Experts say small, planned steps can prevent AI rollouts from failing

In short: Experts say companies can avoid messy AI rollouts by starting with a clear problem, testing early with real customers, and tackling small tasks before harder ones.

What's going on

Several well known companies have pulled back on AI plans after problems hit customers and staff. The Financial Times points to Commonwealth Bank in Australia, which said it would replace 45 call centre staff with a chatbot (a computer program that talks to customers). A month later, the bank reversed course after customer calls increased, overtime rose, and managers had to help clear backlogs.

McDonald’s also stepped back from using AI voice ordering at US drive-throughs after customer complaints about errors. One example mentioned was an order that produced an ice cream cone topped with bacon.

People who advise companies on AI say these cases show a common pattern, moving too fast without enough planning. Suveer Kothari of Sierra says successful teams pick a specific business problem first, then use AI to help, instead of adopting AI and then searching for a use. He also says companies should go live with real customers early, like a small pilot test, and they should not try to solve everything at once.

Swiss Re’s AI business lead, Nicola Parton, says prioritising matters because demand for AI inside companies is high. She said Swiss Re first encouraged simple, everyday uses, like summarising documents and preparing for internal meetings, so staff could get comfortable. Parton also warns that AI should not be treated as just a tech install, since the harder part is changing how work gets done.

What to watch

A key open question is who should own AI rollouts, IT, the CEO, or a dedicated chief AI officer. Experts quoted by the FT say titles matter less than clear accountability, and that top leaders need to stay involved so AI is seen as part of the company’s strategy, not a side project.

Source: Financial Times