A successful AI strategy can’t just be based on technical capability within a business but should be underpinned by an agile culture that’s open to experimentation and a willingness to embrace data-driven decision-making, according to Microsoft’s CTO, Worldwide Services, Norm Judah.
“Every company needs to have an AI strategy,” Judah said. “Because in some form or another — particularly companies that have large volumes of data — they will be using AI to process data, to look for patterns in that data, to make recommendations against the data.”
“Some are there — some are progressing a little more slowly. It depends on the industry, depends on the country, depends on the maturity [of the company],” he added. “But no AI strategy is not a strategy, the same way that no cloud strategy was not a strategy.”
Microsoft has developed an AI maturity model that is intended to help answer the question, “Where do I start?” the Microsoft executive said. It’s based on four pillars, only one of which is technical capability. The other three relate to strategy, organisation and culture.
“Because dealing with AI actually turns out to have a cultural impact on the company — how you deal with data, how you deal with decisions being made,” Judah said. “So the more successful companies are those that actually address the cultural issues inside of a company, the ability to make decisions quickly, to do agile development for example.”
NAB this morning revealed that it had used Azure Cognitive Services for a proof of concept project involving using facial recognition instead of a card to authorise ATM transactions. The proof of concept took two months to build.
“So, when we talk about Agile, that’s a fairly good [example] of how you can actually take a hypothesis — and the hypothesis was could we do cardless access for an ATM — and then actually build something real to evaluate against that hypotheses in eight weeks,” Judah said.
“It's not huge a long process; it's typically a cross-functional team that will do this to actually get it up and running. But in order to get there, the company has to understand the notion of agility.”
“There's another interesting question that comes with that, which is do you make data-driven decisions as a company or are your decisions heuristic; are they intuitive-like decisions rather than data-driven decisions,” he added.
Judah gave the example of monitoring customer waiting times at a bank branch: “Go to a bank manager and pull him aside and say, ‘Do you know how many people were in your bank branch at 3 o'clock yesterday?” The probability is that they have no clue how many people were waiting in each one of the queues. Do you know if anybody came into the branch and turned around because there were too many people? Do you have that data?
“Now we can actually get that data; it's fairly easy to take a video camera and actually synthesise that data and do it anonymously. We don’t have to see their faces, just see that there are people waiting.
“But if I give a bank that data, what would they do? Would they redesign the branch? How would they think about that information? Would they redesign the staff layout so that they actually have a different number of people servicing it? Can they use that data?”
“AI is all about engines that look for patterns of information. If I give you the pattern — if you make intuitive decisions it's just more data to ignore,” Judah said. “So successful AI actually is a little bit about the way the company works and the way they make decisions; it's not just ‘Can I recognise your face?’”