The governed enterprise brain
Meetings, sales calls, Slack threads, and pull requests become one governed record layer: cited, reviewed, and fresh enough for people and AI agents to act on.
what RecallMap does
Decisions happen in meetings, threads, and pull requests, then scatter. RecallMap watches those sources, catches the durable "from now on" moments, and routes them through review before people or agents depend on them.
Here is one ordinary week, as the brain sees it.
"We're cutting enterprise pilots from 45 days to 21. Effective immediately."
captured → pilot-duration proposed, routed to sales ops for review"From now on, any discount above 20% needs VP sign-off before it reaches the customer."
conflict check clear → approved by s.chen as discount-approval v3"v1 webhook endpoints are deprecated. All new integrations must use v2 signed events."
linked to integration-checklist · 2 dependent contracts flagged for review"Customer wants 25% off. Can I just send the quote?"
answered from discount-approval v3 · cites Tuesday's decision · routed to VP queuecategory wedge
Ask both the same question and the difference stops being philosophical. Retrieval hands you fragments and lets you guess. RecallMap hands you the current rule, with its owner, version, and the evidence behind it.
"Can we offer net-60 payment terms on enterprise deals?"
four fragments, three answers. You decide.
Yes. Net-60 is allowed on deals of $50k ARR or more, with CFO approval recorded on the opportunity.
knowledge that survives departures
Months of meetings, customer discussions, sales calls, and engineering decisions are already in the brain as owned, permissioned records. Offboarding becomes a transfer: the departing owner hands their restricted knowledge to whoever takes over, and the successor inherits context instead of an empty inbox.
Org-shared knowledge is never at risk in the first place; it belongs to the company, not the person.
chain of custody
A connector captures the exact Slack message below (text, author, permissions, timestamp) and pins it as a source version.
Decision review extracts the rule, refunds over $500 need finance approval, and checks it against every approved and pending record.
A named owner on the finance team edits or approves it. The old $250 threshold is retired deliberately, not silently overwritten.
v4 becomes memory, graph context, and support-refund.workflow: the exact version agents cite when they act.
Refunds over $500 require finance approval before support issues the refund.
slack / #ops-finance · message v3 · captured 09:42
notion / support-handbook · page v7
read: support, finance · admin: finance-ops
support-refund · v4 · called by 2 agents
current · both sources still at their captured versions
not a scraped wiki
This is one record. Around the sentence itself sit the source versions that prove it, the audience allowed to see it, and the contract it governs, so an answer can be traced, challenged, and re-published when the world changes.
Open the workspaceknowledge → governed contracts
When you put an agent on real work, it calls an approved contract instead of improvising. The contract stays thin: the rules it applies live as approved records in your knowledge base. Change one of those records, and every contract it governs is flagged for review, so an agent never acts on a rule your team has already moved past.
Discounts of 20% or less are auto-approved.
pricing-policy · v2currentDiscounts above 20% route to VP-sales for sign-off.
discount-approval · v3currentbuilt for governed agents
That distinction matters when an agent drafts a refund, updates a customer, or decides which process applies. Every guarantee below is enforced in the data model, not promised in a system prompt.
freshness audit
RecallMap pins the exact source versions behind every record. Sixteen days after the refund rule went live, someone edits the support handbook. The record and its contract are flagged within a minute, and agents keep running on the last approved version, visibly, until a human confirms the change.
common questions
These are the questions buyers, operators, and AI platform teams usually ask first.
An enterprise brain is a living knowledge layer built from a company's own work: messages, documents, code, meetings, sales calls, and customer conversations, organized so people and AI agents can use it. RecallMap adds governance, so the brain carries provenance, review, permissions, and freshness instead of just retrieved text.
Their meetings, customer discussions, sales calls, and engineering decisions are already ingested into RecallMap as owned records. During offboarding, restricted knowledge can be transferred to whoever takes over the role, so the handover becomes a permission change instead of a brain drain.
Knowledge carries an audience. Common operating knowledge is shared org-wide. Sensitive records stay restricted to an individual or group until someone with authority deliberately shares or transfers them. Retrieval respects those grants, so answers never flatten private and public knowledge.
Governed company knowledge is operating information that can be traced to a source, reviewed by an owner, checked for freshness, and safely used by people or AI agents.
Search tells teams what exists. RecallMap tells people and agents what the company currently approved. It stores source evidence, proposed and approved records, permissions, memory, graph context, freshness state, and versioned contracts for agents.
Yes. RecallMap is designed for company sources such as Slack, Notion, GitHub, meetings, and support systems, so decisions made in everyday work can become reviewed operating records.
RecallMap tracks source object versions and marks affected proposals, records, memories, relationships, and contracts stale or needing review when the underlying source changes.
RecallMap separates retrieved evidence from approved operating knowledge. Teams can require owner review before a proposed rule becomes a record or contract that agents use.
Connect a source, approve one decision, and watch raw conversation become knowledge your team and agents read together.