Begin with a decision and a reporting scope
Fin is the AI Marketing Strategist inside SharkBlade. Its role is to help explain supported workspace evidence and explore improvement priorities. A useful question identifies the business objective, the period and the kind of decision you need to make. “What should we do?” leaves much of that context unstated. A more useful prompt is: “For the completed month, which underlying KPIs explain our lowest category assessment, and what should we verify before assigning work?” This is an example question, not a promise that every response contains the same fields.
Three useful question patterns
| Task | Example question | What to verify in the response |
|---|---|---|
| Explain | Which measures explain this category result? | Named KPIs and the correct period |
| Prioritize | Which issue affects the enquiry journey we care about? | A connection to the business objective |
| Plan | What bounded action could we take next? | Owner, assumptions and acceptance evidence |
Ask follow-up questions when the answer jumps from an observation to a conclusion. “What evidence would distinguish a tracking problem from a real decline?” is often more useful than requesting a longer version of the same recommendation.
Distinguish workspace explanation from research
The Fin interface includes a Research current strategies option. A question about saved workspace evidence and a request for fresh external research have different evidence needs. Read the response for its source basis and verify current external claims before using them in a campaign or proposal.
For example, a change in paid-ad efficiency could lead to a collection check, a campaign investigation or an offer review. An external recommendation about advertising policy or a new platform feature needs current vendor documentation; it cannot be established from the demo’s synthetic metrics.
Keep the delivery boundary clear
Fin’s guidance is not the same as making changes to a website, advertising account or lead intake process. Your team or ZeroDark carries out the agreed work. Before delivery, verify required access, dependencies and the acceptance criteria. A hypothetical action record might say: “Review the primary service page’s form-success event and verify that one test enquiry creates the intended record.” That is a testable assignment. “Let AI increase our leads” does not specify what will change or how it will be checked.
Preserve a reviewable trail
Save the question, the relevant period, the response assumptions and the evidence used to accept or reject the advice in your team’s working record. When the next report arrives, revisit the assignment and inspect what actually changed. The demo captures show product behavior using fictional data. They do not establish accuracy for every possible question or prove a real customer’s performance improved. Use the dedicated Fin page to explore the interface and the monthly-review guide to place its advice in a broader decision process.