Key capabilities
- Multi-step pipeline builder with LLM generation, tool calls, conditionals, and branching
- Batch execution across accounts
- Context recipes for reusable research configurations
- Frameworks for reusable workflow templates
- Workflow history and tracking
- Model selection per step - choose the right LLM for each task
Who it’s for
| Role | How they use it |
|---|---|
| RevOps teams & sales operations | Build workflows once and run them at scale across many accounts |
| Account executives | Trigger pre-built workflows for specific tasks (e.g., “generate a meeting brief”) |
| Sales leaders | Ensure consistent output quality across the team |
When to use workflows vs. the AI Assistant
| Task | Use Assistant | Use Workflow |
|---|---|---|
| One-off research question | Yes | No |
| Meeting brief for one account | Either works | Yes, for repeatable process |
| Research report for 50 accounts | No | Yes |
| Exploring an account interactively | Yes | No |
| Standardised competitive analysis | No | Yes |
| Territory-wide change summary | No | Yes |
Rule of thumb: If you’ll do the same task more than 3 times, build a workflow. If it’s a one-off exploration, use the assistant.
Effective workflow patterns
Meeting preparation pipeline
Output: A meeting-ready document per account. Run in batch for a week’s worth of meetings.
Territory health review
Check for changes
For each account in segment: check for Intelligence changes since last review (tool call)
Account research and engagement plan
Output: A complete engagement package per account.
Tips for effective workflows
Use context recipes
Don’t reinvent how Intelligence data is gathered for each workflow. Recipes standardise this.
Test on a small sample first
Before running a workflow across 200 accounts, validate the output on 3–5 accounts.
Review output quality
Workflows produce consistent output, but consistency doesn’t guarantee quality. Review and iterate.
Monitoring and troubleshooting
- Workflow history shows the status of every run: completed, failed, or in progress.
- Step-level detail shows which step failed and why.
- Retry is available for failed runs.
- Output inspection shows the generated content at each step.
