Multi-step workflows with agents

What is the challenge?

• Elle can complete single-step tasks today: draft a description, build a report, analyze a set of records.

• But real product work involves multi-step workflows where the output of one step feeds the next.

• Writing a competitive brief means researching competitors, then comparing their positioning to yours, then drafting the brief.

• Elle cannot chain these steps together in a reliable way.

What is the impact?

• Product managers break complex work into steps manually, prompting Elle repeatedly and copying context between conversations. The value of AI drops significantly when you have to be the orchestrator for every workflow that involves more than one action.

Describe your idea.

• Elle can execute multi-step workflows where each step builds on the previous one. You describe the end goal and Elle determines the steps, executes them in sequence, and produces the final output. A workflow can span internal data (records, reports, notes) and external sources (web research, uploaded documents) in a single run. For example, "prepare me for my customer call with Acme" could pull the account's linked ideas, check recent feature status changes, and draft talking points in one pass.

This supports the semi-autonomous workflow defined below: