AI strategy template with a worked example

AI strategy template

An AI strategy sets out where an organization will apply artificial intelligence first, the data, platforms and skills those uses depend on, and the rules that keep them safe and lawful. It is owned by the chief technology officer and agreed with the executive team, legal and compliance, and the leaders of the operations it will change.

Horizon: 2 years4 priorities · 10 outcomes

Use this template

What this aI strategy template includes

ContextThe situation, the challenges to solve, and what is deliberately left out.
AI ambitionWhat the organization wants AI to do for it, and what it will not use AI for.
Use case portfolioThe candidate uses of AI, scored and prioritized.
Data and technology foundationsWhat must be in place for the use cases to work.
Governance and riskThe policies and oversight that apply to AI use.
Skills and operating modelWho builds and runs AI, and how staff are prepared.
Priorities and measuresThree to five priorities, each with an owner, outcomes with a baseline and target, and the initiatives behind them.

AI strategy example

A complete aI strategy for a fictional organization: national staffing agency placing light industrial, clerical and skilled trades workers through 62 branches.

Worked exampleAI strategy

Organization	National staffing agency placing light industrial, clerical and skilled trades workers through 62 branches (1,500 staff)
Ambition	Fill client orders in half the time by putting the right existing candidate in front of a recruiter within minutes, with a person making every hiring decision.
Horizon	2 years

Context

Where we are and what we are choosing

The agency bills $780M a year and makes about 70,000 placements through 900 recruiters. The applicant tracking system holds 2.3 million candidate records, yet only 22% of placements come from it; recruiters repost jobs because search returns stale and duplicate profiles. First qualified submittal takes 31 hours on average and 68% of orders are filled. Recruiters spend 38% of their week on intake, screening calls and scheduling. Only 34% of associates are redeployed within 30 days of an assignment ending. A staff survey found 44% of recruiters paste candidate details into public chatbots, and two tools already rank applicants without a bias audit.

Challenges

The candidate database is the agency's largest asset and its least used. 18% of records are duplicates and only 27% have skills and availability confirmed in the last 12 months.
Speed decides who fills an order, and clients send the same order to three agencies. At 31 hours to first submittal, a third of orders are lost before a candidate is presented.
AI is already in use without controls. Our largest clients have started asking for bias audits and candidate notice for automated hiring tools in contract renewals, and we cannot supply either.

We will focus on

Matching and rediscovery of candidates already in the database, for light industrial and clerical orders where volume is highest.
Removing recruiter administration: intake notes, job postings, screening questions, scheduling and redeployment outreach.
A governance model in place before anything that touches candidate selection goes live.

We will not

Automated rejection or ranking that a recruiter does not review. It would save the most time and it is the use most exposed to discrimination claims.
Building our own models. We will use the applicant tracking vendor's features and a commercially hosted language model, and accept less differentiation.
Client-facing AI products such as workforce demand forecasting. Order history is too inconsistent across branches to support them before 2029.

AI ambition

Ambition	AI shortens the distance between a client order and a qualified person starting work. It does the searching, drafting, reminding and scheduling so that recruiters spend their time talking to candidates and clients.
Scope	Recruiting operations in all 62 branches: candidate search and matching, order intake, screening and scheduling, redeployment, and timesheet exceptions. Sales, finance and marketing may use approved general assistants but are not part of the funded program.
Out of bounds	No automated decision to reject, rank out or set pay for a candidate. No analysis of video, voice or facial expression. No scraping of personal social media. No inference of age, health, ethnicity or other protected traits, and no candidate data in tools outside the approved list.

Use case portfolio

Use case	Value	Feasibility	Decision
Candidate match and rediscovery from the existing database	High: 2.3 million records, cuts time to first submittal	Medium: needs deduplicated profiles and a skills taxonomy	Fund first; recruiter reviews every shortlist
Redeployment alerts before assignments end	High: 21,000 associates on assignment each week	High: end dates and skills already in the system	Fund first
Order intake notes and job posting drafts	Medium: saves about 20 minutes per order	High: low risk, recruiter edits the output	Roll out to all branches in H1 2027
Conversational pre-screen and interview scheduling	High: 60% of applicants drop out before a recruiter call	Medium: vendor tools exist; needs bias audit and notice	Pilot in 6 branches, then decide
Timesheet and payroll exception triage	Medium: 3.1% of timesheets need correction	High: structured data, clear rules	Second wave, 2028
Automated ranking with auto-reject below a score	High in hours saved	Low: legal exposure and client objections	Rejected
Branch-level client demand forecast	Medium: better recruiter allocation	Low: order data inconsistent across branches	Deferred until data is fixed

