Cheap AI is a trap if your data is trapped

The price of AI is falling fast. Your AI bill is going up anyway.

That is a warning, and it has very little to do with the model you choose.

The models are becoming a commodity

Every few months a new model arrives that is cheaper and better than the last one. The clever money has stopped arguing about which one wins. Whatever you pick today, something better will exist by the spring. Model choice is becoming like choosing an electricity supplier: worth a look, not worth a strategy.

So why do costs keep rising? Because the real work, and the real spend, is everything around the model: getting your information into a shape the machine can read, connecting the systems it needs to act on, and checking the results. The model is the cheap bit. Your mess is the expensive bit.

Where the moat actually is

One question decides this:

If a free, brilliant AI model landed tomorrow, could it actually read and act on your business? Or would it be locked out like everything else?

For most hospitality businesses the honest answer is locked out. The menus live in a design file. The pricing lives in a spreadsheet on someone’s laptop. The trading knowledge lives in the heads of two managers. The systems don’t talk. An AI, however clever, can only work with what it can reach.

The pub company that wins with AI will not be the one that buys the cheapest model. It will be the one whose estate data, trading knowledge and operating context sit in a form a machine can reach. That is the moat. Nobody can buy it off a shelf, and nobody can copy it overnight.

What to do about it

Spend less time shopping for models and more time freeing your own context:

• Get your operational truth into one reachable place. Not a five-year data programme. Start with what guests and machines ask about most: opening hours, menus, prices, facilities, bookings.
• Prefer systems with an open door. When you next buy software, ask before anything else: can we get our own data out, automatically, in a standard format?
• Write down what your best people know. The trading rules of thumb, the supplier quirks, the recovery playbook. If it is only in someone’s head, no machine can ever take the task off them.

None of this is glamorous. All of it compounds. Every piece of context you free today makes every future model, including the free ones, more useful to you and less useful to your competitors.

Cheap AI is coming either way. The only question is whether your business will be in a state to use it.