‹ Sales recipesSales · Agent Blueprint
Sales enablement, messaging
Recurring objections across mid-market deals. The same pushback, said three different ways, by ten different prospects this quarter.
Side by side: token usage, with and without Paperbase
~136k tokens saved · 85% less work for the agentWithout Paperbase
~160k tokens
- Pull every deal record and every lost-deal note from the CRM.
- Search call transcripts for the word "objection" and related terms.
- Skim deal reviews and win-loss analyses across the quarter.
- Re-read discovery calls to reconstruct what prospects actually pushed back on.
- Group objections by theme from memory and rough categorization.
- Write messaging from anecdotes, not the actual record.
With Paperbase
~24k tokens
- Pull the quarter's Interactions across deals, with extracted Sentiment.
- Ask Paperbase to cluster recurring objections by theme, with quoted examples.
- Pull Decisions on messaging and positioning with Rationale.
- For each objection theme, get the deals it appeared in, the stage it killed, and the language used.
- Output the objection map with the exact words prospects used.
Agent prompt
You are building the objection library for this quarter's messaging.
Using Paperbase:
1. Pull the last {{days}} days of Interactions across all deals - calls, emails, Slack.
2. Cluster the objections and hesitations prospects raised, by theme. Quote the exact language used.
3. For each theme, list the deals affected, the stage where it appeared, and the outcome if known.
4. Pull Decisions about messaging or positioning, with Rationale, so the new library stays consistent.
Output:
- The 5-10 objection themes, ranked by how often they appeared.
- For each: the exact prospect language, quoted and dated.
- The counter-positions we have already taken, with sources.
- One recommendation: the theme most worth a dedicated response.Placeholders in {double_braces} are inputs the agent will ask for at runtime. Give the prompt to any agent connected to Paperbase, and the rest grounds in your own memory graph.