Data and technology foundations

Data readiness	Not ready for matching today. Records must be deduplicated, resumes parsed into one skills taxonomy of about 1,800 skills and certifications, and availability refreshed by text message. Assignment history and client feedback, which predict a good match better than a resume, need to be joined to the candidate record.
Platform	The applicant tracking system stays the system of record and recruiters work inside it. A language model is accessed through our cloud provider under a contract that bars training on our data and keeps processing in country. A search index over candidate profiles supports matching, and every prompt and output that involves a candidate is logged.
Build, buy or partner	Buy matching, screening and scheduling features from the applicant tracking vendor and specialist suppliers where they pass our risk review. Build only the redeployment model and the connections between systems, since those depend on our own assignment data. Partner with an independent auditor for bias testing.

Governance and risk

Policy	An AI use policy lists approved tools and the data each may handle. Candidate personal data may only enter tools on the approved list. Candidates are told when AI assists with their application and can ask for a person to handle it instead.
Human oversight	A recruiter makes every decision to submit, reject or place a candidate. AI output is a suggestion with the reasons shown. Recruiters can see and search candidates the system did not surface, and overrides are recorded and reviewed monthly.
Risk review	Every use case that touches candidate selection is risk-assessed before pilot and has an independent bias audit before launch and every 12 months after. The test is the selection rate for each group against the highest group, with action required below 0.8. Vendors must supply their own audit results.
Accountable owner	The chief technology officer is accountable. An AI review board with the general counsel, the head of compliance, the chief operating officer and two branch directors meets monthly and approves each use case before it goes live.

Skills and operating model

Team	A central team of nine: a product director, three data engineers, two machine learning engineers, an analyst, a governance lead and a change lead. It reports to the chief technology officer and is funded at $2.4M a year, including licenses and audit fees.
Training	All 900 recruiters complete a three-hour course on the approved tools, what may not be entered into them, and how to challenge a suggested match. Branch managers get an extra session on reading override and adverse impact reports.
Change support	Each branch names one recruiter as its AI lead with four hours a week to coach colleagues and report problems. New features go to six pilot branches for eight weeks before national rollout, and recruiter bonus plans are unchanged during a pilot.

Priorities and measures

Priority 1Owner: Director of AI Products

Fill orders from the candidates already on file

Only 22% of placements come from 2.3 million existing records because search cannot be trusted. Fixing the data and adding matching is the fastest route to cutting the 31-hour submittal time.

Outcome	Baseline	Target	Due
Average time from client order to first qualified submittalMilestones: 23 hours by Q4 2027	31 hours	14 hours	Q4 2028
Placements sourced from the existing candidate databaseMilestones: 30% by Q4 2027	22%	40%	Q4 2028
Active candidate records with skills and availability confirmed in the last 12 monthsMilestones: 48% by Q4 2027	27%	65%	Q4 2028

Initiatives

Deduplicate candidate records and map parsed resumes to one skills taxonomy
Refresh availability by text message every 90 days for candidates placed in the last three years
Launch match suggestions inside the applicant tracking system, with reasons shown and recruiter review required
Add assignment history and client feedback to the signals used for matching

Targets are illustrative. Adapt them to your baseline before you start.

Priority 2Owner: VP Branch Operations

Keep associates working between assignments

Two thirds of associates leave the agency's books when an assignment ends, and each one has to be replaced by new recruiting. End dates are known weeks ahead and nobody acts on them.

Outcome	Baseline	Target	Due
Associates redeployed within 30 days of an assignment endingMilestones: 41% by Q4 2027	34%	48%	Q4 2028
Applicants who drop out between application and recruiter screenMilestones: 52% by Q4 2027	60%	40%	Q4 2028

Initiatives

Alert recruiters 14 days before an assignment ends with three open orders that fit the associate
Pilot conversational pre-screening and self-service interview scheduling in six branches after a bias audit
Send associates matched openings by text message with one-tap interest

Targets are illustrative. Adapt them to your baseline before you start.

Priority 3Owner: Chief Operating Officer

Take administration off recruiters' desks

Recruiters spend 38% of their week on tasks that need no judgment. Returning that time is how fill rate rises without adding headcount.

