ConceptsGuidesAPI ReferenceChangelog
Automation

Automation

How GTM Engine workflows turn CRM activity, external data, AI analysis, and integrations into structured revenue intelligence.

GTM Engine automation is built for revenue teams that need more than "if this, then that." It is a first-class CRM data platform for processing internal activity, external research, enrichment, and AI analysis into information your team can query, trust, and reuse.

Automations can write structured data back to CRM fields, send unstructured messages to users, expose reusable workflow tools to agents, and feed downstream workflows that make the next decision.

The Big Idea

GTM Engine as a CRM data platform

Input evidence

Signals arrive from every GTM motion

CallsEmailsCRM updatesResearch

Activity, fields, transcripts, messages, and external context become record-linked evidence.

->

GTM Engine

The CRM record graph stays current

GTM Engine attaches evidence to accounts, contacts, opportunities, and activities so AI and workflows can update the data layer with context.

1

Match evidence to records

2

Evaluate with AI + workflows

3

Keep structured fields current

Structured fields

Durable, filterable CRM data.

Attached evidence

Source context for agents and workflows.

->

Outputs

Reusable data and actions

CRM fields

Deal healthForecastClose riskNext step

Actions and integrations

Agent answersSlack alertsWorkflow actionsCRM sync
GTM Engine connects raw go-to-market activity to CRM records, keeps both the source evidence and structured fields, then uses that context across workflows, agents, and connected tools.

1

Sources

CRM records, field changes, emails, calls, meetings, imports, enrichment, web research, and user actions.

2

Workflow graph

Triggers map source data into workflow inputs. Steps run tasks that fetch, analyze, enrich, branch, loop, and write.

3

CRM intelligence

Structured fields are updated: predicted close date, deal health, next steps, gaps, scoring, and custom fields.

4

Reuse

People, agents, reports, forecasts, and other workflows can query or act on the updated data.

Automation turns raw GTM signals into durable CRM intelligence and actionable messages.

What the Automation Platform Can Do

CRM data

Process internal CRM signals

Workflows can react to record creation, record updates, stage movement, owner changes, imports, and manual runs from record pages. Trigger outputs are mapped into workflow inputs so every run has an explicit data contract.

Sources

Use external and conversational data

Tasks can use enrichment vendors, web research, scraped pages, call transcripts, emails, meetings, and imported prospect lists as workflow context.

Durable data

Write structured fields

Use Update Record, AI Prompt with structured output, enrichment, and field-task workflows to write values that reports, forecasts, reps, managers, and agents can query later.

Human action

Send unstructured messages

Use Slack Message, Send Email, Instantly Sequence, and other side-effect tasks when the result should notify a person, start outreach, or publish content.

The Core Building Blocks

Triggers

Triggers start workflows. A trigger can be an event, schedule, webhook, or system action. Triggers are standalone resources, not hidden workflow settings, so one trigger can call many workflows and one workflow can be called by many triggers.

Examples:

  • A Record Updated trigger fires when an opportunity stage changes.
  • A scheduled trigger runs weekly account scoring.
  • A webhook starts a workflow from an external system.
  • A manual run starts from a record page or batch action.

Trigger outputs are mapped explicitly into workflow inputs. The old implicit {{ trigger.* }} model is gone; the mapping is visible and validated.

Inputs and Outputs

Every workflow has an input/output contract.

  • Inputs are values the caller provides, such as account_id, contact_id, opportunity_id, campaign_id, or search_criteria.
  • Outputs are values the workflow returns to its caller, such as research_summary, health_score, next_best_action, or created_record_id.

This contract is what makes workflows reusable. The same workflow can be called by a trigger, a record action, another workflow, or an agent tool because each caller maps its own data into the same declared inputs.

Steps and Tasks

A workflow is a graph of steps, and each step contains one or more tasks.

  • Steps run in order.
  • Tasks inside the same step run in parallel.
  • Downstream steps can reference outputs from prior steps.
  • Tasks in the same step should not depend on each other because they run at the same time.

Use this pattern when designing workflows:

  1. Put independent fetches or research tasks in the same step so they can run in parallel.
  2. Put tasks that depend on those results in the next step.
  3. Use For Each when an array needs item-by-item processing.
  4. Use Nested Workflow when a process should be reusable.

1

Step 1: Load context

Get Record, Search, and Research tasks can run independently.

2

Step 2: Analyze

AI Prompt or Research Signal reads Step 1 outputs and produces structured findings.

