Researchers analysing hundreds of executive interviews and public records found only a tiny fraction of AI pilots – those meeting demanding criteria – achieved a positive return on investment after six months.
The news comes hot on the heels of the disastrous rollout of ChatGPT 5, compounding the fears of some industry observers that expectations of AI’s transformative effects on business are overblown.
It is a sobering warning to Phuket people wrestling with how to incorporate AI into their lives and work. But the research also offers lessons on what to do – and what not to do – to up your AI game.
Encouragingly, the authors attributed the high failure rates not to shortcomings in the technology itself, but to poor design and management. Most pilots, they say, are doomed by a lack of clear goals, poor integration, and leadership failures.
Let’s take a deeper look at MIT’s recommendations and see how they apply to your Phuket business.
Decision-making
First, forget about “transforming” your business with AI. Successful projects were modest and focused on narrow, well-defined problems, especially in back-office or operational workflows. For your Phuket hotel, ditch the idea of an all-singing AI concierge. Instead, think about AI optimising housekeeping rosters based on checkout times, or analysing consumption to reduce food waste.
Second, integration with existing information systems predicted success. If an AI pilot could read from and write to the systems you already use, it had a much better chance of delivering ROI gains.
Third, decentralised decision-making was key. Where the project was run by the line manager – your accountant or your chef – who would actually use the tool, it succeeded. Where it was run by a central tech or “innovation” team, it tended to lose money.
Perhaps unsurprisingly, successful projects included clear KPIs and strong feedback systems. And as with any change, buy-in from senior management was essential.
Less obviously, the MIT researchers found that pilots tapping the expertise of external vendors or solution providers were twice as likely to succeed as purely in-house efforts.
Clean data
One of the first things external vendors always tell you is to get your data in order. Before you can use AI to personalise offers, you need to know what your guests like. If their preferences are currently scattered across three different messy spreadsheets, AI can’t help you. Clean data comes first.
Finally, realistic expectations are crucial. The MIT study’s timeline was just six months – a shortcoming its authors acknowledged. An AI project may well take longer to bear fruit.
Putting that all together, what does it mean for your Phuket business?
First, forget the moonshots. Don’t try to build a revolutionary AI tour guide. Instead, automate one tedious, expensive job, like responding to the same ten booking questions you get on Facebook every day.
Second, integration is key. Make sure AI connects to your core systems, whether that’s your property management software or your accounting package.
Third, empower your people. Let the bookings manager be the AI champion, not the IT guy. They know the real problem and can see if the solution is actually working.
Fourth, buy, don’t build. If you can find an off-the-shelf tool that solves your problem, use it instead of hiring a coder.
And finally, measure what matters. Before you start, figure out how much your problem is costing you. That’s the only way you’ll know if your investment in AI is a success and you’ve joined the winning 5%.
Joe Smith is Founder of the AI consultancy 2Sigma Consultants. He studied AI at Imperial College Business School and is researching AI’s effects on cognition at Chulalongkorn University. He is author of The Optimized Marketer, a book on how to use AI to promote your business and yourself. Contact joe@2Sigmaconsultants.com.