One in three hospitality businesses still don't use any AI tools at all. That's according to recent research from The Caterer - and honestly, it doesn't surprise me.

Not because operators are behind the curve. But because most AI tools aren't built for them.

The AI hype machine runs on enterprise budgets. The demos show 500-site restaurant chains with dedicated data teams. The case studies reference "predictive demand modelling" and "autonomous supply chain optimisation". Then a café owner with 10 seats and three suppliers sees the same marketing and thinks: is this even for me?

The honest answer is - some of it is. Most of it isn't. And knowing the difference will save you from wasting money on tools that sound impressive but don't actually help you make better croissants or tighter margins.

What Actually Works Right Now

These are AI tools and features that deliver measurable value for independent cafés and small restaurant operations - not "eventually" or "at scale", but this month.

Invoice Scanning and Data Extraction

This is probably the single most useful application of AI for independent operators right now. You photograph a supplier invoice. The system reads it, extracts every line item, matches products to your ingredient list, and tracks price changes over time.

The good systems don't just do OCR (reading text from images). They learn your specific suppliers' formats - so the fifth invoice from your dairy supplier processes faster and more accurately than the first.

Why it works for independents: you're already receiving the invoices. There's no extra data collection. You just point your phone at a piece of paper and the system does what used to take you 45 minutes on a Tuesday evening.

Pattern Spotting From POS Data

Let's be clear about what this is and isn't. This is not the "predictive demand forecasting" that enterprise software companies sell. That requires thousands of data points, multiple locations, and often a data scientist to interpret the output.

What works for independents is simpler and more useful: pattern spotting. Your POS data already tells you that Tuesdays are slow, Saturdays need extra prep, and your oat milk flat white outsells your latte 3 to 1 after 2pm.

You probably know some of this instinctively. But having it in actual numbers changes how you make decisions. You stop overstaffing Tuesdays. You prep more pastries for Saturday. You adjust your milk order.

It's not revolutionary. It's just useful.

Recipe Costing That Updates Itself

Manual recipe costing is a pain. You build a spreadsheet, enter your ingredient costs, calculate your dish costs, then three weeks later your butter price changes and everything is wrong again.

AI-assisted recipe costing connects to your invoice data. When your supplier charges you more for double cream, your recipe costs update automatically. Your menu margins recalculate. You can see immediately which dishes are still profitable and which ones need repricing.

This is the kind of costing software that actually earns its subscription fee - because it stays accurate without you having to maintain it.

Rota Suggestions Based on Trading Patterns

If your POS knows that Friday evenings do 3x the covers of Monday evenings, your scheduling tool should know that too. AI-assisted rota tools look at your historical trading data and suggest staffing levels that match actual demand.

You still build the rota. You still know that Dave can't do Wednesdays and Sarah is finishing her degree. But the starting point is data-driven rather than guesswork, and that means fewer shifts where you're paying two people to stand around, and fewer where one person is drowning.

Simple Chatbots for Customer Questions

A basic chatbot on your website or social media that handles "what time do you close?", "do you take bookings?", and "is your sourdough vegan?" - this works. It's not glamorous. It won't transform your business. But it stops you answering the same five questions on Instagram every day.

The key word is simple. If a chatbot tries to handle complex complaints, dietary requirements, or anything that needs genuine human judgement, it'll cause more problems than it solves.

What's Still Hype for Most Independents

These technologies exist. Some of them work brilliantly - for the right operation. But for a typical independent café or small restaurant, they're not worth your money yet.

Fully Autonomous Ordering Systems

The pitch: AI analyses your stock levels, predicts what you'll need, and places orders with your suppliers automatically.

The reality: your head baker already knows you need more strong flour on Wednesday. Your ordering is probably 5-10 suppliers, and most of them you order from by WhatsApp or a quick phone call. Autonomous ordering adds complexity to a process that isn't actually that complex for a small operation.

When it makes sense: 20+ suppliers, multiple sites, a dedicated purchasing role. Not a single-site café.

AI-Generated Menus

Some tools claim to analyse trends, customer preferences, and margin data to generate optimal menus for you. In practice, your menu is a creative expression of your kitchen, your skills, and your suppliers. The idea that an algorithm should decide whether you serve a Victoria sponge or a Basque cheesecake misses the point of independent hospitality.

