back

Don’t Blame AI, That’s Just Mean

A finance team blames an agentic tool for a 15% miss on cloud spend projections. The tool isn’t at fault. A human is.

AI executes instructions. Humans write them, feed the data, set limits, set how creative it should be, and decide what to accept. When outputs go wrong, responsibility doesn’t live in the model. It lives with us.

This hits hard in scaling companies, where AI is driving tech, business decisions, and more every day.

AI Lacks Independent Agency

AI follows human-set rules. No intent. No judgment. Regulators agree: humans own the outcomes. Blaming AI dodges the real choices: data quality, boundaries, tuning, grounding, oversight.

Shared vs. Diffused Responsibility

Cross-functional AI spans finance, tech, and ops. Shared work, blurred ownership. Finance assumes tech validated the integrations. Tech assumes finance set the rules. Ops assumes the reviews happened.

Result: an agent reallocates cloud spend wrong. Everyone says “AI glitch.” No one says “our setup failed.” In scaling firms, this repeats: errors linger, fixes slow.

Map it out: who owns data, thresholds, reviews, and post-incident tweaks. Diffusion isn’t inevitable. It’s fixed with a clear chain of responsibility.

An Example From Practice

In February 2025, Advance Local Media LLC and other publishers sued in the Southern District of New York (Case No. 1:25-cv-01305-CM). They alleged that Cohere’s AI model generated hallucinated news articles falsely attributed to their trademarked brands without proper output controls to prevent misattribution.

The AI didn’t decide to misattribute content. The issue is training data and output controls, or the lack of them, that allowed the problem to occur. Tools don’t err alone. Oversight does.

Why Blame Shifts

AI feels separate. Studies show people assign less blame for automated harm. In growth mode, AI speeds planning and cloud ops. Issues hit? “System fault.” A comfortable excuse that skips examining the choices. Blaming the tool is easy. Owning the process isn’t.

Maintaining Accountability

Steps for AI in workflows:

  • Assign one owner per high-stakes output.
  • Verify suggestions against sources.
  • Log decisions and escalations.
  • Set hard autonomous boundaries.
  • Involve finance, tech, and compliance from rollout.

These build stable growth. They turn AI into reliable support.

Conclusion

AI fails when we let it. Trace problems to human decisions: design, data, deployment, tuning.

For scaling leaders, clear accountability, especially cross-functionally, cuts risk, earns trust, and maximizes AI value. Map responsibility. Own the outcomes. Get AI working the way it should to support the business.