Short answer
An AI roadmap template sets out which AI initiatives a business will run over the next 12 months, in what order, with owners, budgets, dependencies and the measure each must hit before the next starts. This version uses business goals, a prioritisation score, three horizons (foundations, first projects, scale), gate measures and a quarterly review, so the plan changes when the evidence does.
- Start from two or three business goals with numbers, not a list of tools.
- Score candidates on value, feasibility, data readiness and risk.
- Foundations first: policy, approved tools, data inventory and training.
- Nothing moves to the next horizon until its gate measure is met.
Use this template after you have run the AI readiness checklist and listed a few candidate workflows. It is built for a business of roughly 10 to 300 people planning the next 12 months. Fill in the five parts in order; each one feeds the next.
The AI roadmap template
Part A: Business goals
Two or three goals the roadmap serves. If an initiative does not move one of them, it does not go on the roadmap.
| Goal | Measure | Baseline | Target | By |
|---|---|---|---|---|
| [e.g. Respond to customer enquiries faster] | Median first response time | |||
Part B: Candidate scoring
Score each candidate from 1 to 5 on four criteria, then calculate:
Priority score = (Value × 2) + Feasibility + Data readiness + (6 − Risk). The range is 5 to 25.
| Criterion | 1 | 5 |
|---|---|---|
| Value | Little effect on any goal | Large, measurable effect on a goal |
| Feasibility | Needs skills and tools we lack | Doable with current tools and people |
| Data readiness | Data scattered, missing or unreliable | Data current and accessible |
| Risk | Little harm if it goes wrong | Serious harm to people, money or reputation |
| Initiative | Value | Feasibility | Data | Risk | Score | Horizon |
|---|---|---|---|---|---|---|
Part C: Roadmap by horizon
| Horizon | Initiative | Owner | Start | Gate measure | Budget (NZ$) | Depends on | Status |
|---|---|---|---|---|---|---|---|
| Foundations (months 0–3) | |||||||
| First projects (months 3–6) | |||||||
| Scale (months 6–12) |
Part D: Foundations checklist
- AI use policy approved and briefed (template).
- Approved tools on business plans, with personal accounts retired for work use.
- Data inventory for the first projects (template).
- Training needs assessed and the first sessions booked.
- Risk register and incident log in place.
- A named owner for AI across the business.
Part E: Quarterly review agenda
- Gate measures: which initiatives met their gate, and which did not.
- Spend against budget, including model usage and staff time.
- Incidents and near misses since the last review.
- Changes in tools, prices and models that affect the plan.
- Re-score remaining candidates and update the horizons.
- Decisions, owners and the next review date.
How to build the roadmap
- Agree goals with the decision-makers. One meeting, two or three goals, each with a measure.
- List candidates from real work. Ask team leads for the tasks that eat the most time; map the top ones with the AI workflow mapping template.
- Score them together. Two or three people score independently, then agree.
- Put foundations first. Foundations usually take weeks, not months, and every project depends on them.
- Set one gate per initiative. For a pilot, use the decision rule in the AI pilot scorecard.
- Book the quarterly reviews now. A roadmap without review dates becomes a wish list.
Worked example
Illustrative example: a 70-person Canterbury manufacturer wants to cut time lost searching for maintenance information and processing supplier paperwork.
| Initiative | Value | Feasibility | Data | Risk | Score |
|---|---|---|---|---|---|
| Supplier document extraction | 4 | 4 | 4 | 2 | 20 |
| Maintenance knowledge search | 4 | 4 | 3 | 2 | 19 |
| Quality inspection report triage | 5 | 2 | 2 | 3 | 17 |
| Customer quote drafting | 3 | 3 | 2 | 3 | 14 |
| Website chatbot | 2 | 4 | 3 | 4 | 13 |
| Horizon | Plan | Gate |
|---|---|---|
| Foundations, November to January | Policy, approved tools, data inventory for supplier documents and manuals, training for office and maintenance leads | Policy approved; all office staff on business accounts |
| First projects, February to April | Pilot supplier document extraction | Pilot score 75 or more, with all gates passed |
| Scale, May to October | Run extraction in production; pilot maintenance knowledge search; clean inspection data so triage can be re-scored | Extraction live with monitoring; search pilot scored |
Inspection triage scored highest on value but low on data readiness, so the roadmap fixes the data first instead of starting a project likely to fail. The website chatbot was dropped: it served no goal.
Common mistakes
- Starting from tools. “Roll out Copilot” is not a roadmap item; “cut enquiry response time” is.
- Too many projects at once. One or two at a time for a business this size.
- No gates. Without them, every pilot continues by default.
- Ignoring price and product timing. Introductory prices end: Gemini 4 Argon’s API price doubles from 1 December 2026 (Google), and Microsoft’s discounted Copilot Business price in New Zealand applies to the first year of annual commitments made by 31 December 2026 (Microsoft). Budget for the full price.
- Skipping the exit question. MBIE’s quickstart asks businesses to think about what they will do if something goes wrong or they want to stop using an AI tool (MBIE).
Next step
Write a business case for the first project on the roadmap, and plan the rollout with planning a safe AI rollout. If you want an outside view on priorities, our AI implementation roadmap service builds this plan with your team.
Sources
- AI Advisory Pilot announcement, New Zealand Government (Beehive.govt.nz), 13 January 2026
- Quickstart: Businesses using AI tools, Ministry of Business, Innovation and Employment, accessed 19 October 2026
- Gemini Developer API pricing, Google, last updated 16 October 2026
- Microsoft Copilot business pricing (New Zealand), Microsoft, accessed 19 October 2026
Written by the Levy Street team in Wellington. Last updated 19 October 2026.