The Context Graph is the connected customer context Bigmind can use for a Deal or Account. It links CRM records and their relationships to conversations, communications, notes, signals, workspace configuration, and the artifacts Bigmind maintains.
What is available depends on the systems connected to your workspace, the data imported from them, and the associations that connect each record. Bigmind does not assume that unrelated activity belongs to a Deal or Account.
Core concepts
| Concept | Meaning |
|---|---|
| Record | An imported CRM object such as a Deal, Account, Contact, or Lead and its current fields. |
| Relationship | An association or attribution that connects records, people, and interactions. |
| Evidence | Customer information that can support a finding, such as a transcript, email, CRM activity, or linked note. |
| Workspace context | Your Frameworks, Playbooks, meeting configuration, and Library knowledge. |
| Scope | The Deal, Account, or meeting whose context is being assembled. |
| Artifact | A maintained result for a specific record, such as a Framework assessment, Deal summary, or Deal warning. |
Records and relationships
The graph starts with CRM objects and the links between them.
- Record state: current Deal, Account, Contact, Lead, and other imported CRM fields. For a Deal this can include owner, stage, amount, close date, pipeline, lifecycle flags, and custom fields. See the CRM integration overview.
- Associations: the CRM relationships connecting a Deal to its Account and people, or an Account to its Deals, contacts, activities, and notes.
- Identity and attribution: the participants, senders, recipients, organizers, and owners used to connect customer interactions to the right records. See Activity attribution.
CRM fields describe the current state of a record. Associations and attribution determine which other records and interactions belong in its context. They are separate from the activity itself.
Customer evidence
Once records are connected, Bigmind can use evidence from the systems available in your workspace:
- Meetings and calls: participants, timing, recordings or transcripts when available, summaries, action items, and meeting-level analysis such as Scorecards and Trackers.
- Email and calendar: attributed inbound and outbound messages, threads, invitations, and scheduled or completed events from connected communication systems.
- CRM timeline records: imported calls, tasks, events, and other activity records that form part of the CRM history.
- Linked notes and documents: CRM notes and other records explicitly associated with the Deal, Account, or related people.
- Signals: relevant Account events and the context supplied by configured signal sources.
- User-provided context: corrections, edits, and information users intentionally add through Bigmind.
A source can only contribute what its connection and permissions make available. A meeting without a transcript can still provide metadata, but it cannot support conclusions that require conversation content.
Workspace context
Workspace configuration tells Bigmind how to interpret evidence for your business:
- Frameworks are customer-configured assessments that define the Deal or Account components to track, their applicability, and any scoring rules. A workspace can start from a template or build its own MEDDICC, Deal Score, Account Health, or another assessment through the UI or MCP.
- Playbooks provide guidance for a GTM motion and can reference relevant Frameworks and Library resources.
- Meeting configuration defines Scorecards, Talking Points, Trackers, categories, and other meeting-level analysis.
- Library knowledge provides approved company, product, customer, and process context when it is relevant to the work being performed.
Deal context and Account context
| Scope | Context |
|---|---|
| Deal | The Deal record, its associated Account and people, Deal-linked records, and customer interactions attributable to that sales motion. Relevant Account context can inform the review without replacing Deal-specific evidence. |
| Account | The Account record and the broader customer relationship across associated people, Deals, meetings, communications, notes, and signals. |
| Meeting | One interaction and its participants, transcript or recording, meeting analysis, and associated CRM records. |
From evidence to artifacts
Evidence is the supporting customer information. Artifacts are the maintained results users review in Bigmind, such as Framework findings, Deal summaries, and Deal warnings. Existing artifacts are considered during a later review so Bigmind can preserve, update, or resolve prior findings instead of starting from an empty state.
See Deal and Account artifacts for the artifact model and Frameworks for configuration.
Next, see How it works for a concrete example and refresh behavior.
How changes become updates
Connecting a source makes information available. It does not mean every synced record starts an AI review. Automatic artifact maintenance follows this flow:
- A change arrives. A selected CRM field changes, an attributed email or activity arrives, a linked meeting finishes processing, or a relevant account signal is dispatched.
- Bigmind finds the affected account or deal. CRM associations and activity attribution determine where the information belongs. Unrelated messages do not become customer evidence just because an integration is connected.
- Tracking and triggers are checked. The account must qualify for tracking and fit its allowance. The change must also be a relevant trigger. Deals inherit tracking from their account.
- Changes are grouped. The first relevant change starts a refresh window for the affected target. Further changes can join that pending review instead of each starting a separate AI run.
- Artifacts are reviewed. Bigmind uses current evidence and existing results to decide what to create, update, preserve, or resolve.
For example, a meeting summary can supply new evidence for an associated deal summary, warning, or Framework assessment. The meeting summary remains a meeting artifact; the later review maintains the deal's own artifacts. Exact outputs are decided during processing, so a queued change does not promise that every artifact will change.
Sources, Signals, and Orchestration
- Sources manages connections and the data imported from them.
- Signals manages account events and their independently configured detection, notifications, and automations.
- Orchestration controls which accounts receive automatic artifact maintenance, which changes trigger it, and when grouped reviews become eligible to run.
Open Context graph → Orchestration. Its tabs are Settings, Changes, and Outputs. Tracking controls graph refreshes; it does not switch off integration sync or independent signal automations.
Every Orchestration setting
Automatic artifact updates
Enable automatic artifact updates is the master switch for automatic graph artifact maintenance. Turn it on and save to allow relevant changes to queue reviews. Turn it off and save to stop automatic updates. Existing artifacts are preserved, and account tracking membership is separate from this switch.
