AI assistants (MCP)
Example prompts
Questions and requests that work well with an AI assistant connected to Optima, grouped by role.
You talk to a connected assistant in plain language; it chooses the tools. The prompts below work with the key setups on Permissions and privacy. Prompts that create something need the campaign drafter or report builder roles; the others need only the viewer roles.
Tips
- Name the environment's terms. "Customers" and "orders" work best when they match your entity and event type names. Ask the assistant to describe the data model first if you are unsure.
- Give a time range. "Last 30 days" or "in September" keeps answers small and fast.
- Ask for counts before lists. "How many…" is cheaper than "show me all…", and large lists are cut off.
- Review what it drafts. Segments, campaigns and journeys an assistant creates are drafts or inactive. Open them in the console before you use them.
Business owners and campaign managers
- "How did last month's campaigns perform? Compare reach, conversions and lift against the control group."
- "Which journey step loses the most people in the welcome journey?"
- "Using last-touch and linear attribution, which campaigns get credit for purchases this quarter?"
- "Show me the headline analytics for this month and anything unusual compared with last month."
Campaign creators
- "Draft a segment of customers who bought twice in the last 90 days but not in the last 30, and tell me how many there are."
- "Create a draft campaign for that segment called 'Win-back October'."
- "Draft a journey: send the win-back email, wait three days, and send a reminder to those who did not open it. Check it but don't save it yet."
- "Render the 'Win-back' email for profile 123 and check whether that profile would be reached."
- "Which offers would profile 123 get right now, and why are the others excluded?"
Analysts
- "Run the 'Weekly revenue' report and summarise the trend."
- "Build a funnel from product view to add to cart to purchase for the last 14 days."
- "Create a metric for daily sign-ups that alerts me when it drops more than 20 % below its usual value."
Integration specialists
- "Describe the data model: entity types, event types and custom datasets."
- "Did the data connector runs in the last 24 hours succeed? Which ones failed?"
- "Show me the last five 'purchase' events for profile 123 with all their fields."
- "Which identity cluster would the current identity rules form for the profile with email
jane@example.com?"
Privacy officers and auditors
- "Show the audit log entries for profile 123 in the last week: who read it, under which purpose and which attributes."
- "Which consents does profile 123 hold, and what receipts were recorded?"
- "Show yesterday's audit log entries made by the key
claude-analyst."