An example I made up.
The head of service at a regional bank's card centre in Kuala Lumpur has a decision to make before 1 March. From that date, her 800 front-line staff could settle a card dispute while the customer is still on the phone. Today they pass every dispute to a back office, which writes to the customer within 10 days.
The project plan in front of her says: "Staff trained, system tested, back office agreed. Ready for 1 March."
Every part of that line is true. It's also written only from the staff's point of view. It tells her the staff can do the new work. It doesn't tell her what the customer will see after the conversation ends.
So before she approves the date, she asks a team leader to follow one dispute, step by step, from the customer's point of view. She asks for the whole path, from the moment the customer phones to the day the next card statement arrives.
The team leader finds the gap. On the phone, the customer hears "your refund is approved". But the refund doesn't reach the card account before the next statement goes out, so that statement still shows the charge. A customer who was told the matter was settled now reads a statement that says it isn't. That customer will phone again, and the second conversation will be harder than the first.
Her decision is to start on 1 March, with one change: what the customer is told about the next statement. Staff tell each customer, while they're still talking, what the next statement will show and when the refund will appear. The confirmation message the customer receives afterwards is rewritten to say the same thing.
She tells her team leaders first, because they'll explain the change to their own teams. Then she tells the head of the back office and the chief operating officer.
Three months later, 9 in 100 customers have phoned back about the same dispute, against 24 in 100 under the old process. And every later change in the centre is followed from the customer's point of view before its start date.
Illustrative figures.
Where I learned it
HSBC Global Resourcing, 2005 to 2008: 33,000 front-line employees and 10 million customers. The name for that work was service transformation.
That is the only fact from those years that I use here. The example above is made up and is not from those years. The fact shows the scale of the two groups a change like this reaches. They are the people who do the work and the customers who receive it.
What it taught me
Everything around a rollout decides whether the result lasts: trust, judgment, leadership, process, skills, data, responsibility and culture. A project plan can report on only some of those. The plan in my example covered the staff, the system and the back office. It didn't cover the process the customer goes through after the conversation ends, and that's where the result would have failed to last.
Finding an answer is rarely the hard part. The hard part is knowing which answer to act on. "Ready for 1 March" was an answer, and a reasonable one. The head of service still had to decide whether to act on it, and on what terms. She kept the plan and the date, and she changed one thing. That was her decision, and she told the people who needed to know before the date.
A change to front-line work may now come with AI in it. An AI tool drafts the reply to the customer, or suggests the outcome of a dispute before a person looks at it. If the path the customer follows has a gap, an AI tool takes more customers to that gap, and faster.
So the AI point follows from the example. AI can speed up existing work. People still decide what needs to change.
In my example, what needed to change was one sentence on the phone and one confirmation message. Nobody would have found them by reading the plan. Someone found them by following one dispute from the customer's point of view to the next thing the customer sees.
The example above is made up, and so are its numbers. I'm careful not to pretend to be an expert in an industry I haven't worked in, so I don't claim the example fits yours. And nothing I give at an event measures or scores anyone.
If you're about to change how a large front-line team works, with AI or without it, that check is worth doing before the start date. Pick one real case and follow it past the conversation, to the next statement or message the customer receives. Then decide what changes before the date, and who needs to be told.
I give keynotes, briefings and workshops on this.