Agentic AI speeds curriculum drafting at General Assembly
Summary
General Assembly built a proprietary agentic AI system called GAIA (General Assembly Intelligence Application) on the CrewAI framework to accelerate curriculum development without replacing human experts. GAIA orchestrates four specialised agents—instructional architect, QA architect, subject matter expert and learning experience designer—each trained on the company’s content, frameworks and standards. The system produces strong first drafts in minutes, enabling humans to focus on judgement, client context and quality.
Under CTO Danielle Chircop’s direction, the company reports about a 90% reduction in first-draft development time, greater capacity to run multiple programmes in parallel, and preservation of a human-in-the-loop review process. They chose a proprietary approach for enterprise-grade security, white-glove service and to handle structured, practical learning assets like labs and walkthroughs. Implementation required more engineering effort, tighter evaluation criteria and change management than initially expected.
Key Points
- GAIA is a multi-agent system (built on CrewAI) that mirrors four real curriculum roles to generate review-ready first drafts.
- The tool reduced first-draft development time by roughly 90% versus previous estimates of ~33%.
- Humans remain central: learning experience designers and SMEs perform final judgement, refinements and client-specific tailoring.
- General Assembly opted for a proprietary system to meet enterprise security and quality standards rather than using generic tools like ChatGPT.
- Implementation required unexpected investments: developer/engineering support, rigorous evaluation standards and a focused ‘tiger team’.
- Organisational benefits included higher throughput, consistency across programme portfolios and internal promotions as staff handled larger workloads.
Why should I read this?
Fancy slashing your curriculum drafting time and still keeping humans firmly in charge? This piece shows a real-world playbook: agentic AIs that spit out quality first drafts, humans who add the nuance, and the org-level fixes you actually need (training, a tiger team, and more engineering than you think). If you work in L&D, training ops or enterprise learning, this is a neat shortcut to see what’s possible—and what to watch out for.
Context and relevance
This story matters because it illustrates a maturing use of agentic AI in enterprise content production: orchestration of specialist agents, preservation of human oversight, and a focus on security and bespoke output for premium clients. It aligns with broader trends—human-in-the-loop deployments, bespoke enterprise AI stacks, and the rise of agent frameworks for complex workflows—making it highly relevant for organisations scaling learning, knowledge work and productised services.
Author style
Punchy: the article is direct about benefits and trade-offs. If you’re evaluating agentic AI for scaling knowledge work, the interview with GA’s CTO gives sharp, practical guidance rather than hype—worth reading in full if you’re planning a pilot.