Science & Tech · AI · 10 hrs ago
Why Bank Staff Still Overruled an Approved AI Credit System
A global bank approved an AI system for routine credit decisions in late 2022.
In pilot tests, it handled applications faster than people, made fewer errors and was expected to cut costs.
Six months later, most decisions still received human review, despite the system continuing to assess applications and make recommendations.
Staff added more review steps and manual overrides, so the AI rarely made the final call.
The explanation is that workplace accountability rules made relying on AI professionally risky when its decisions differed from a person’s.
The same pattern has appeared with earlier tools such as enterprise planning and credit-scoring systems, which could make decisions but did not gain authority.
The concern is that companies may adopt AI without changing who is trusted to decide, limiting its impact even when the technology works.
A global bank approved an AI system for routine credit decisions in late 2022, but staff continued to review most cases.
The system processed applications faster than people, reduced errors and promised cost savings in pilot projects.
Six months after approval, human reviews and exceptions remained common, and temporary safeguards had become permanent steps.
The article argues that accountability structures and rules about who can make trusted decisions prevented the system from gaining real authority.
It says staff could face professional risk if they accepted the system’s decisions when they disagreed with them.
- Who
- A global bank and its staff used the AI credit system.
- What
- The bank approved an AI system for routine credit decisions, but human staff continued to review most cases and usually prevailed when they disagreed with it.
- When
- The system was approved in late 2022; six months later, human review remained common.
- Where
- Not stated.
- Why
- The article says the bank’s accountability structures made it professionally risky for staff to defer to the system.
This story does not have two clearly opposing sides.
A global bank approved an AI system designed to automate routine credit decisions.
The system continued to assess applications and make recommendations, but most cases still received human review.
- Bank
- A global bank
- Approval
- Late 2022
- Purpose
- Automate routine credit decisions
- Pilot results
- Faster processing, fewer errors and promised cost savings
- Six months later
- Most cases still passed through human review