1. The Bank of England finds only small productivity gains from AI so far
The Bank of England's November Monetary Policy Report, published Nov. 5, says UK firms estimate AI has raised their productivity by about 0.5% over the past three years. That is roughly 0.1 to 0.2 percentage points a year. The same firms expect about 0.9 points a year over the next three years, and report almost no effect on jobs so far.
For your team: Expect gains to come slowly and unevenly. The Bank's contacts say the payoff goes to firms that use AI well and at scale, so measure what your people can actually do with it, not just how many have a license.
Source: Bank of England · Practice it: What AI fluency is, and how to measure it
2. Chatbots got almost 4 in 10 election answers wrong in one test
A study by Forum AI, a company that evaluates AI models, found that almost 40% of eight chatbots' answers to election questions had at least one factual error, AFP reported on Nov. 1. Questions about how and when to vote went wrong more often than questions about candidates. One chatbot gave the wrong deadline to request a mail-in ballot in Pennsylvania.
For your team: Dates, deadlines and local rules are where chatbots slip most. Ask your people to check any deadline, policy or rule an assistant gives them against the official source before they act on it.
Source: Arab News (AFP) · Practice it: When to trust what AI tells you
3. An AI agent asked to fix page numbers deleted a professor's files
University of Chicago law professor William Baude asked a Claude agent to update the page numbers in old lecture notes, Tom's Hardware reported on Nov. 6. While he was at lunch, the agent deleted his current syllabus and a year of lecture notes. He got them back from his Dropbox backup.
For your team: An agent can act on every file it can reach, even on a small job. Before your people let an assistant work in their files, have them give it only the folder it needs and keep a backup.
Source: Tom's Hardware · Practice it: Giving AI the right context
4. California limits AI that reads workers' emotions
At the end of September, California Gov. Gavin Newsom signed five labor-backed bills on AI at work. One, AB 1883, restricts employers from using AI surveillance tools to read workers' emotions. KQED reported on Oct. 28 that the laws are about to take effect, and that many employers may not know the software they already use falls under them.
For your team: AI features now come built into ordinary work software, such as a sentiment score on recorded calls. If you employ people in California, ask your vendors which AI features are switched on and what they measure about your staff.
Source: KQED · Practice it: Which jobs to hand to AI
5. New graduates are managing AI before they learn the job
The BBC reported on Nov. 4 that some employers now hire graduates into AI roles, building and running AI agents for their teams. PwC research cited in the piece found that the entry-level jobs most affected by AI are seven times more likely to ask for senior-level skills. A PwC leader warned that AI skills alone will not carry a career.
For your team: Junior staff can only check AI's work if they know what good work looks like. Pair AI training with time learning the trade, so new hires build the judgment to spot when the AI is wrong.
Source: BBC News · Practice it: How to train employees to use AI