The enabled setting displays your organization's refresh window. The default is six hours; some organizations use a longer window, such as 24 hours. It is configured for your organization and read-only here. A window starts with a relevant change, not at a fixed time every day. Its end is the earliest refresh time, not a guarantee that processing finishes then.
Existing workspaces: before Orchestration settings have been saved for the first time, previous automatic refresh behavior may remain active. An unsaved switch or zero tracked count is not proof that updates are paused. To explicitly stop automatic maintenance, switch it off and save.
Account tracking
Tracked accounts and Untracked accounts show the current stored tracking statuses in your CRM connection. Click either card to browse those accounts. The Filter page action shows or hides the filter bar. Use the tracking filter to switch between tracked and untracked accounts, and account filters to narrow the list. Hiding the bar keeps filters applied. View activity opens an account's recorded graph activity.
When your plan includes a company allowance, it appears below the tracked count. Matching a rule and being admitted within the allowance are separate steps. An Out of bounds card appears when rule matches exceed the allowance.
Tracking rules
Use Add tracking rule to define which accounts qualify. Use Add account filter inside a rule to select CRM account fields, including available owner, status, and custom fields. Choose a comparison and value supported by that field. Remove rule removes it from the draft settings.
- Within a rule: match all conditions (AND) or any condition (OR), as selected in the editor. Groups have their own operator; the rule's outer operator combines groups and other conditions.
- Must have a deal matching these filters: enable this option, then use Add deal filter to select fields from your CRM’s deal schema. At least one linked deal must satisfy the complete deal filter expression, in addition to the account filters. With no deal filters, the account must have at least one deal.
- Between rules: matching any one rule is enough.
- No rules: all accounts qualify, subject to exclusions and the allowance. A rule with no field conditions and no matching-deal requirement also includes all accounts. Removing every rule does not pause updates; use the master switch for that.
For example, use an account filter for active customers and a deal filter for Is Closed = false. Add a separate rule with Is Won = true AND Close Date is within the last 365 days to also include recently won customers. Field names and options come from your CRM. Rolling date windows are re-evaluated over time; existing open-deal rules retain their behavior.
Preview changes, below the rules when settings are unsaved, evaluates the draft before saving. It shows matching accounts, accounts that would be tracked after the allowance, new matches, accounts no longer matching, and any excess. Account exceptions are included. The preview does not save settings or start reviews; the cards above still show current stored tracking statuses.
Account exceptions
Always track includes selected accounts even if no rule matches, including when they have no open deal. Never track excludes selected accounts and takes precedence over both rules and Always track.
Exceptions determine account eligibility independently of integration trigger settings. They do not bypass the organization allowance or master switch, or turn every change to an included account into a refresh trigger.
Refresh triggers
This section shows connected integrations that produce graph triggers. Click an integration to open its settings drawer:
- CRM: choose Account fields and Deal fields to watch. Relevant changes to selected fields trigger CRM-driven reviews. Restore standard fields restores the default selection. Trigger fields are separate from rule fields: rules choose accounts; trigger fields choose which CRM changes prompt a review.
- Email and calendar: attributed messages and events can contribute updates. The drawer shows relevant streams and data; their graph triggers are read-only here.
- Slack and other messaging: synced activity must be associated with an account or deal. For example, a Slack channel linked to an account can contribute message activity. Trigger information is read-only here.
- Meetings and calls: processing a captured, associated session can trigger a review. Importing a recording alone is not enough.
- Signal sources: dispatched account signals can contribute to graph reviews. Their independent detection and automation behavior remains configured under Signals.
Not every connected service is a direct trigger. Enrichment results or indexed website documents do not independently trigger account reviews merely because they were imported. A resulting change to a watched CRM field can follow the normal CRM trigger path.
Save settings
Click Save settings to apply the switch, rules, exceptions, and selected CRM fields. Bigmind evaluates existing accounts in background batches and updates their tracking statuses. Large CRMs take time to scan, so counts need not change immediately. Newly included accounts can receive an initial review; these reviews are scheduled gradually rather than all starting at once.
Tracked accounts and allowances
The decision order is Never track → Always track → tracking rules → account allowance. Deals use their account's tracking eligibility; their lifecycle and applicable artifact settings still determine which work runs.
| Status | Meaning |
|---|---|
| Tracked | The account qualifies and has been admitted for tracking. |
| Untracked | The account is not currently marked for tracking. During setup, this can include accounts the background scan has not evaluated yet. |
| Tracked · Over allowance | The account is tracked above the allowance because an organization-specific exception permits it. |
| Untracked · Over allowance | The account qualifies but exceeds the allowance. This is the default for excess accounts. |
Over-allowance behavior is configured for the organization, not in these settings.
Understanding Changes and Outputs
Changes shows incoming activity over time, stacked by Queued, Processing, Completed, Skipped, and Failed. One source event can affect both an account and a deal, so the chart is not necessarily a count of unique emails, meetings, or CRM edits. Choose 24 hours, seven days, or 30 days; click a bar to inspect its time bucket. The refresh marker shows when the next queued review becomes eligible.
Outputs shows recorded artifact changes, such as warnings created or resolved and deal summaries updated. A completed review can produce no changes. Conversely, a review can save an output and fail later; that saved output still appears. Meeting processing and other audited executions can also produce outputs, so not every output represents a scheduled account refresh.
The activity list groups entries by account and day where the association is available. Click a group to open a drawer answering What happened, What will be updated (or What was updated), and Status. Multiple reviews can be browsed separately. Before processing, exact outputs may not yet be known. Recorded outputs include counts and links where available.
History covers up to 30 days. It is an operational history, not a permanent record of every source-field value or artifact version.