployed systems, lessons learned | Performance metrics improving quarter over quarter |
Scale Phase Milestones:
The second production deployment targets month 14. Year 1 ROI assessment takes place at month 15. The Year 2 strategy and roadmap draft is prepared by month 16. Month 18 concludes with the roadmap completion review.
Decision Gate Questions (Month 18):
The final gate evaluates what business value has been created, what capabilities have been built, what works and what doesn't, and what the next 18 months should focus on.
SOP: Quarterly Roadmap Review
Roadmaps drift without disciplined review. This Standard Operating Procedure ensures roadmaps remain relevant and accountable.
Purpose
Quarterly review ensures roadmap alignment with strategy, tracks progress against milestones, and enables course corrections.
Frequency
Reviews occur quarterly, aligned with the business planning cycle.
Participants
The review requires the AI/Digital Leader as Chair, the Executive Sponsor, IT Leadership, Business Unit Representatives, a Risk/Compliance Representative, and a Finance Representative.
Pre-Meeting Preparation (Owner: AI/Digital Leader)
One week before the meeting, the AI/Digital Leader collects status updates from all workstream leads, compiles milestone progress across completed, in-progress, at-risk, and not-started categories, prepares variance analysis comparing actual against planned outcomes, documents emerging risks and issues, gathers budget status comparing actual against planned spend, and prepares draft roadmap adjustments for discussion.
Meeting Agenda (90 minutes)
| Time | Topic | Owner |
|---|---|---|
| 0:00 | Strategic context update | Executive Sponsor |
| 0:10 | Milestone review | AI/Digital Leader |
| 0:30 | Variance analysis and root causes | AI/Digital Leader |
| 0:45 | Risk and issue discussion | All |
| 1:00 | Budget review | Finance |
| 1:10 | Proposed roadmap adjustments | AI/Digital Leader |
| 1:20 | Decision and action items | Chair |
Decision Types
| Decision | Authority | Criteria |
|---|---|---|
| Minor timeline adjustment (<4 weeks) | AI/Digital Leader | Documented rationale |
| Milestone change or addition | Steering Committee | Majority approval |
| Scope change (add/remove workstream) | Executive Sponsor | Business case required |
| Budget reallocation (>10%) | Executive Sponsor + Finance | Approval required |
| Phase gate decision | Steering Committee | Defined criteria met |
Post-Meeting Actions (Owner: AI/Digital Leader)
Within one week, the AI/Digital Leader distributes meeting notes and decisions, updates the roadmap document with approved changes, communicates adjustments to affected teams, updates the risk register, and schedules follow-up on action items.
Documentation
Quarterly review records should include meeting attendance, milestone status at time of review, decisions made, roadmap changes approved, and action items with owners.
Roadmap Template
Phase 1: Foundation (Months 1-6)
| Month | Data | Infrastructure | Governance | Skills | Pilots |
|---|---|---|---|---|---|
| 1 | Data inventory begins | Cloud assessment | Policy drafting | Training needs assessment | Use case shortlisting |
| 2 | Quality assessment | Environment design | Policy review | Training program design | Pilot selection |
| 3 | Governance framework | Environment build | Policy approval | Training pilot | Pilot planning |
| 4 | Data quality remediation | Integration architecture | Committee formation | Training rollout | Pilot 1 kickoff |
| 5 | Continued remediation | Security implementation | First committee meeting | Continued training | Pilot 1 execution |
| 6 | Foundation review | Environment operational | Phase gate review | Training milestone | Pilot 1 results |
Phase 2: Build (Months 7-12)
| Month | Pilots | Production | Process | Capability | Governance |
|---|---|---|---|---|---|
| 7 | Pilot 1 wrap-up | Production planning | Process documentation | Advanced training | Risk register setup |
| 8 | Pilot 2 kickoff | Deployment prep | Process review | Specialized hiring | Board report drafted |
| 9 | Pilot 2 execution | First deployment | Process refinement | Vendor evaluation | First board report |
| 10 | Pilot 3 kickoff | Monitoring setup | Process training | Partnership exploration | Policy review |
| 11 | Pilot 3 execution | Optimization | Process adoption | Capability assessment | Compliance review |
| 12 | Build phase review | Production stable | Process documented | Capability gaps identified | Phase gate review |
Phase 3: Scale (Months 13-18)
| Month | Expansion | Operating Model | ROI | Planning | Improvement |
|---|---|---|---|---|---|
| 13 | Second deployment planning | Model design | ROI tracking setup | Retrospective | Optimization begins |
| 14 | Second deployment | Model piloting | Q1 ROI analysis | Lessons learned | Performance review |
| 15 | Third deployment planning | Model refinement | Year 1 ROI | Strategy refresh | Metrics analysis |
| 16 | Third deployment | Model operational | ROI reporting | Year 2 roadmap draft | Continuous improvement |
| 17 | Expansion assessment | Model optimization | Value communication | Roadmap review | System optimization |
| 18 | Roadmap completion | Model mature | Final Year 1 report | Year 2 approval | Improvement roadmap |
Common Failure Modes
1. Overloading the Roadmap
Attempting too many initiatives simultaneously dilutes focus and exhausts resources. The fix is to limit active workstreams. Two major initiatives executing well beats five struggling.