Outcome	Baseline	Target	Due
Recruiter time spent on intake, screening calls and schedulingMilestones: 31% by Q4 2027	38%	24%	Q4 2028
Placements per recruiter per monthMilestones: 6.8 placements by Q4 2027	6.5 placements	7.4 placements	Q4 2028
Client orders filledMilestones: 71% by Q4 2027	68%	76%	Q4 2028

Initiatives

Draft job postings and intake summaries from the order call for the recruiter to edit
Automate interview and start-date reminders to candidates and client supervisors
Run a time study in 12 branches each half year to measure where recruiter hours go
Triage timesheet exceptions automatically from 2028

Targets are illustrative. Adapt them to your baseline before you start.

Priority 4Owner: Chief Technology Officer

Put controls around AI in hiring

Two ranking tools run without an audit and 44% of recruiters use public chatbots with candidate data. One complaint or a failed client review would stop the whole program.

Outcome	Baseline	Target	Due
AI tools touching candidate selection with a current independent bias auditMilestones: 50% by Q2 2027	0%	100%	Q4 2027
Recruiters reporting use of unapproved AI tools with candidate dataMilestones: 15% by Q4 2027	44%	5%	Q4 2028

Initiatives

Publish the AI use policy and approved tool list, and give every recruiter an approved assistant
Audit the two ranking tools already in use and switch off any that fail
Set up the AI review board and a register of every use case with its owner and risk rating
Add candidate notice and a request-a-person option to every application flow

Targets are illustrative. Adapt them to your baseline before you start.

Blank aI strategy template

The same structure with a prompt in every field. Copy it, or download it as a Word or Excel file.

Your planAI strategy

Organization	[Describe your organization in a few words, with its size]
Ambition	[One sentence on where you want to be at the end of the plan]
Horizon	[How long the plan runs, for example 12 months]

Context

Where we are and what we are choosing

[Where things stand today: the facts that make this strategy necessary]

Challenges

[A problem the strategy has to solve]
[A problem the strategy has to solve]

We will focus on

[Something you are choosing to concentrate on]
[Something you are choosing to concentrate on]

We will not

[Something you are deliberately leaving out]
[Something you are deliberately leaving out]

AI ambition

Ambition	[Your ambition]
Scope	[Your scope]
Out of bounds	[Your out of bounds]

Use case portfolio

Use case	Value	Feasibility	Decision
[Use case]	[Value]	[Feasibility]	[Decision]
[Use case]	[Value]	[Feasibility]	[Decision]
[Use case]	[Value]	[Feasibility]	[Decision]

Data and technology foundations

Data readiness	[Your data readiness]
Platform	[Your platform]
Build, buy or partner	[Your build, buy or partner]

Governance and risk

Policy	[Your policy]
Human oversight	[Your human oversight]
Risk review	[Your risk review]
Accountable owner	[Your accountable owner]

Skills and operating model

Team	[Your team]
Training	[Your training]
Change support	[Your change support]

Priorities and measures

Priority 1Owner: [Role accountable for this priority]

[Name the priority in a few words]

[Why this priority matters now, in one or two sentences]

Outcome	Baseline	Target	Due
[Metric you will track]	[Where it is today]	[Where it should be]	[When]
[Metric you will track]	[Where it is today]	[Where it should be]	[When]
[Metric you will track]	[Where it is today]	[Where it should be]	[When]

Initiatives

[A project or action that moves the outcomes above]
[A project or action that moves the outcomes above]
[A project or action that moves the outcomes above]
[A project or action that moves the outcomes above]

Priority 2Owner: [Role accountable for this priority]

[Name the priority in a few words]

[Why this priority matters now, in one or two sentences]

Outcome	Baseline	Target	Due
[Metric you will track]	[Where it is today]	[Where it should be]	[When]
[Metric you will track]	[Where it is today]	[Where it should be]	[When]

Initiatives

[A project or action that moves the outcomes above]
[A project or action that moves the outcomes above]
[A project or action that moves the outcomes above]

Priority 3Owner: [Role accountable for this priority]

[Name the priority in a few words]

[Why this priority matters now, in one or two sentences]

Outcome	Baseline	Target	Due
[Metric you will track]	[Where it is today]	[Where it should be]	[When]
[Metric you will track]	[Where it is today]	[Where it should be]	[When]
[Metric you will track]	[Where it is today]	[Where it should be]	[When]

Initiatives

[A project or action that moves the outcomes above]
[A project or action that moves the outcomes above]
[A project or action that moves the outcomes above]
[A project or action that moves the outcomes above]

Priority 4Owner: [Role accountable for this priority]

[Name the priority in a few words]

