START
Internal work with human review
Campaigns, draft replies and reports; final selection follows data and capacity checks.
StrategyDirection01 / 05
Every team has an AI tool.
What is the company’s plan?
An AI strategy explains where this change should take the company: value for customers and the business, where to invest, and what to leave out for now. Start with that company-wide direction; connect opportunities and execution to it next.
Company AI strategyVision, goals, priorities, people, data and tools, governance, investment and review: one company-level document.
| Company goal | Today → 6-month target | Quality condition | Owner |
|---|---|---|---|
| G1 / Support | 10 → 8 minutes per request | Answer quality: at least 4.3/5 | Support lead |
| G2 / Growth | 5 → 3 days per campaign | Brand quality ≥ 4/5; conversion ≥ 2.2% | Growth lead |
| G3 / Finance | 4 → 2 days per monthly report | Every reported figure must have a source | Finance lead |
The portfolio follows these goals.
For G2, the internal campaign is one opportunity. That opportunity enters GODAE; G1 and G3 remain in the company programme.
FIELD, its decisions and figures are educational examples. A company’s actual artifact requires its own data, owners and evidence.
Strategy makes choices within real constraints.
Which teams start, what time and funding are available, and what knowledge and access do they need? Record what you will defer too. Revisit these choices using evidence from the first implementations.
NotionCompany AI strategy / v0.1ILLUSTRATIVE · DRAFTDEFER
Autonomous financial authority
Failure disruption is high. Financial decisions and final sends stay with people for now.
RESOURCES
Time, funding and shared capability
The sponsor confirms first-wave capacity and funding before delivery commitments.
REVIEW
Evidence updates the strategy
Portfolio review: value, adoption, cost, risk and next decisions.
FIELD, its decisions and figures are educational examples. A company’s actual artifact requires its own data, owners and evidence.
Company AI strategy
Use these company goals to collect departmental opportunities in one backlog.
NotionCompany AI strategy / v0.1ILLUSTRATIVE · DRAFTFIELD company AI transformation strategy
Owner: programme leadHorizon: 6 monthsReview: monthly
01Executive decision
FIELD is a 50-person online commerce company. People use AI individually, while team work still stalls across scattered knowledge, files and approvals. The six-month direction is to bring trusted company knowledge into daily work and improve three business outcomes.
02Objectives and success measures
| Goal / owner | Example baseline | 6-month target | Quality condition |
|---|---|---|---|
| G1 / Support lead | 2,000 requests/month; 10 min/request | 8 min; 20% less handling time | Quality ≥ 4.3/5 |
| G2 / Growth lead | 4 campaigns/month; 5-day lead time; 2% conversion | 3 days; conversion at least 2.2% | Brand quality ≥ 4/5; approved release |
| G3 / Finance lead | 4 days/monthly report | 2 days; 50% less elapsed time | Every figure traceable to a source |
03Strategic choices
Invest in three shared capabilities: trusted retrievable knowledge, bounded workflows with human review, and evaluation alongside execution. AI-101, the internal campaign, is the first pilot because publication can remain human-controlled. AI-102 support reply drafts and AI-103 financial reporting need knowledge and access reviews before delivery commitments.
Defer autonomous financial decisions, unapproved sending and company-wide tool purchases. Time improvement alone does not justify expansion; quality and adoption also require review.
04People, capacity and investment
| Responsibility | Example weekly capacity | Main accountability |
|---|---|---|
| Executive sponsor / CEO | 2 hours | Resources, priority conflicts and investment decisions |
| Programme lead | 4 hours | Portfolio, roadmap and value review |
| Analysis and architecture | 8 hours | Discovery, brief and system design |
| Build and evaluation | 20 hours | Execution, tests, quality and correction |
| Team champions | 4 hours combined | Practice, adoption and feedback |
These are planning assumptions to confirm against actual calendars. Approve model, tool and data costs before connection. Report freed time as capacity, not cash savings, unless financial savings actually occur and are measured.
05Operating boundaries and tool choices
Use only approved knowledge and data. Human review approves the exact AI draft before publication. The architect evaluates tools against quality, cost, provider eligibility and data handling. A tool’s presence in this example does not establish company eligibility.
The business owner can stop the workflow. Access or quality failures stop execution and return work to the brief and manual process. Model, source and tool changes require new tests and evaluations.
06First wave and the next decision
| Period | Programme output | Evidence condition |
|---|---|---|
| Days 1–30 | Strategy, backlog, owners, knowledge/access; AI-101 brief | Sponsor confirms resources; access and criteria defined |
| Days 31–60 | Build/test AI-101; team practice; support knowledge review | Tests, quality and stop procedure before real use |
| Days 61–90 | Review results/adoption; iterate or release an accepted version | Expand only with quality, value, adoption and operating ownership |
07Review and six-month direction
Each month, the programme lead reports value, full cost, adoption and risk to the sponsor. Each initiative receives an expand, iterate, pause or retire decision. The next wave depends on team capacity, pilot evidence and knowledge/data dependencies. Update company goals and the roadmap using the same evidence.
FIELD, its decisions and figures are educational examples. A company’s actual artifact requires its own data, owners and evidence.