THE SHORT ANSWER
Start with the business objective, the customers you can serve, the problem you solve and the reason they should choose you. Then decide where AI can improve research, execution or evaluation. An AI adoption plan is not a substitute for a marketing strategy.
A strategy must make some options less attractive
‘Use AI across all channels’ does not explain which market to pursue, what to offer or why the business should win. A strategy makes trade-offs. It might prioritise helping a narrow group evaluate a complex service over attracting a broad audience with inexpensive content.
Consider a hypothetical maintenance company that can support local commercial premises but lacks capacity for household emergencies. A plan built around total enquiry volume could overwhelm operations. A useful marketing choice is to reach the right buyers with an accurate service promise, even if that reduces the apparent size of the audience.
AI can help organise objections from existing conversations or test whether a draft brief contradicts itself. It cannot establish that an unsupported positioning claim is true simply by finding persuasive wording.
Write the decision brief before the technology brief
| Choice | Question to answer |
|---|---|
| Business objective | What valuable change are we trying to create, and within what constraints? |
| Customer problem | Whose difficulty are we solving, based on what evidence? |
| Positioning | Why should they choose us, and what proof supports that reason? |
| Channels | Where can we reach or assist them at the relevant moment? |
| Measurement | What outcome would justify continuing, changing or stopping? |
| AI enablement | Which step could AI improve, and who checks the result? |
If the customer evidence is weak, fund learning before scaling production. If fulfilment is the constraint, generating more demand may make the business less reliable. The brief should expose those limits rather than conceal them behind an innovation agenda.
Choose a small portfolio of useful experiments
Compare candidate uses by decision value, evidence quality, reversibility and review effort. Summarising approved research for an internal meeting is a different commitment from automatically changing offers shown to customers. A pilot should have an owner, a comparison and a stopping rule.
NIST's generative-AI profile provides risk-management guidance across the lifecycle. It is useful context for assigning responsibility; it is not a marketing strategy or a certificate that a particular use is safe.
Evidence & context: NIST
Review the business decision, not the adoption score
In the maintenance example, evaluate whether qualified enquiries become profitable, deliverable work. Also examine complaints, rejected jobs and the time needed to correct generated material. A lower cost per form submission can coexist with a worse commercial outcome.
Use audience research to challenge assumptions and analytics to test explanations. Keep a decision record: the assumption, the evidence, the action and what would make you revise it.
Sources & further reading
- Generative Artificial Intelligence Profile (NIST AI 600-1)
NIST. Risk-management guidance, including confabulation. It does not establish a universal error rate.
Examples and exercises are illustrative unless attributed to a source. No independent expert review is claimed.
A correction, a counterexample or an experience worth sharing?
Join the conversation ↗