3

Step 3: Act

Update Record, Slack Message, Send Email, or Instantly tasks write data or notify people.

Steps provide order. Tasks inside a step provide parallel work.

Variables and Liquid

Variables move data through the graph. Task fields can reference:

NamespaceWhat it meansExample
inputs.*Values passed into the workflow{{ inputs.account_id }}
steps.<N>.*Outputs from prior steps{{ steps.1.account.name }}
outputs.*Values returned by the workflow{{ outputs.summary }}

Liquid is the template language used in prompts, messages, mappings, and integration payloads:

Research {{ steps.1.account.name }} ({{ steps.1.account.domain }}) and summarize the top 3 buying signals.

Use Liquid for field insertion and light formatting. Use workflow tasks for real branching, side effects, and durable business logic.

Task Reference

The Task reference documents every supported workflow task. Each task page explains:

  • What the task is for.
  • Side effects and credit behavior.
  • When to use it.
  • Setup tips.
  • Practical examples.
  • Outputs and common pitfalls.
  • Exact required fields, variable-aware fields, Liquid-rendered fields, and default configuration.

The task reference is generated from the workflow task registry on every docs build, so examples and field details stay aligned with what is available in the builder.

Structured Data vs. Messages

Automation usually produces one of two kinds of output.

Structured CRM Fields

Structured fields are the backbone of GTM Engine automation. When a workflow writes to a field, that value becomes available to:

  • Record pages and lists.
  • Reports and dashboards.
  • Forecasting.
  • Agent tools.
  • Other workflows.
  • CRM sync and field mappings.

Examples:

  • predicted_close_date
  • deal_health
  • best_next_step
  • forecast_category
  • meddic_score
  • competitor_summary
  • account_angle

Structured data is best when the result should be queryable, filterable, reportable, or reusable later.

Unstructured Messages

Messages are best when the result should prompt a person to act now.

Examples:

  • A Slack alert for a high-risk deal.
  • An email summary before a customer meeting.
  • A generated outbound personalization note.
  • A Sanity document draft or published content.

Messages can include workflow variables and AI-generated text, but they are less reusable than structured fields. If the information should power future automation, write it to a field too.

Example: Forecasting as Automation

Forecasting is a good example of why GTM Engine treats automation as a CRM data platform.

1

Configured signal

An activity-created or record-updated trigger fires for configured categories such as email, meeting, transcript, or CRM field changes.

2

Field workflows

Workflows extract facts, update structured fields, and keep record context current.

3

Forecast analysis

AI and analysis tasks use schemas and validation to update predicted close date, deal health, gaps, next steps, and scoring.

4

Action layer

Managers, reps, reports, Genie, and follow-up workflows can all query or act on the updated intelligence.

Forecast intelligence is built by processing configured activity and record-change signals into structured fields, then analyzing those fields.

For example, after a configured transcript trigger fires, GTM Engine can:

  1. Read the related account, opportunity, contacts, and transcript.
  2. Extract structured signals from the conversation with an AI Prompt or analysis task.
  3. Update CRM fields such as deal health, close-date confidence, next step, gaps, and MEDDIC/MEDDPIC evidence.
  4. Run follow-on analysis using those updated fields.
  5. Surface the result in Forecast, record pages, reports, Genie, or another workflow.

The key is that analysis does not live only in a chat answer. It becomes CRM data your team and agents can use later.

Design Patterns

Activity driven

Triggered field intelligence

When a configured call, email, meeting, or transcript trigger fires, update structured fields that explain deal health, gaps, and next steps.

Many records

Batch data operations

Run the same workflow across a filtered record set, track progress in Batch History, and rerun failures.

Workflow tool

Agent as interface

Expose a workflow as a tool so an agent can choose when to run a controlled operation.

Outside data

External enrichment

Pull in enrichment, research, or scraped content, then normalize it into CRM fields or messages.

Action now

Human notification

Send Slack or email when a workflow finds something that needs immediate attention.

Nested workflow

Reusable subflows

Keep common operations like enriching one contact or generating one account summary as reusable workflows.

Where to Go Next

Publish content to Sanity from a workflow

Generate a Sanity document in a workflow, validate it against a schema, and publish it safely.

Task reference

User-facing guidance and generated reference details for every workflow task type.

On this page

What the Automation Platform Can DoThe Core Building BlocksTriggersInputs and OutputsSteps and TasksVariables and LiquidTask ReferenceStructured Data vs. MessagesStructured CRM FieldsUnstructured MessagesExample: Forecasting as AutomationDesign PatternsWhere to Go Next