How best to avoid a bad AI rollout

{"text":[[{"start":5.44,"text":"Disastrous attempts to roll out AI strategies by Commonwealth Bank and McDonald’s are often cited as examples of what can go wrong when they are not planned properly."}],[{"start":14.88,"text":"Just a month after Commonwealth Bank announced it would replace 45 call centre staff last year with an AI chatbot, the Australian lender was forced to reverse its plans when the move resulted in a jump in customer calls, requiring increased overtime from the remaining staff and the drafting in of senior management to help clear backlogs. The bank apologised to the dismissed employees and offered them their jobs back or voluntary redundancy."}],[{"start":40.825,"text":"McDonald’s also retreated two years ago from rolling out AI voice-assisted ordering across its US drive-through restaurants after customers complained of a number of mishaps, including the arrival of an ice cream cone topped with bacon."}],[{"start":56.72,"text":"Experts helping companies roll out AI strategies say these high-profile cases demonstrate how not to do things, and why patience and planning are key to successful deployment."}],[{"start":64.88,"text":"Rather than dwell on what might go wrong, Suveer Kothari, head of Europe at Sierra, a start-up that assists groups with their AI strategies, prefers to focus on what successful companies get right: “First, they focus on a clear business issue and then apply AI — versus what we call AI tourism, where teams decide they need an AI strategy, often as the result of a top-down mandate, and then go looking for a problem to solve.”"}],[{"start":90.381,"text":"Second, he says companies that do well go live with real customers quickly, which enables teams to understand what is working and what might need improvement."}],[{"start":null,"text":""}],[{"start":99.187,"text":"“And third, they don’t try to boil the ocean. They pick a manageable issue that AI can solve to launch with,” he says. “Once they’ve cracked that, they choose use cases that get progressively more complex over time.”"}],[{"start":111.189,"text":"Nicola Parton, group AI business lead at insurer Swiss Re, agrees that prioritisation is important."}],[{"start":117.128,"text":"“A real pitfall is trying to do too much at once. There is enormous demand for AI, so prioritisation matters,” she says. “You need to focus technology and business expertise on the areas where AI can genuinely unlock better outcomes, while building the foundations — data, technology, governance and organisational activation — that allow successful solutions to scale.”"}],[{"start":137.2,"text":"Parton says Swiss Re also made AI tools available across the company early for non-core work, such as summarising documents and preparing for internal meetings. The thinking was that if people did not become comfortable using AI in their day-to-day work, then there was little chance they would use it effectively in key business areas, which she says now includes helping underwriters carry out compliance tasks and finalise contracts."}],[{"start":163.12,"text":"Tools for non-core work “are relatively simple applications, but they help people build familiarity and confidence with AI through regular use”, she says."}],[{"start":171.486,"text":"Another pitfall is treating AI as a technology rollout rather than as a business transformation that is integrated into an overall strategy, she says. “It is relatively easy to launch tools or individual use cases, but it is much harder to change how work actually gets done and translates into measurable business value.”"}],[{"start":187.84,"text":"At UK bank Griffin, Ed Carter, chief operating officer, adds that “measuring the wrong things” is a waste of time. “Asking ‘are we using AI enough?’ leads to counting hours. We should rather measure output and encourage experimentation in the right ways.”"}],[{"start":204.613,"text":"While there is agreement that honing strategies leads to better AI outcomes, there is less consensus on who should be responsible for rolling those strategies out."}],[{"start":null,"text":""}],[{"start":214.08,"text":"For some companies that job sits with the IT department, for others it rests with the chief executive, while for many, such as HSBC, it is the responsibility of the chief AI officer, whom the bank appointed for the first time in April."}],[{"start":227.494,"text":"Parton is not convinced that all companies need a chief AI officer to get implementation right. She says there is no catch-all approach, rather that clear accountability is needed. Whether or not that involves the appointment of a chief AI officer depends on the organisation and its stage of maturity."}],[{"start":244.67,"text":"“A dedicated chief AI officer can make sense at a certain stage, particularly when an organisation needs to create momentum or co-ordinate activity across many parts of the business. But there is also a risk that appointing one inadvertently signals that AI belongs to one person or function.”"}],[{"start":260.24,"text":"Florian Mueller, a partner in the applied technology and AI practice of consultancy Bain, agrees. “A chief AI officer can be a great builder but also a dangerous alibi,” he says. “Someone has to own the platform, the orchestration layer, and the talent, and that can absolutely be a chief AI officer."}],[{"start":279.758,"text":"“However, the big decisions on where to place the bets, how to redesign the organisation and how to engage the employees can’t be delegated. If employees only hear about AI from the AI chief, it is a project. If they hear it from the CEO and the entire executive team, repeatedly . . . it is a strategy.”"}],[{"start":302.72,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1793349523_1721.mp3"}