2. Ignoring Dependencies
AI initiatives have dependencies on data, infrastructure, and skills. Ignoring them creates bottlenecks. The fix is to map dependencies explicitly and never schedule work that depends on incomplete prerequisites.
3. Fixed Thinking
Treating the roadmap as immutable when circumstances change leads to wasted effort. The fix is quarterly reviews with genuine authority to adjust. The roadmap serves the strategy, not the other way around.
4. Milestones Without Meaning
Milestones that don't represent genuine progress, like "complete documentation" rather than "pilot delivers 20% efficiency improvement," create a false sense of momentum. The fix is to define milestones in terms of outcomes, not activities.
5. Disconnection from Budget
Roadmap plans without corresponding budget allocation are wishes, not plans. The fix is ensuring every roadmap element has allocated resources. Unfunded initiatives should be removed or deferred.
Checklist: AI Roadmap Development
Preparation
Confirm that the AI strategy is approved and available. Identify and secure the executive sponsor. Assemble the roadmap development team. Complete the current state assessment. Understand resource constraints before beginning.
Phase Design
Define three phases with clear objectives. Identify workstreams for each phase. Define milestones as outcomes rather than activities. Map dependencies between workstreams. Establish decision gates with specific criteria.
Resource Alignment
Allocate budget to each phase. Assign or plan for people resources. Specify technology requirements. Identify external support needs.
Governance
Establish the quarterly review process. Define decision authority. Clarify the escalation path. Set the reporting cadence.
Metrics to Track
| Metric | What It Measures | Frequency |
|---|---|---|
| Milestone completion rate | % of milestones met on time | Monthly |
| Budget variance | Actual vs. planned spend | Monthly |
| Scope changes | Number and impact of changes | Quarterly |
| Decision gate passage | Whether criteria met at gates | Per gate |
| Business value delivered | ROI of deployed initiatives | Quarterly |
Next Steps
A roadmap transforms AI ambition into executable plans. It creates accountability, enables tracking, and provides a structure for learning and adjustment.
If you have a strategy but lack a concrete implementation plan, roadmap development is your next step.
Book an AI Readiness Audit with Pertama Partners to develop a roadmap grounded in your specific context, capabilities, and constraints.
Related Reading
- [Building Your First AI Strategy]
- [7 AI Strategy Mistakes That Derail Implementation]
- [AI Investment Prioritization: Allocating Budget for Maximum Impact]
Common Questions
The first 90 days should focus on foundation-setting rather than technology deployment. This includes completing an AI readiness assessment, identifying and prioritizing 3 to 5 potential use cases based on business impact and feasibility, auditing data assets required for priority use cases, selecting an initial pilot project with a clear success metric, and securing executive sponsorship with defined governance roles. Skipping this foundation phase is the primary reason AI roadmaps derail in later stages.
At the 18-month mark, success should be measured across four dimensions: business value delivered (quantified ROI or cost savings from deployed AI solutions), operational maturity (number of AI models in production with proper monitoring and maintenance), organizational capability (percentage of target employees trained on AI tools and processes), and strategic positioning (documented pipeline of future AI initiatives with clear business cases and resource plans). Avoid measuring success solely by number of projects completed.
References
- AI Risk Management Framework (AI RMF 1.0). National Institute of Standards and Technology (NIST) (2023). View source
- ISO/IEC 42001:2023 — Artificial Intelligence Management System. International Organization for Standardization (2023). View source
- Model AI Governance Framework (Second Edition). PDPC and IMDA Singapore (2020). View source
- Enterprise Development Grant (EDG) — Enterprise Singapore. Enterprise Singapore (2024). View source
- What is AI Verify — AI Verify Foundation. AI Verify Foundation (2023). View source
- OECD Principles on Artificial Intelligence. OECD (2019). View source
- ASEAN Guide on AI Governance and Ethics. ASEAN Secretariat (2024). View source
