By Paul Sandle

LONDON, April 25 (Reuters) - Britain could gain 400 billion pounds ($533 billion) from AI-driven growth if it trained its workforce, Google said, after a pilot scheme in the UK showed workers could save more than 120 hours a year by using AI in administrative tasks.

Simple steps such as giving workers permission to use AI and a few hours of training to get them started could help double the adoption of the new technology, and in turn boost economic growth, Google said in a report on its pilot scheme, published on Friday.

The U.S. tech giant, which developed the Gemini AI chatbot, said that according to analysis by Public First, its partner in the scheme, two thirds of workers - particularly older women from lower socio-economic backgrounds - had never used generative AI at work.

Debbie Weinstein, Google's Europe, Middle East and Africa president, said the AI Works pilots - conducted in a small business network, educational trusts and a union - showed workers could save on average 122 hours a year by using AI in administrative tasks.

But one barrier stopping some from dipping a toe into the water was a concern that using AI in their job was not legitimate nor fair.

"People wanted 'permission to prompt'", Weinstein said in an interview. "'Is it okay for me to be doing this?' And so giving them that reassurance was really important."

Once they started, a few hours of AI training to build their confidence resulted in them using the technology twice as much, she said, and they were still using it several months later.

These simple interventions helped to narrow the AI adoption gap amongst the participants in the pilot studies, Google said in its AI Works report.

Before training, for example, only 17% of women aged above 55 in its cohorts used AI weekly and only 9% daily.

Three months later, 56% were using it weekly and 29% had made it a daily habit.

($1 = 0.7509 pounds)

(Reporting by Paul Sandle; Editing by Susan Fenton)

((paul.sandle@thomsonreuters.com; +44 20 7542 6843; Reuters Messaging: paul.sandle.thomsonreuters.com@reuters.net))