How to Create an AI Roadmap: From Strategy to 12–24 Month Execution
Many organisations now have an AI strategy.
The strategy explains why AI matters.
It identifies opportunities.
It establishes ambition.
It may describe priority areas, governance principles and expected business outcomes.
Then leadership asks the practical question:
What happens next?
This is where strategy can begin to break down.
An organisation may know it wants to:
increase productivity;
improve customer experience;
automate processes;
introduce AI agents;
strengthen decision-making;
create new products.
But those ambitions do not automatically tell leadership:
what should happen first;
which capabilities need building;
which use cases should begin now;
what dependencies exist;
who owns delivery;
how much investment is required;
what evidence should be produced;
when decisions should be revisited.
That is the role of an AI roadmap.
A good AI roadmap translates ambition into a sequence of decisions, capabilities, initiatives and measurable outcomes.
It connects:
where we want to go
with
what we need to do next.
What is an AI roadmap?
An AI roadmap is a structured plan showing how an organisation will move from its current AI position towards its desired future state over a defined period.
It should connect:
business strategy;
AI opportunities;
organisational readiness;
technology and data;
governance;
people;
investment;
execution;
value.
An AI roadmap is not simply a list of AI projects with dates beside them.
It should explain:
what needs to happen
why it needs to happen
what needs to happen first
who owns it
what evidence demonstrates progress
and
what decisions come next.
For more on ensuring AI strategy begins with business outcomes rather than technology, read AI Strategy for Business: Why Technology Isn't the Problem.
AI strategy and AI roadmap are not the same thing
The distinction is simple.
AI strategy asks:
Where should we use AI and why?
AI roadmap asks:
How do we move from where we are now to where we want to be?
The strategy provides direction.
The roadmap creates sequence.
Without strategy, a roadmap can become a list of disconnected technology projects.
Without a roadmap, strategy can remain a presentation.
The two need each other.
For more on keeping AI connected to commercial outcomes, read AI Strategy Without Business Value Is Just a Technology Plan.
Why 12–24 months?
AI moves too quickly for many organisations to create a detailed five-year implementation plan.
Technology will change.
Costs will change.
Vendors will change.
Capabilities will improve.
Regulation may change.
Business priorities will change.
But planning only three months ahead can create another problem:
the organisation never builds the capabilities required to scale.
A 12–24 month horizon is therefore useful.
It is long enough to establish direction and dependencies.
But short enough to adapt as evidence changes.
The roadmap should become progressively less detailed further into the future.
For example:
0–3 months
Specific actions and owners.
3–6 months
Defined initiatives and expected outcomes.
6–12 months
Capability and portfolio direction.
12–24 months
Strategic destination and major milestones.
This gives leadership direction without pretending it can predict AI perfectly.
The mistake: building the roadmap around technology
A weak roadmap might look like:
Q1 — deploy Copilot
Q2 — introduce chatbot
Q3 — pilot agents
Q4 — deploy predictive analytics
This is a technology deployment plan.
It does not explain:
which business problems matter;
what outcomes should improve;
what organisational capabilities are required;
whether employees will adopt the technology;
what governance is needed;
how value will be measured.
A stronger roadmap starts with:
business outcomes
then works backwards into:
capabilities
and finally:
technology.
The Stratify AI Roadmap Framework
I would build an AI roadmap across seven connected layers:
OUTCOMES → OPPORTUNITIES → FOUNDATIONS → PORTFOLIO → TRANSFORMATION → VALUE → DECISIONS
Each layer answers a different question.
1. OUTCOMES — What are we trying to change?
Begin with the organisation's strategic priorities.
Examples might include:
increase operating margin;
grow revenue;
improve customer retention;
reduce cost-to-serve;
increase employee capacity;
improve decision speed;
reduce operational risk;
create new digital services.
The roadmap should identify where AI can materially contribute to those outcomes.
This prevents the roadmap becoming:
What AI technology should we deploy?
Instead:
What business outcomes are important enough to justify AI investment?
2. OPPORTUNITIES — Where can AI create value?
Once outcomes are clear, identify opportunities.
These may include:
productivity;
automation;
decision support;
customer experience;
revenue growth;
risk reduction;
new products;
agentic workflows.
The objective is not to pursue every opportunity.
It is to create an opportunity portfolio.
For more on identifying high-value opportunities, read Where Should Your Business Use AI First? A Framework for Finding High-Value Opportunities.
3. FOUNDATIONS — What needs to be true?
This is where many AI roadmaps are too optimistic.
A use case may look attractive while the organisation lacks:
appropriate data;
integration;
technical infrastructure;
governance;
skills;
ownership;
change capacity.
The roadmap should therefore include the capabilities required to enable the portfolio.
These may include:
AI governance framework;
approved technology platforms;
data access;
integration;
security;
executive AI literacy;
employee capability;
AI policy;
risk classification;
measurement capability.
Some foundations enable multiple use cases.
That makes them strategically important.
Foundation work is not “non-value work”
Leadership can become impatient with:
governance;
data;
integration;
training.
These can look like costs rather than value.
