Skip to content

How I build with AI

I design and build AI agents for workplace messaging platforms, including Slack, Microsoft Teams and Google Chat. They work across a team’s email, files and CRM. A person approves anything that goes out.

See the case study Contact

Agent loop.

Start with a clear outcome.

I learn how the work gets done, define the result that matters and agree on boundaries: what the agent handles and when a person steps in.

Agents I’ve built.

Team Workspace

One shared AI room per channel on a team’s messaging platform. The orchestrator plans each request and runs the specialists it needs.

  1. 01notes · follow-upsSales AgentProposes follow-ups from sales notes, each cited, and flags overdue ones.Explore
  2. 02plans · calls toolsOrchestratorPlans each step and runs up to three tools at a time, then answers with sources.Explore
  3. 03team chat historyChat Search AgentFinds what the team already said, even in threads.Explore
  4. 04accounts · dealsCRM AgentLooks up the account, the deal and its history.Explore
  5. 05files in · outDocuments AgentTurns shared files into new spreadsheets and documents.Explore
  6. 06inbox · draftsEmail AgentReads the thread and drafts a reply to approve.Explore
  7. 07public webWeb Research AgentSearches the public web and reads the pages it finds.Explore
  8. 08items · pricingQuote AgentEnters RFQ items into the quote and fills the workbook from the price library.Explore
See them work together

From workflow to agent.

TestTraceImprove
How I improve an agent