AI Roadmap Planning
The order you build things in decides whether the programme survives its first year.
What is AI roadmap?
An AI roadmap is a sequenced delivery plan setting out which pieces of work happen in which order, what each depends on, how long each takes, what it costs, and the points at which the business can stop or change direction.
Programmes rarely fail because a single project was wrong. They fail because the largest project went first, nothing visible existed by the third budget review, and the funding quietly evaporated.
Sequencing fixes that. Something useful reaches daily use early, earns its keep, and pays for what follows. The most interesting project waits until the programme has credibility.
A roadmap also drags dependencies into the open. Work that cannot begin until two systems are connected or a dataset is consolidated is far cheaper to discover on paper than in month three of a build.
Why our plans expect you to change course
Duara plans AI programmes for Kenyan organisations from Nairobi. We have watched enough initiatives lose funding in their second quarter to treat sequencing as the main event rather than an administrative detail.
We build plans that assume you will change your mind. Every stage ends in a genuine decision, and stopping is designed in as a normal outcome rather than treated as failure.
Who this is for
- You have several AI ideas and one budget
- A previous project stalled after the pilot
- Your board wants checkpoints before releasing more money
- Departments disagree about what comes first
Industries we serve
- Banking and microfinance
- Insurance
- Logistics and transport
- Manufacturing
- Retail chains
- Public sector bodies
Problems this solves
Roadmap work is usually commissioned after a programme has already lost momentum once.
- The biggest project went first. Months pass with nothing in daily use, so each budget conversation gets harder until the funding stops.
- Dependencies discovered mid-build. Work halts because a dataset was never consolidated or a system turns out to have no usable interface.
- No defined stopping points. The programme continues on momentum because nobody agreed where a genuine continue-or-stop decision belongs.
- Stages without owners. Work slips quietly because accountability sits with a committee rather than a named person.
What we deliver
A staged plan with dependencies mapped, owners named, durations estimated and stop points built in, ordered so early value funds later work.
A stage by stage plan with named owners and timelines
Cost and duration estimated for every stage
Decision points where you continue, adjust or stop
The preparatory work required before later stages can start
A dependency map showing exactly what blocks what
A one page version for the board
Features
- Value-first ordering. The stage that pays back soonest goes first, so the programme funds its own continuation.
- Dependency mapping. What must be finished before each stage can begin is written down before anyone commits money.
- Named internal owners. Every stage has an accountable person inside your business, not only on the delivery side.
- Designed stop points. Each stage closes with a real decision, so halting is an ordinary outcome rather than an admission of failure.
- Rolling revision. The plan is updated as stages complete instead of being written once and quietly ignored.
Benefits
- Funding survives the first year. Visible results land before anyone starts questioning the line item.
- Expensive surprises move earlier. Dependencies surface during planning rather than during construction.
- Board conversations get shorter. Directors see checkpoints instead of an open-ended commitment with no exit.
- Changing direction becomes cheap. Redirecting at a checkpoint costs a fraction of abandoning a half-finished system.
Technologies we use
- Dependency mapping. Documented links between datasets, systems and stages so blockers are visible before commitment.
- Cost and duration modelling. Estimates per stage, including the running costs that begin the moment a stage goes live.
- Your existing project tools. The plan lives wherever your teams already work, including a shared spreadsheet if that is the truth.
- Integration feasibility checks. Which of your systems can realistically be connected, and what each connection would involve.
- Benefit tracking method. A simple, repeatable way to check whether each completed stage delivered what it promised.
Roadmapping is planning work. The tools involved exist to make dependencies and costs visible, not to build anything yet.
How the process works, step by step
Roadmap work takes two to three weeks and produces a plan your board can approve stage by stage.
-
Step 1
Free planning call
We establish what has already been attempted, what funding exists, and who has to approve each release of it.
-
Step 2
Opportunity inventory
We collect every candidate project already proposed inside the business and put them on one comparable footing.