But if one foundation unlocks ten high-value AI opportunities, it may be one of the highest-leverage investments on the roadmap.
The important question is:
Which foundations unlock the greatest amount of future value?
4. PORTFOLIO — What happens first?
Now prioritise.
Not every valuable opportunity should begin immediately.
Some may have:
greater strategic importance;
higher readiness;
lower implementation complexity;
better evidence;
more manageable risk.
Others may depend on foundations that do not yet exist.
For more on prioritising consistently, read How to Prioritise AI Use Cases: A Practical Framework for Business Leaders.
The roadmap should distinguish between:
NOW
Ready to execute.
NEXT
Attractive but dependent on current work.
LATER
Strategically relevant but not yet ready.
WATCH
Emerging opportunities or technologies worth monitoring.
This gives us the first practical Stratify roadmap tool:
NOW → NEXT → LATER → WATCH
The roadmap should show dependencies
Imagine leadership wants to deploy an AI agent that can:
access customer information;
update CRM;
send communications;
issue refunds.
That use case may depend on:
risk classification;
data controls;
agent permissions;
human oversight;
integration;
decision rights;
monitoring.
The roadmap should make those dependencies visible.
Otherwise leadership sees only:
Deploy customer-service agent — Q3
without seeing the organisational work required to make Q3 realistic.
For more on how the wider organisational system needs to work together, read What Is an AI Operating Model? A Practical Guide for Leadership Teams.
5. TRANSFORMATION — What needs to change in the organisation?
Technology implementation is only part of the roadmap.
AI may require changes to:
workflows;
roles;
management;
skills;
performance measures;
employee behaviour.
These need explicit roadmap activities.
For example:
Technology milestone
Deploy AI customer-service capability.
Transformation milestone
Redesign customer-service workflow.
People milestone
Train employees in exception handling.
Management milestone
Introduce new performance measures.
Value milestone
Reduce cost per resolved case.
That is a much more complete roadmap.
For more on the difference between deploying AI and transforming the business, read AI Transformation vs AI Adoption: What's the Difference — and Why Does It Matter?
6. VALUE — What evidence should appear?
Every roadmap should include value milestones.
Not simply:
system deployed
or:
pilot completed.
Examples:
Adoption reaches 60%.
Handling time reduces by 15%.
Conversion increases by 5%.
Backlog reduces by 30%.
£500,000 of validated capacity is created.
Cost per transaction falls by 10%.
This changes the roadmap from an activity plan into an outcome plan.
For more on ensuring someone owns those benefits, read Who Owns AI ROI? Why Every AI Investment Needs a Business Value Owner.
7. DECISIONS — What happens when evidence changes?
An AI roadmap should not assume everything on it will happen.
That is crucial.
Each major initiative should eventually reach a decision:
SCALE
CHANGE
PAUSE
STOP
This makes the roadmap adaptive.
An initiative that produces strong evidence may move forward faster.
An initiative that underperforms may stop.
A new technology may change priorities.
A regulatory development may alter sequencing.
For more on making those investment decisions, read When Should You Stop an AI Project? A Framework for Scale, Change or Stop Decisions.
The Stratify AI Roadmap Horizon
I would structure the roadmap across four horizons.
HORIZON 1 — ESTABLISH
0–3 months
Focus:
baseline;
strategy;
ownership;
governance;
portfolio;
immediate opportunities.
Typical outputs:
AI baseline.
Priority outcomes.
Use-case inventory.
Named owners.
Governance foundations.
Initial business cases.
HORIZON 2 — PROVE
3–6 months
Focus:
controlled implementation;
evidence;
adoption;
measurement.
Typical outputs:
priority pilots.
workflow redesign.
adoption activity.
risk assessments.
baseline measurement.
initial value evidence.
HORIZON 3 — SCALE
6–12 months
Focus:
expanding what works.
Typical outputs:
scaled use cases;
integration;
operating-model changes;
stronger automation;
agent deployment;
benefits realisation.
HORIZON 4 — TRANSFORM
12–24 months
Focus:
enterprise capability.
Typical outputs might include:
AI embedded into operating models;
cross-functional workflows;
greater agentic capability;
portfolio optimisation;
workforce redesign;
new products or services;
measurable enterprise value.
So the roadmap becomes:
ESTABLISH → PROVE → SCALE → TRANSFORM
This is one of the frameworks I would preserve for the future Stratify methodology.
Roadmaps should contain decisions, not just dates
Traditional roadmaps can create false certainty.
They say:
AI agent deployed — September
But what if the evidence in June shows the agent is not reliable enough?
The roadmap should instead contain decision gates.
For example:
June
Review pilot evidence.
Decision
Scale / Change / Pause / Stop.
If Scale
Proceed to wider deployment.
That creates an evidence-led roadmap.
The organisation commits to making a decision.
It does not pretend the outcome is already known.
The Stratify AI Roadmap Card
Every major roadmap initiative should have a simple card.
OUTCOME
What business result are we trying to create?
OWNER
Who is accountable?
USE CASE
What AI capability is involved?
DEPENDENCIES
What needs to be true first?
INVESTMENT
What resources are required?
RISK
What governance applies?
