AI Crew Recap: Build Agentic AI Workflows that Actually Scale
We kicked off June with a fantastic AI Crew session featuring
The AGENT Framework
The heart of Jen's presentation was a five-step framework she uses with organizations to operationalize AI at scale. The acronym AGENT breaks down like this:
- A — Assess: Identify high-value workflows worth improving, not AI for AI's sake
- G — Ground: Give the AI enough context to make good decisions (garbage in, garbage out)
- E — Execute: Orchestrate multiple agents working together, each with a clear role (intake, research, policy, action, QA) rather than isolated one-shot prompts
- N — Navigate: Keep humans in the loop — human oversight is a feature, not a limitation
- T — Tune: Continuously measure, improve, and scale; building AI into a workflow is a starting point, not a finish line
Live Demo: Support Ticket Triage
Jen also walked through a live demo of a support triage system built in Airtable with OpenAI, showing how a customer submitting a three-field form kicks off a full agentic workflow: automatically matching the submitter to their account, pulling in CRM context, analyzing sentiment and priority, generating a recommended action plan, drafting a customer response email, and even creating a linked feature request, all before a human ever touches it. The human's job at the end? Review, make any tweaks, and hit approve. Clean, fast, and actually useful.
Common Pitfalls
Jen also walked through the mistakes that break agentic systems: automating a process that's already broken, skipping human escalation paths, context-starving your agents, and over-engineering before you've proven value. Her advice: start narrow, measure aggressively, and expand intentionally.
Watch the Recording
Missed it live or want to rewatch the demo? The full recording is here.
We hope to see you at the next one!
