Skip to main content
Memory is Superhawk’s durable record of what has been true about your accounts. Every call, email, ticket, and usage change it processes leaves something behind, and that accumulated record is what lets an answer cite a specific sentence from a call eight months ago. It is the difference between a tool that summarises this week and one that remembers the relationship.

The three layers

Memory is organised in three layers, each answering a different kind of question.

Facts

Specific things that are true about an account and the people at it. Who the champion is, which team owns the workflow, what they use a feature for, which system they need it to integrate with. Facts are extracted from what was actually said or observed, and they carry their source.

Episodes

Individual observations, in time order. A call that happened, a ticket that was raised, an email thread, a usage change. Episodes are the raw record that facts and patterns are derived from.

Patterns

Derived aggregates — what holds true across many episodes rather than in any one of them. Recurring themes across calls, behaviour that repeats across quarters, tendencies that only appear when you look at the whole history.

Searching memory

The Memory page lets you search across all three layers, scoped to your whole workspace or narrowed to one account. Useful questions to ask it:
  • What has this customer said about a particular feature, over time?
  • Who at this account has ever been described as a decision maker?
  • What has come up more than once and never been resolved?
  • What did we say we would do, on any call, ever?

What memory is used for

Memory is not only a page you visit. It feeds:
  • Account health — historical context behind a band
  • Signals — recognising that something is a change rather than normal
  • Briefs — what changed since last time, and what is still open
  • Chat — most of what makes an answer specific rather than generic

Tenancy

Memory is scoped to a workspace, and within a workspace to an account. One workspace’s memory is never visible to another, and it is never mixed into another workspace’s answers.
  • Chat — the easiest way to query memory
  • Briefs — memory applied to a specific meeting