Chapter 25: Common Pitfalls and How to Avoid Them

Organisations that struggle with agentic AI often make the same mistakes. Learn from their experience.

Pitfall 1: Starting Too Big

The mistake: Attempting a complex, high-stakes, enterprise-wide transformation as your first agentic AI project.

The fix: Start small. Pick a bounded use case. Prove value. Then expand.

Pitfall 2: Underinvesting in DNA

The mistake: Assuming a powerful model will figure it out. Rushing past instruction and knowledge design.

The fix: Treat DNA design as a first-class discipline. Invest time in crafting instructions and curating knowledge.

Pitfall 3: Skipping Governance

The mistake: Moving fast without establishing accountability, oversight, or guardrails.

The fix: Establish basic governance before agents go live. It doesn't have to be heavy — but it has to exist.

Pitfall 4: Ignoring the Human Element

The mistake: Focusing on technology while neglecting change management, training, and communication.

The fix: Bring people along. Explain what the agents do and don't do. Train staff to work alongside agents.

Pitfall 5: Set and Forget

The mistake: Treating agents as finished products rather than living systems.

The fix: Plan for ongoing operations. Budget for maintenance. Build feedback loops. Continuously improve.

Pitfall 6: Chasing Autonomy for Its Own Sake

The mistake: Pushing for maximum autonomy regardless of whether it's appropriate.

The fix: Match autonomy to context. Use the maturity model to find the right level, not the highest level.