[Why this priority matters now, in one or two sentences]

Outcome	Baseline	Target	Due
[Metric you will track]	[Where it is today]	[Where it should be]	[When]
[Metric you will track]	[Where it is today]	[Where it should be]	[When]

Initiatives

[A project or action that moves the outcomes above]
[A project or action that moves the outcomes above]
[A project or action that moves the outcomes above]
[A project or action that moves the outcomes above]

How to adapt it

01
Start from the operating numbers the business already tracks, such as time to fill or cost to serve, and ask where AI could move them. Avoid starting from a list of tools.
02
Score each candidate use case on value and feasibility, and record the rejected ones with the reason. The rejected list stops the same idea returning every quarter.
03
Write the out-of-bounds list with legal and compliance before choosing use cases. In regulated activities such as hiring, lending or healthcare it will remove some of the most attractive ideas.
04
Check data readiness for the first two use cases in detail. Most delays come from duplicate, stale or unjoined records, not from the model.
05
Decide who is accountable and which forum approves a launch, and give that forum the power to stop one.
06
Budget for training and for people in each team who support their colleagues. Adoption is slower than the technology.

How to run it

A plan only works if it is reviewed on a rhythm. This is the cadence the example assumes.

Check-in	Outcome owners update progress and confidence weekly.
Review	Priority owners review outcomes and unblock initiatives monthly.
Reset	Leadership revisits priorities and targets quarterly.

Tability keeps this cadence for you. Each priority becomes an objective, each outcome a key result with its baseline and target, and each initiative a tracked task. Owners get a check-in reminder and the plan shows what is on track.

Track this plan in Tability

Still comparing tools? See the 10 best strategy software tools.

Common mistakes

Funding ten pilots at once, none of which has an owner in operations or a measure it is expected to move.
Treating governance as a document written after launch. In high-risk uses, the audit and notice requirements shape the design.
Measuring adoption by logins or prompts sent when the point is a business result such as hours saved or orders filled.
Ignoring the AI staff already use. Banning public tools without offering an approved alternative only hides the practice.

Questions, answered.

What should an AI strategy template include?

It should include the ambition and what is out of bounds, a scored portfolio of use cases with a decision on each, the data and platform foundations needed, governance and risk controls, and the skills and operating model. Add measurable business outcomes so that AI spending is judged on results and not on the number of pilots.

How do you prioritize AI use cases?

Score each use case on business value and on feasibility, then fund the few that are high on both. Value means a measurable effect on revenue, cost or risk. Feasibility covers data readiness, technical difficulty, legal exposure and how much working practice must change. Record the decision and the reason for every candidate, including those rejected.

Who should own the AI strategy?

One executive should be accountable, usually the chief technology officer or chief information officer. Because AI changes how operational teams work, the leaders of those teams must own the outcomes for their use cases. Legal and compliance need a formal role in approving anything that affects customers, candidates or employees.

Does a company need an AI policy before starting?

Yes, a short one. Staff are almost certainly using public AI tools already, so the first policy should say which tools are approved and what data may be entered. A fuller governance model, with risk assessment and audits, must be in place before any use that affects decisions about people goes live.

How long does it take to see results from an AI strategy?

Drafting and summarizing tools can show time savings within three months. Use cases that depend on cleaning data or changing a core workflow usually take 9 to 18 months to move a business measure. Plan for both: early wins that build confidence, and a two-year horizon for the outcomes that justify the investment.

AI strategy template

Get the Word file

What this aI strategy template includes

AI strategy example

Context

AI ambition

Use case portfolio

Data and technology foundations

Governance and risk

Skills and operating model

Fill orders from the candidates already on file

Keep associates working between assignments

Take administration off recruiters' desks

Put controls around AI in hiring

Blank aI strategy template

Context

AI ambition

Use case portfolio

Data and technology foundations

Governance and risk

Skills and operating model

[Name the priority in a few words]

[Name the priority in a few words]

[Name the priority in a few words]

[Name the priority in a few words]

Get the Word file

How to adapt it

How to run it

Common mistakes

Questions, answered.

What this aI strategy template includes

AI strategy example

Fill orders from the candidates already on file

Keep associates working between assignments

Take administration off recruiters' desks

Put controls around AI in hiring

Blank aI strategy template

[Name the priority in a few words]

[Name the priority in a few words]

[Name the priority in a few words]

[Name the priority in a few words]

How to adapt it

How to run it

Common mistakes

Questions, answered.

Related templates

Product strategy template

IT strategy template

Data strategy template