ADOPTION
Who needs to change behaviour?
EVIDENCE
What proves this is working?
DECISION DATE
When will leadership review it?
NEXT DECISION
Scale / Change / Pause / Stop.
This could become a practical consultancy template.
And eventually a core object inside Stratify Intelligence.
The roadmap needs multiple workstreams
A mature AI roadmap should not have only a:
Technology workstream.
It may need parallel workstreams across:
STRATEGY
Direction and portfolio.
TECHNOLOGY & DATA
Platforms, architecture, integration and data.
GOVERNANCE
Policy, risk, controls and oversight.
PEOPLE
Skills, adoption and workforce change.
TRANSFORMATION
Workflow and operating-model redesign.
VALUE
Business cases, measurement and benefits realisation.
This prevents leadership discovering six months later that technology moved ahead while the organisation did not.
Who should own the AI roadmap?
There should be one coordinated enterprise roadmap.
But responsibility will be distributed.
Executive leadership owns direction.
Business leaders own outcomes.
Technology owns enabling capability.
Risk and governance functions support control.
HR supports workforce capability.
Finance supports investment and value.
A CAIO, AI leader, transformation leader or Fractional CAIO may coordinate the roadmap.
The coordinator should not own every action.
They ensure:
dependencies are visible;
owners are clear;
decisions happen;
actions progress;
evidence is reviewed.
The Fractional CAIO Roadmap Review
This creates a very obvious recurring Fractional CAIO process.
Every month or quarter:
DIRECTION
Have priorities changed?
PORTFOLIO
Which opportunities are active?
DEPENDENCIES
What is blocking progress?
ACTIONS
What needs leadership attention?
EVIDENCE
What have we learned?
VALUE
What outcomes are emerging?
DECISIONS
What scales, changes, pauses or stops?
ROADMAP
What changes next?
The roadmap becomes a living executive management tool, not a document created once and forgotten.
Ten questions to test your AI roadmap
Ask:
Is the roadmap connected to business strategy?
Are desired business outcomes explicit?
Are AI opportunities prioritised?
Are organisational foundations included?
Are dependencies visible?
Does every major initiative have a named owner?
Are workforce and process changes included?
Are value milestones defined?
Are decision gates built into the roadmap?
Is the roadmap regularly updated as evidence changes?
If several answers are no, you may have an AI project plan rather than an AI roadmap.
What should leadership see?
Leadership does not need hundreds of tasks.
It needs a concise view of:
Where are we going?
What are we doing now?
What happens next?
What is blocked?
What evidence is emerging?
What value is being created?
What decisions are required?
That is the executive purpose of the roadmap.
An AI roadmap should reduce uncertainty, not pretend it doesn't exist
AI strategy involves uncertainty.
That is unavoidable.
The answer is not to create a detailed plan pretending leadership knows exactly what will happen over the next 24 months.
The answer is to establish:
direction;
priorities;
dependencies;
ownership;
evidence;
decision points.
Then adapt.
A good AI roadmap therefore does not say:
This is exactly what will happen.
It says:
This is where we're going, this is what we're doing next, this is what we need to learn, and this is when we'll decide what happens after that.
That is a much more useful management tool.
Frequently Asked Questions About AI Roadmaps
What is an AI roadmap?
An AI roadmap is a structured plan showing how an organisation will move from its current AI position towards its desired future state through prioritised initiatives, enabling capabilities, ownership, investment and measurable outcomes.
What is the difference between an AI strategy and an AI roadmap?
AI strategy defines where and why the organisation intends to use AI. The roadmap translates that direction into sequenced initiatives, capabilities, owners, milestones and decisions.
How long should an AI roadmap be?
A 12–24 month horizon can provide useful direction while allowing adaptation as technology, evidence and business priorities change. Near-term activity should generally be more detailed than longer-term plans.
What should an AI roadmap include?
It can include strategic outcomes, AI opportunities, priority use cases, organisational foundations, technology and data requirements, governance, workforce change, investment, value milestones and decision gates.
Who should own an AI roadmap?
Executive leadership should own strategic direction, while a CAIO, AI leader, transformation leader or Fractional CAIO may coordinate the roadmap across business and specialist functions.
How often should an AI roadmap be updated?
The roadmap should be reviewed regularly and when material evidence, business priorities, technology, risk or organisational capability changes.
Should an AI roadmap include governance?
Yes. Governance, ownership, risk classification, decision rights and controls are enabling capabilities required for responsible AI scale.
How do you prioritise an AI roadmap?
Prioritise according to strategic value, business impact, feasibility, organisational readiness, risk, dependencies, investment and evidence.
From AI Ambition to AI Execution
An AI strategy creates direction.
But leadership needs a practical route from ambition to execution.
Stratify's Executive AI Readiness Assessment helps organisations understand the foundations already in place and the gaps that need addressing before AI can scale.
Assess areas including:
Strategy and leadership
Commercial opportunity
People and adoption
Technology and integration
Governance and risk
Test Your AI Readiness
Because the purpose of an AI roadmap is not to predict the future perfectly.
It is to make the next important decisions visible.
Comments