The year 2020 will usher in a new era in the deployment of artificial intelligence (AI) as corporations seek to scale up pilots they’ve been dabbling in and make their businesses more intelligent. DXC Technology calls it the industrialisation of AI: “A bit like the introduction of electricity, it’s the new fuel for anything”, says Jerry Overton, DXC Fellow and head of AI at DXC. Unlike the proof-of-concept pilots that companies have been engaged in to date, industrialised AI calls for new levels of investment, expertise and knowledge.

Prepare for greatest disruptive force

Banks are deploying industrialised AI in a radical move towards price discovery: In the future everyone will see what a transaction service costs, as World Economic Forum (WEF) Financial Innovation Lead Jesse McWaters noted in an interview with Deutsche Bank. “The norm will be instant and real-time fast services, and as the move to digital distribution accelerates, extensive branch and broker networks look set to make way for digital connections (e.g., via apps) to become the norm.”

As this future disruption in banking indicates, decision makers in all sectors must get clued-up about what AI and machine learning (ML) can deliver. While no one doubts our future will be artificially intelligent, many headlines are misleading, alternately depicting a dystopian world controlled by bots or peddling a utopian vision. The reality lies somewhere in between, says Overton, and having a firm grasp on reality is vital for companies wishing to industrialise AI.

Get familiar with the four pillars of industrialisation

Overton sets out the four pillars of industrialisation in the practical language of “how to do AI”.

  1. How to operationalize: It’s a new arena, and the services you create have to cooperate with your other systems and applications.
  2. How to integrate: People come out of the woodwork and say they want the ML capability to plug into their applications, too.
  3. How to manage AI “drift”: ML can quickly become obsolete through new developments and customer trends; you have to regularly retrain the ML and algorithms and have procedures in place to ensure regular refreshes. AI that is not properly managed can also become a security or ethics risk.
  4. How to monetize: Organisations can spend vast sums of money scaling AI. They need to see a return on investment and think about that up front.

“Given that the industrialisation is not a sequential process, it needs a conductor to ensure the AI strands play to a coherent whole,” Overton says. “That’s DXC’s strong suit.”

Unlock data avoid AI silos

Organising the next phase of AI around these four tenets is sensible groundwork, but there are still problems ahead, Overton warns. A common obstacle he comes across is data being locked in different systems and in a variety of formats. If this isn’t sorted out at the beginning, there’s a danger of AI solutions being developed from different data sets and operating in silos, which would derail industrialisation.

Educating the board and other senior decision makers — as well as the workers who will be expected to work alongside cobots — is another foundational task in industrial AI. “Everyone should take machine learning classes to learn about the implications and ethics of using data, and to ensure it is not anthropomorphised. You can’t just leave it to the data scientists,” Overton advises.