-
Step 3
Dependency and effort mapping
We work out what each candidate depends on and roughly what it would take to deliver.
-
Step 4
Sequencing workshop
We order the work with your team, arguing the trade-offs openly rather than presenting a finished answer.
-
Step 5
Plan issued with checkpoints
You receive the staged plan, the dependency map, the costs and the board summary.
Quality assurance and how we work
What we commit to on roadmap work.
- Stopping is a designed outcome. Every stage ends in a decision that includes stopping, and we never treat that as a failure of the plan.
- Estimates carry their confidence. Durations and costs are labelled as rough or firm, so nobody plans cash flow against a guess.
- Dependencies are stated, not implied. If a stage cannot start until something else finishes, that appears in writing before approval.
- The plan is revised, not defended. When reality contradicts the sequence, we change the sequence rather than argue for the document.
- Owners are agreed with the owners. Nobody is named as accountable for a stage without having agreed to it first.
Frequently asked questions about AI roadmap
What is an AI roadmap?
A sequenced delivery plan setting which work happens in which order, what each stage depends on, what it costs, and where the business can stop.
How far ahead should a roadmap look?
Twelve to eighteen months. Beyond that, both the technology and your own priorities shift enough to make the detail guesswork.
How does this differ from a strategy?
A strategy decides what is worth doing. A roadmap decides the order, the dependencies and where the checkpoints sit.
What if our priorities change halfway?
Expected. The decision points exist precisely so you can change direction deliberately rather than discovering a problem too late.
Who owns each stage?
A named person inside your business, alongside whoever delivers it. Stages owned by a committee tend to slip.
How long should a single stage run?
Four to eight weeks. Much longer and the checkpoint arrives too late to influence anything.
Can stages run in parallel?
Occasionally, where they share no dependencies. More often parallel work means neither finishes on time.
What happens at a decision point?
You compare what the stage delivered against what it promised, then choose to continue, adjust or stop.
Does the roadmap include running costs?
Yes. Running cost begins the moment a stage goes live, so it appears from that stage onward.
What if a dependency cannot be satisfied?
The stage is resequenced or dropped. Mapping dependencies first is what makes that a cheap decision.
Is a roadmap worthwhile for a small business?
Yes, though shorter. Often three stages across a year rather than a large portfolio of parallel work.
Do you keep updating it?
Yes. A roadmap written once and never revised stops describing reality within about a quarter.
Can another supplier deliver the plan?
Yes. It is written so any competent team, including your own, can execute it without us.
What is the most common sequencing mistake?
Putting the largest and most interesting project first, leaving nothing visible when the budget is next reviewed.
How do you compare very different proposals?
We put them on one footing using expected return, effort and dependency weight, then argue the ranking openly with your team.
Do you include work that is already underway?
Yes. Existing commitments are part of the sequence, and sometimes the recommendation is to stop one of them.
What if two departments both want to go first?
The workshop settles it in the room using the same scoring for both, rather than by seniority.
Does the board need to attend?
Not the workshops, but they should see the one page summary before approving the first stage.
How much does roadmap work cost?
It depends on how many candidate projects exist. We quote a fixed figure after the free planning call.
Can you roadmap work we will build ourselves?
Yes. That is a common arrangement and the plan is written to be handed straight to an internal team.
Glossary of terms
Planning vocabulary used in this document, defined plainly.
- Roadmap
- A sequenced plan setting what gets built when, with dependencies, owners and decision points.
- Dependency
- Something that must be completed before another piece of work is able to start.
- Decision point
- A scheduled moment where the business chooses to continue, adjust or stop the programme.
- Stage
- A discrete block of work with its own scope, cost, owner and defined outcome.
- Sequencing
- Choosing the order of work so value arrives early and pays for what comes next.
- Scope creep
- Work quietly expanding beyond what was agreed, generally without the budget expanding to match.
Related services in AI Consulting
Talk to us about AI roadmap
Bring us whatever proposals are already circulating internally. The planning call is free and often shortens the list considerably.
Start your project