‹ Marketing recipesMarketing & Positioning · Agent Blueprint
Positioning, web copy, ads
Language our best customers actually use. Phrases your customers used to describe your product: ranked, dated, sourced.
Why this is memory-only. The words your customers use about you are scattered across every call, support thread, and review - and they are never written down as phrases customers use about us. Only a memory layer that keeps those conversations can mine the language on demand.
Side by side: token usage, with and without Paperbase
Memory onlyWithout Paperbase
Not recoverable from ordinary content search
- Interview the sales team for phrases customers use.
- Skim call notes for memorable quotes.
- Guess at the words that resonate.
- Write copy in your own language, not the customer's.
- Measure later why it did not land.
With Paperbase
Persistent memory layer
- Ask Paperbase for the language customers use about the product, extracted from real interactions.
- Rank phrases by frequency and recency.
- Pull the quote, date, and context behind each.
- Compose copy from the market's own words.
Agent prompt
You are writing positioning and copy for {{product_or_offer}}.
Using Paperbase:
1. Pull the language customers actually use about {{product_or_offer}} from interactions - calls, support, reviews, emails.
2. Rank phrases by how often they recur and how recent they are.
3. For each phrase, return a quote, its date, and the account.
4. Distinguish the phrases buyers use before purchase from those used after.
Output:
- The ten most recurring customer phrases, ranked.
- For each: the quote, source, and whether it is pre-purchase or post-purchase language.
- Two positioning angles grounded in the customer's own words.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.