Conversations
Prompt like a teammate, not a ticket
Give each message a clear objective, context, and success criteria—your agents will reward you with better output.
- State the goal first: e.g., "Book a 30-minute intro" or "Summarize injury impact".
- Attach assets: drop briefs, PDFs, or URLs so the agent can cite the same source your team uses.
- Confirm approvals quickly: when the agent asks for sign-off, reply Approve or provide edits inline.
- Reset context when needed: start a new thread for a new campaign so analytics stay clean.
Persona tuning
Iterate on the character sheet intentionally
Small edits go a long way. Treat the sheet like your brand bible in prompt form.
- Keep traits and communication style succinct—three to four high-signal words each.
- Update the do/do-not list whenever legal or policy teams flag a new guideline.
- Refresh examples every quarter to reflect winning conversations and new objections.
Versioning idea
Track changes in a shared doc or ticket. When you push an update, note the version inside the character sheet so analysts can tie performance shifts to prompt revisions.
Operationalize
Keep agents fast and trusted
Blend automations with human checkpoints.
- Reply with a clear Approve or next edit so runs keep moving without back-and-forth.
- Review recent conversations each week to make sure tone and guardrails are holding.
- Keep the Approvals tab open during busy periods or enable email notifications so nothing stalls.
- For prediction agents, maintain a lightweight results sheet to compare projected vs actual outcomes over time.
Measurement
Measure impact from day one
Quantify how agents support the business so you can justify expansion.
Business assistants
- Meetings booked vs targets.
- Approval turnaround time (how long drafts wait).
- Response acceptance rate (messages approved without edits).
Prediction assistants
- Number of analyses delivered per week.
- Edge accuracy vs actual closing line.
- Research turnaround time (from request to draft).