Good menu engineering tools help you understand the financial performance of your existing menu. That's different from generating a new one.

Predictive Analytics That Need Huge Datasets

"Predict your demand with 95% accuracy." Maybe - if you have three years of daily sales data across 50 locations. For a café that opened 18 months ago, there isn't enough data for meaningful predictions. The model is guessing, and your gut feeling is probably more accurate.

This is the big gap between enterprise AI marketing and independent reality. The tools work at scale. They don't work with small datasets. And nobody in the sales process tells you that.

Robot Baristas and Kitchen Automation

They exist. They're genuinely impressive technology. But a robot barista costs more than your annual turnover, and your customers chose your café over the chain specifically because a human made their coffee and remembered their name.

Automation makes sense in high-volume, low-personalisation environments. That's not what independent hospitality is.

The Only Test That Matters

When someone pitches you an AI tool, ask one question:

Will this save me time or money this month?

Not "eventually". Not "once you've been using it for six months". Not "when you add your second location". This month.

If the answer is yes - and you can see specifically how - it's probably worth trying. If the answer involves words like "scale", "over time", or "when the model has enough data", it's not built for you yet.

How to Evaluate Any AI Tool

Before you subscribe to anything, ask these five questions:

What data does it need from me? If it needs data you don't already collect, the setup cost is higher than they're telling you.
How long before I see value? Anything over 30 days for an independent is too long. You should see results within a week or two.
What happens if I stop paying? Can you export your data? Or is everything locked inside their platform?
Is it learning from my data or generic data? Generic AI models don't know your suppliers, your recipes, or your local market. The best tools learn your specific patterns.
Does it replace something I'm already doing, or add new work? The best AI tools eliminate tasks. The worst ones create new tasks disguised as "features".

What We're Building at Brikly

I'm not going to pretend Brikly is neutral in this conversation. We're building AI features into our tools, so I have skin in the game. But our approach is specific and worth explaining.

Every AI feature in Brikly learns from your data - your invoices, your POS sales, your recipes. Not a generic industry model. When CostingBrik processes your invoices, it learns your suppliers' formats and your ingredient names. When MenuBrik analyses your sales, it's looking at your actual trading patterns, not an average across thousands of restaurants you have nothing in common with.

The modular approach matters here too. You don't need to buy an entire platform to get AI invoice processing. You subscribe to CostingBrik, and the AI features are built in. If you later add MenuBrik, the data flows between them - your invoice costs feed your recipe costs feed your menu margins. But you're never paying for AI features you're not using.

The goal is tools that help you build a café that runs without you being buried in admin every evening. Not AI for the sake of AI. AI that gives you your Tuesday nights back.

The Bottom Line

AI in hospitality is real. Some of it works brilliantly for independents right now - invoice processing, recipe costing, POS pattern analysis, smart scheduling. These tools pay for themselves quickly and don't require a data science degree.

But a lot of what's marketed as "AI for restaurants" is really AI for restaurant chains that's been repackaged with a lower price tag. The underlying technology needs data volumes and operational complexity that most independents simply don't have. That gap narrowed again this month - Starbucks, Shake Shack, Deliverect and a ChatGPT ordering launch all shipped in the same week, and we broke down what each one actually means for independents in the 20 April Weekly Grind. Two weeks later, Papa John's launched its own conversational ordering agent and Toast bundled voice AI into a new POS - we tracked that next wave in the 4 May edition. And in the 11 May edition we covered an actual AI-run cafe in Stockholm that ordered 120 eggs for a kitchen with no stove - a useful reminder that AI is a tool, not a substitute for an experienced operator's judgement.

Be selective. Be sceptical. And always ask: does this help me this month?

The one in three operators who haven't adopted any AI tools yet aren't wrong to be cautious. They're just waiting for tools that are actually built for the way they work. And that's a perfectly reasonable position.

Ed O'Brien has run Hunters Cake Company for 17 years across cafés in Witney, Burford, and a bakery in Carterton, Oxfordshire. He's building Brikly - modular tools that give independent café owners the same data the big chains have, without the big chain price tag.