Some questions can't be answered with a single query. They require pulling data from multiple sources, running analyses in parallel, synthesizing conflicting signals, and presenting a unified answer. Sub-agents make this possible.
How it works
When Adapt receives a complex request, it decomposes it into discrete sub-tasks and delegates each to a specialized agent. These agents work in parallel, and their results are synthesized into a single coherent response.
You ask one question. Adapt orchestrates everything behind the scenes.
Example:
Give me a full picture of why churn spiked in March.
Pull usage data, support tickets, and any relevant Slack discussions.
Adapt might spawn three sub-agents simultaneously:
- One querying your analytics platform for usage drop-off patterns
- One searching support tickets for themes during that period
- One scanning Slack for customer mentions and internal discussions
Results come back together, cross-referenced and synthesized.
Why this matters
Previously, multi-source analyses required you to run several separate queries and manually connect the dots. Sub-agents collapse that into a single interaction. You get a complete answer without having to orchestrate the research yourself.
This is particularly useful for:
- Root cause investigations across product, support, and usage data
- Competitive research pulling from web, CRM, and internal notes simultaneously
- Executive briefings that need context from engineering, finance, and sales data in one place
- Incident retrospectives correlating events across multiple systems
Transparency
Adapt shows you what each sub-agent was tasked with and what it found. You can expand any sub-agent's work to see the raw query, the sources it used, and its intermediate conclusions before the final synthesis. Nothing is hidden.
Limits
Sub-agent tasks run within your existing integration permissions. An agent won't access a data source you haven't connected, and it won't take actions you haven't enabled. The same authorization model applies whether the request comes directly from you or from an orchestrating agent.
