Ep159: Why Agentic AI Projects Fail (and How To Avoid It)
Update: 2025-10-16
Description
Industry leaders from Coder, Scale AI, and Suger reveal why 95% of AI pilots fail—and share the frameworks that actually work to get agents into production.
Topics Include:
- Panel features leaders from Coder, Scale AI, and Suger discussing agentic AI.
- MIT report reveals 95% of AI pilots fail to reach production.
- Challenges are rarely technical—they're organizational, mindset, and people-driven instead.
- Companies lack documented tribal knowledge needed to train agents effectively.
- Many organizations attempt AI where deterministic, rules-based automation would work better.
- "Freestyle agents" concept: Some problems shouldn't be solved by agents at all.
- Regulated industries struggle when asking agents to handle highly differentiated, complex tasks.
- Common mistakes: building one universal agent or separate agents for every use case.
- Post-billing workflows and business-critical operations aren't ready for AI's black box.
- VCs pressure companies to define "AI-native"—but nobody has clear answers yet.
- Scale AI uses five maturity levels; Coder uses three tiers for adoption.
- Success metrics span operational readiness, business impact, and technology performance indicators.
- Production requires data governance, context, A/B testing, and robust fallback mechanisms.
- Even Anthropic uses agents conservatively: research tasks and log triage, no write-access.
- Path to 50% success requires agile frameworks, people change, and proper AI talent.
Participants:
- Ben Potter - VP of Product, Coder
- Raviteja Yelamanchili - Head of Solutions Engineering, Scale AI
- Jon Yoo - CEO, Suger
- Adam Ross - US, Partner Sales Sr. Leader, Amazon Web Services
See how Amazon Web Services gives you the freedom to migrate, innovate, and scale your software company at https://aws.amazon.com/isv/
Comments
In Channel
Download from Google Play
Download from App Store
United States00:00
00:00
x
