AI is not a savior, it's a power-up.

I've been working as an advisor for a recently acquired (search fund) company. New management started doing some spring cleaning right away. They focused on a well-paid data analyst. After inspecting her work, they decided that the AI is here, and she is not as important, so they can let her go. They asked her to describe her work, processes, and edge cases, and got an external company to develop an AI automation that will do her work. All sounded like an amazing plan. The company was expected to grow quickly, and they wanted to ensure they would nail data analysis with AI and avoid the unnecessary cost of hiring more experts. The system was turned on, they fired the expert, and called it a day. Fast forward 3 weeks after the data analyst left, and the financial operative noticed that one of the numbers in the summary was off. After carefully checking where it went bad, it traced to the AI - an undocumented anomaly occurred, and the AI didn't have real-life experience to deal with it, so it made a probable number up. The anomaly was rare, but they added the adjustment to make it work next time it happens and got the finance person to double-check key metrics for a week to ensure the problem is gone. Luckily (?), another rare anomaly happened after just 2 days. And then another after 5 days. Now this was a serious issue, as the data was crucial for running the business. This is where I was brought - asked to determine a solution or at least a way to mitigate the risk. The tech wasn't too hard - the AI can catch those anomalies with a bit of coding and logic, but someone needs to govern if 1 plus 1 equals 2, and also teach AI and lead when the anomalies are detected. The financial department refused to take on a new role (rightfully so, given the type of business, this would put too much responsibility on them and add to their workload). So in the end, the only reasonable solution was to hire a new expert data analyst, who will oversee AI and ensure all data and edge cases are handled correctly. Not only did the new expert have a higher starting salary than the previous one, but also needed training, and was a liability for the finance team, as it needed their guidance at the beginning. Overall, the entire finance department was furious because of all of this - they had to check all the data submitted by AI manually. It was a lot of additional work for them, a massive setback for the company, and the cost was substantial. The company would save money long-term if it did not believe in the stories about AI being able to completely take over a human role, and if it used AI to empower existing employees (and save on future hiring) rather than trying to reduce costs right away. Guarantees given by the dev company that this system will work flawlessly didn't help either (they downplay the role of a human in all of this). Now I'm posting this story as a reminder - a revolution is happening, but rushing it will cost you a lot. Using AI to automate an entire job role is a miss and a hit. Using AI to power up employees in a controlled way is mostly a win.