How to get the business to buy into AI

Show the board how AI can make every employee an expert says insight and tech chief of financial services firm BGL Group

Artificial intelligence (AI) advocates within enterprise IT teams must focus their efforts on solving "real business problems" if they are to convince the board to support the technology.

That was the key message from BGL Group's insight and technology manager, who was speaking at Computing's Big Data and Analytics Summit 2016 this morning.

Mike Maddock began by spelling out the differences between the public's perception of AI, and what it can now realistically achieve for the enterprise.

He described Gartner's recent prediction that, by 2020, five per cent of financial transactions across the world will be carried out by AI, as a "bold claim".

He also cast doubt on the usefulness of hypothesis-driven AI, as epitomised IBM's Watson, which achieved fame after its victory in the US gameshow Jeopardy.

"I'm not convinced by [it]," said Maddock. "It's greater than traditional systems, but [hypothesis-driven AI] is certainly no match for us at the moment."

Instead, Maddock described a future where AI can make any employee "an expert" while also freeing up serious data scientists to do more complex, and potentially more lucrative, tasks.

"A key thing is everybody becoming an expert," said Maddock.

"For lawyers, it's about looking at case histories, but using AI, every employee can become an expert [in law].

"[Also], we don't want to tie up our data scientists dealing with trivial stuff - so not asking data scientists to look at data of trivial value, but moving to saying ‘Here's a big area to explore, can you move your big brain to that' [while AI deals with lesser tasks]."

Maddock said he recently ran an explorative AI workshop at BGL, but "only invited business people".

"IT people love the tech, but it's about solving business problems now," he said. "I strongly suggest you explore AI with business problems to get that buy-in.

"And make it so you're not trying to solve a whole problem - break it into little steps. The value of learning from this is excellent - it's not just monetary."

Maddock said his experiences with AI showed that there is still some way to go before insurance providers like BGL can use AI to recommend policies.

"Training AI for [text] recognition is lengthy process," he explained.

"For example, if someone asked, 'am I covered to ride a motorbike in Malayasia?' your AI could scan a whole document but miss one clause on page 2 that says '...except in Malaysia'.

"But for historical data it's brilliant," he added, explaining that for working quickly within more defined parameters, AI can take the burden off people who could spend their time more profitably on more complex tasks.

Maddock warned that using AI in the enterprise is still about "managing expectations".

"Senior people tend to want things within the lifespan of their particular role - they're not really investing to make their successors' jobs easier," he said. In other words, many senior executives may not have the patience to explore AI's potential.

Customers, too, need to be introduced to AI in the right way, he said.

"If you've got something like AI as a customer-facing interface, as a cost-saving measure, think what happened when call centres started to be outsourced to India - it wasn't well received."