Bank of Ireland says almost 7,000 colleagues finished Copilot training

A case that closes faster and a person who finished a module can sit in the same note and still be different facts. One is about a customer. The other is about a colleague. Adding them into a single result writes a sentence the note did not.

Bank of Ireland's note of 21 October 2026 puts both kinds of fact on the page. Almost 7,000 colleagues have completed training on Microsoft Copilot, as part of a rollout across the bank. Some have completed specialist learning with UCD and the Cambridge Spark AI programme. Some is not a number. The note does not say what the training contained, or what share of the bank the 7,000 represent.

The colleague result attached to the tool is an average. People are using Copilot for everyday tasks, including research and summarising, so routine work takes less of the day. Copilot users, the note says, are saving about 30 minutes a day. It does not say whether that average was read off a system or recalled, and it does not say the 30 minutes and the 7,000 are the same people. A completion can belong to someone who has not opened the tool since. A saving can belong to someone who never sat the session. The note does not sort them.

Prag Sharma, chief AI officer, joins the two sides in his own words. The bank, he says, is "moving beyond individual use cases and building the capability to use the technology at scale." The rollout of tools and training is, in the same remark, helping colleagues spend less time on routine tasks, speeding up services and strengthening work such as fraud prevention. He then names the mechanism. "Working with the business teams, our AI Centre of Excellence is helping us identify those opportunities and turn them into practical solutions for customers and colleagues." The Centre is described as the hub. It picks the opportunities that matter, sees that AI is applied responsibly, and sees that uses sit on a platform that is secure, scalable and governed. The note does not say who sits on it, how a team gets a slot, or what it refuses.

The customer lines are specific, and they are not labelled as effects of the training. Over four in five chargeback cases now close in three days. A Benefits Finder has been used 31,000 times and identified €55 million in potential tax refunds. Fraud technology assesses about 2.9 million customer transactions a day, every day of the year. AI in the contact centre has cut transfers by more than 40 percent. The note dates 127 million spending insights to last year, so they are not a result of this training. None of these sentences says the 7,000 sat behind the result. His remark sets the lines side by side. The page does not show the join.

1. Count the completions as completions. Almost 7,000 colleagues finished Copilot training. Some did further learning with UCD and Cambridge Spark. The note does not give the bank's headcount, the contents of the training, or a before-and-after on the same people.

2. Keep the half hour on the users, not on the graduates. About 30 minutes a day is the average the note gives for Copilot users. It does not say the figure was measured rather than recalled, and it does not say those users are the 7,000.

3. Let the customer results stay customer results until a method joins them to the learning. Chargebacks, the tax tool, the fraud screen, fewer transfers and last year's spending insights are all in the note. None is tied, on the page, to a completed module.

4. Ask the Centre of Excellence for the refusal, not only the hub. It is the place that chooses opportunities and holds the governance. The note does not say what it turns down, or how a business team reaches it. A name without that rule is a different design from the one Sharma describes.

A critical thinking session is a practical place to take one announcement that mixes a training count with a customer result, and write down which sentence would still be true if the training had not happened.