The 99% Accuracy That Brought the Boss Back to Data Scientists.

AI is powerful. Until it isn't. Then you need someone who actually understands the remaining 1%.

Author

Glenda Carnate
August 14, 2026

Read time: 2.5 minutes

“We've got AI; we don't need data scientists anymore." A few moments later: “Are you busy for a second?" The original irony continues.

The boss announced that data scientists are finished. Everything can be done by AI. Then the boss learned the model was working with 99% accuracy. However, that 1% was very costly for the business. So in the end, the company realized it needed an expert to fix it.

3 Ways to Employ AI in Data Science without Losing Expertise Value

1. Use AI to complement, not replace, data science expertise.
❌ Thinking that AI can do everything.
✅ Employing AI to quicken the process of model building but relying on experts for interpreting and validating.
AI can create models, but data scientists must interpret and clarify them.


2. Keep data scientists for that 1% AI cannot explain.
❌ Believing AI's 99% blindly.
✅ Employing data scientists to validate the 1% where context/patterns do matter.
AI sees the patterns; data scientists understand the sense.


3. Document the rationale of model decisions.
❌ Keeping expertise in people’s heads.
✅ Writing down the reasons behind models.
Experience can be taught, but only if shared.

💡Key Takeaway: 

Though AI can build models, it cannot adequately demonstrate the importance of the 1%. Thus, data scientists are still required.

👉 LIKE this if you find yourself in this situation.

👉 SUBSCRIBE now to this blog post for more insights.

👉 Follow Glenda Carnate to find out why AI does not replace everything.

👉 COMMENT below and tell us about your experience with AI failing.

👉 SHARE this with any data scientist who may need this.

Reply

or to participate.