Why Even Your Best Employees Override AI Systems That Work
When employees retain personal accountability for outcomes, overriding a working AI is rational self-protection—not resistance. Genuine AI adoption requires realigning authority, accountability, and incentives so employees can trust AI decisions without bearing personal liability for outcomes they did not choose.
Summary
Narayandas argues firms deploy AI faster than they redesign governance, leaving employees accountable for outcomes even when following AI advice, so overriding it is rational self-protection, not resistance. He illustrates this with a bank whose accurate AI credit system was quietly undermined because authority shifted to the algorithm while accountability did not. His fix: realign roles and incentives with AI new authority.
Editor's opinion
If a human still owns the blame when AI is wrong, overriding it is rational self-protection, not a change-management failure. Fixing adoption means moving accountability to match AI authority.
Key concepts
AI governance · accountability structures · organizational AI resistance · human-AI collaboration