Not another one

There is a famous British political meme that dates back to the 2017 snap election. In a vox pop, Brenda from Bristol, when asked about the forthcoming election, says: “You’re joking! Not another one”. The same goes for articles about AI. And yes, this is indeed another one (sorry, Brenda).

As a business and finance journalist covering hospitality and related tech developments, I have a simple rule of thumb for dealing with people who make predictions about AI. Rather like the old Yogi Berra joke that predictions are hard, especially about the future, if somebody starts speaking about AI without using the phrase “I don’t know” in at least every other sentence, I suspect they are either a fool or a charlatan.

But this uncertainty surrounding AI, particularly the impact of Gen AI since the launch of ChatGPT at the end of November 2022, does not mean you cannot construct a framework for thinking about the new tech. And while, yes, I don’t know, I can share my framework for approaching the subject.

Let’s start with setting some boundaries. To help set these, I want to reach for the biggest brains in our society, and I think the 2024 Nobel Prize winners are a good place to start.

At the bottom end, by which I mean somebody who is deeply sceptical about the hype, I offer you the 2024 Nobel Prize winner for economics, Daron Acemoglu. He is a professor at MIT and won the prize, jointly, for work exploring why different nations exhibit different economic performances.

Acemoglu reckons that AI will improve global growth by just 1% to 1.5% in total, and mostly thanks to the current investment boom. This is a long way from the McKinsey forecast that global growth is going up three to four percentage points each year.

There is particular scepticism from Acemoglu around the potential for productivity gains, reckoning that total factor productivity (tasks that are automated multiplied by the cost savings) will be just 0.66% over 10 years or just 0.06% a year.

At the opposite boundary, I offer Geoffrey Hinton, the Nobel prize winner for physics in 2024. Hinton is sometimes described as the godfather of AI. A typical quote from him: AI is going to change the world more than anything in the history of humanity. More than electricity.

Hinton is currently doing the media rounds, warning of a Terminator-style scenario in which the bots take over.

My final boundary-setting Nobel prize winner from 2024 is Demis Hassabis, the CEO of Google’s DeepMind. Hassabis won the prize, jointly, for chemistry for his work on AlphaFold, which predicts 3D structures of proteins. This AI-enabled process has transformed the economic logic of structural biology.

The Hassabis view is that AI could bring about a transformation even greater than the Industrial Revolution. This ties in with the World Economic Forum’s view that sees AI as a cornerstone of a phase in which digital, biological and physical technologies converge to deliver structural, rather than simply incremental, change.

This boundary-setting exercise concludes in a ludicrously wide range, going from a nothing burger to the end of humanity. My view is that Acemoglu and Hinton are both wrong, even if IDK precisely has to be declared.

It is pretty clear already that Gen AI represents a platform shift at least as significant as the internet or the introduction of smartphones. And, since AI is essentially a probabilistic process of determining which token comes next that matches the pattern, Hinton’s fears of a triumph of the bots seems a way off.

Hassabis is at the hype end of the spectrum, but I would suggest he represents our upper bound. For the lower bound, I would head to the technology consultant and commentator Benedict Evans, who says AI is the next 10- to 15-year platform shift following mainframes, PCs, the web, and smartphones (though he keeps his bets open but suggesting it might be more).

As an aside, I think Evans’ statement that when things are exciting, people get excited is spot on. He points out that boosters confuse by drawing straight lines on log scale charts. But he also points out that AI is different from previous platform shifts because, with those previous changes, we knew the physical limits of what could happen next year; with AI, we don’t.

At the micro level for hotels, there are three broad types of systems that will be impacted by AI: digital, physical, and service / operational.

The digital impact will be on revenue management, PMS, CRM and guest-journey analytics. Physical is smart building automation, robotics and sensor networks. Finally, labour optimisation, predictive maintenance and wellness or biometric personalisation is where the impact on service and operational systems will be seen.

Let’s take a quick look at the impact of AI on one of those roles, revenue management. If you were to take a discipline most suitable to AI-disruption, revenue management is probably the one because it aggregates huge amounts of data, but the output does not require absolute precision (as in law or finance).

Could AI deliver the same outcome as the automatic lift did for elevator operators? Perhaps eventually, but in the meantime, a significant reduction in headcount is almost certainly on the cards.

AI is already better than humans at demand forecasting, price-elasticity modelling, inventory optimisation, channel mix optimisation, macro pattern detection (of things like events, weather, competitor rates) and real-time repricing at scale.

Humans are still needed in the loop for context interpretation, narrative construction to explain demand forecasts to stakeholders, cross-functional commercial strategy, which requires reconciling conflicting goals rather than optimising a single metric (an example is profit versus guest experience), data quality triage where flawed inputs have to be accounted for, and ethical or brand-sensitive decisions about how far to push pricing.

Over the next five years or so, as many as half of all revenue management jobs may go in the hotel sector (this is a guestimate pumped out by, of course, AI when I submitted suitable context data). Hit hardest will be revenue managers working at economy and budget hotels, with more stickiness for more complex properties at the luxury end or those with significant MICE business.

AI is good at explicit data. In fact, not just good, but typically much better than humans. It is weak at dealing with tacit information, that is, information that is uncodified and context dependent. This might be a GM’s sense of how a competitor might react or a sales manager’s intuition about a key account.

Tacit information is hard to access because it is not typically recorded, and AI has nothing to learn from. It can often contradict structured data and be ambiguous. It is socially learned and requires judgement.

Given the challenges of tacit knowledge, hospitality jobs, particularly higher-value roles, look set to remain much less vulnerable to AI disruption.

This brings me to the macro-level impact of AI on hospitality. There are two angles I want to explore here.

The first is what the productivity improvements will bring to overall economic growth, and where the extra wealth created will be spent. The last part is easier to predict than the quantum of the first.

We do know that in high-income countries, particularly among the better off, we are seeing a prioritisation of spending on experiences rather than goods. Evidence to support this contention comes from the US Consumer Expenditure Survey, which shows that spending on physical goods has been declining for over 20 years while real household spending on experiences such as travel, dining and entertainment has risen meaningfully. Similar data from the UK and EU can be found.

You can thus take whatever level of extra GDP growth from AI you buy into and translate that into meaningful tailwinds for hospitality and travel.

My second angle is on culture. We have seen in liberal democracies over the past decade or so that James Carville’s dictum that “It’s the economy, stupid” has much less purchase than it did in Clinton’s successful 1992 presidential run.

Looking at cultural positions, progressive versus small “c” conservative, explains as much, perhaps more, about politics than economics alone does.

With AI shifting economic power away from explicit human knowledge, the codified sort tested in exams, towards tacit knowledge, which is normally learnt on the job, we are heading towards a re-ordering of the workplace that might be as big as the political re-ordering we are currently confronting.

In this workplace re-ordering, skills embedded in hospitality, such as the empathy demanded of front-of-house staff or the craft skills of chefs, will be valued more highly. The cognitive skills that lead to the fantastic pay packets of higher professions like law or banking face a comparative decline.

There is much more to say on this cultural shift. But it requires a separate column on AI. Another one.