Key takeaways
- An AI cannot follow a strategy that lives only in people's heads.
- Ask for the reasoning and the sources behind each recommendation, not only the answer.
- Test advice against explicit rules: target customers, positioning and things you never do.
- A person accepts or rejects each recommendation and records why.
Why do AI recommendations drift from strategy
AI recommendations drift because the model fills gaps with generic good practice. If it does not know that you sell only to regulated firms, it will happily suggest a campaign aimed at young start-ups.
Drift also comes from the question. Ask how to get more leads and you get advice about volume; ask how to win more mid-sized insurers without discounting and you get advice that fits the business you are actually building.
What should the AI know about your strategy
The AI needs the decisions that shape every sales choice: your ideal customer, your positioning, your pricing stance and your limits. Write them on one page in plain words.
- Who you sell to, and who you deliberately do not.
- The problem you solve and why you beat the alternatives.
- Your pricing stance, such as no first-year discounts below list.
- Channels and tactics you have ruled out, and why.
- The goals for this quarter.
How do you check a recommendation before acting on it
Check a recommendation with three questions: does it fit our target customer, does it match our positioning, and can we see why the AI suggested it? A suggestion that fails any one of them goes back.
Ask the assistant to cite its sources and state its assumptions. A recommendation built on a competitor's public pricing page is worth weighing; one with no traceable source is a guess.
For example, an assistant suggests targeting retail chains because they are hiring quickly. If retail sits outside your ideal customer, the hiring signal is real but the recommendation still fails the first test.
How do you stop the same mistakes repeating
Record why you rejected a recommendation and add the reason to the briefing the AI reads. Over a few weeks the briefing becomes a map of the strategy's edges, and suggestions that cross them become rarer.
Review the briefing whenever the strategy changes. An AI working from last year's ideal customer will keep pointing you at last year's market.
How to keep AI advice on strategy, step by step
Set the strategy down first; the checks only work against something written.
- Write your strategy on one page: target customer, positioning, pricing stance, ruled-out tactics and current goals.
- Store that page where the AI assistant reads it every time, not inside a single chat.
- Ask specific questions that name the customer and the constraint.
- Ask for the reasoning, assumptions and sources behind each recommendation.
- Check each recommendation against the written strategy and accept or reject it.
- Add the reason for every rejection to the briefing, and review the briefing each quarter.
Where Pebbles Ai fits
- FoundationCompany IntelCompany Intel keeps your business knowledge (Company, Approach, Offerings) and brand voice in one place that every assistant works from.
- FoundationAgent memoryTell an assistant a fact about your company or a project and it carries it into later chats; Company memory reaches colleagues' assistants, and you correct a fact in plain words.
- CapabilityResearch & analysisResearch replies come with inline numbered citations and a source list, so you can see where a recommendation came from.
Common questions
Can AI set our sales strategy?
It can help you research markets and test options, but strategy is a choice about where to compete. That choice belongs to the people accountable for it.
What if the AI keeps suggesting the same off-strategy idea?
Your briefing probably does not rule it out. Add the idea to the list of tactics you have rejected, with the reason, so the assistant has something to check against.
How often should we update the strategy briefing?
Whenever a decision changes it, and at least each quarter when you set new goals. Out-of-date context is a common reason good tools give poor advice.
About this page. Written by the Pebbles Ai team for people who run sales, marketing and growth. Last updated .
