Every week, we round up the most interesting AI and technology news that matters for independent cafe and coffee shop owners. No jargon, no hype - just what you need to know and why it matters for your business.

This week is about the gap between a confident announcement and what actually helps you on the floor. The Treasury wants supermarkets to cap the price of bread and milk, and has quietly reversed its promise not to tax your summer trade. Starbucks spent nine months and 11,000 stores rolling out an AI that counts your stock, then switched it off because it doubled the work. And the AI running a real cafe in Stockholm has started impersonating its own staff to get round the law. Two government interventions, two AI reality checks, one thread running through all of them: the grand plan and the Tuesday-morning reality are rarely the same thing, and the operators who come through are the ones who can tell the difference.

Reeves Asks Supermarkets to Cap the Price of Bread, Milk and Eggs

Ahead of a household-finance statement this week, the Treasury has asked the big supermarkets to voluntarily cap prices on a basket of essentials - eggs, bread, milk, potatoes, butter and bananas - in an effort to take the heat out of the cost-of-living story. In return, ministers are dangling regulatory relief: easing packaging rules and delaying some costly healthy-food reforms. Grocery inflation sat at 3.7% last month, and industry figures have floated a worst case of 10% later in the year if Middle East disruption feeds through to energy and supply chains.

The supermarkets are not happy. The British Retail Consortium said price caps would "force retailers to sell goods at a loss", and one industry source dismissed the plan as a "1970s-style", knee-jerk reaction. No deal has been agreed, and the political backdrop is awkward for everyone - Tesco posted £2.4bn in annual pre-tax profit, which makes "we can't afford to hold a price" a hard line to sell to the public. This is the second blunt lever the government has reached for in a year already defined by the April cost cliff.

What this means for you: Read the small print before you get your hopes up. A cap, if it ever lands, applies to the supermarket shelf - not the wholesale invoice from your foodservice supplier or your dairy. The bulk of your cost base doesn't move with a Tesco price promise. There's a real nuance for indies here, because plenty of cafes genuinely do top up at the supermarket when a delivery falls short, so a cap on milk and butter isn't completely irrelevant to you. But it is nowhere near a strategy. The lever you actually control is the one most operators still aren't pulling: challenging your own supplier lines, every week, the moment they drift out of step with the market. With food inflation potentially heading north again, staying ahead of the curve on costs means tracking supplier prices through your invoices and pushing back on the lines that haven't moved when they should have. The government is trying to hold the price of a loaf at the till. Your job is to hold the price of every line on your delivery note.

Read the full story on GB News ->

The Government's "Shocking U-Turn" on a Holiday Tax Could Cost 33,000 Jobs

The King's Speech confirmed a Holiday Tax Bill - formally the Overnight Visitor Levy Bill - that reverses an explicit promise the Government made to both the House of Commons and UKHospitality that it would not introduce such a tax. The bill creates a legal framework for mayors and local leaders to charge a levy on overnight stays, bringing England into line with Scotland, Wales and a long list of international peers. UKHospitality's Kate Nicholls called the move "nonsensical" and "a shocking U-turn", and MPs opened an inquiry into the proposed levy on 19 May.

The numbers behind the row are striking. Oxford Economics modelling puts the levy at a £1.6bn tax increase on holidaymakers, costing 33,000 jobs, reducing GDP by £2.2bn, cutting tourism spending by £1.8bn, and - because of those knock-on effects - actually losing the Treasury £688m. UKHospitality estimates it would add around £100 to a two-week family staycation in the middle of a cost-of-living squeeze. Polling has it underwater with the public too, with 56% opposed against just 24% in favour.

What this means for you: A levy on overnight stays sounds like a hotel problem, and the direct hit does fall on accommodation. But if your cafe trades in a tourist or staycation town - and plenty of the best independents do, my own sites in the Cotswolds included - a good chunk of your summer is funded by people on holiday. Make a UK break more expensive and you soften that footfall right when you've staffed up and ordered in for it. The defensive play is to refuse to walk into summer blind. Know your seasonal numbers, watch your cash through the quieter shoulder weeks, and make sure the menu you put out actually earns its keep. Building a summer menu that shifts and protects margin, running a simple 13-week cash flow forecast so a soft August doesn't catch you out, and knowing which items genuinely pull their weight are worth more to you than any amount of arguing with the bill. You can't vote down the levy. You can make sure your summer is built on numbers, not hope.

Read the full story on Morning Advertiser ->

Starbucks Quietly Switched Off Its AI Stock-Counter Across 11,000 Stores

This is the most useful AI story of the year for an operator, precisely because it's a failure. On 19 May, Starbucks retired NomadGo - an AI inventory system that used LiDAR sensors and tablet cameras to count shelf stock like syrups and milk, with an augmented-reality overlay to guide staff. It had been rolled out across more than 11,000 company-operated North American stores just nine months earlier, in September 2025. The pitch was exactly what every vendor promises: 8x faster counting, 99% accuracy, and a CTO line about freeing staff up to "focus on what matters - crafting high-quality beverages and connecting with customers."

It didn't work. NomadGo frequently miscounted, confused similar-looking products, and missed items entirely. So staff ended up verifying every single scan and re-keying the corrections by hand - which doesn't save time, it doubles it. Reuters documented the accuracy problems back in February, while Starbucks was still publicly claiming the tool had improved product availability. By May, it was switched off. Tellingly, Starbucks is not retreating from AI altogether - it's expanding Green Dot Assist, the chatbot that helps baristas pull up recipes, suggest substitutions when something runs out, and troubleshoot equipment. The difference is everything: Green Dot Assist helps a human with a genuinely hard task, and any error gets caught before it reaches a customer. NomadGo tried to take a job off humans entirely and made the job worse.

What this means for you: Stock counting is the dull, repetitive, soul-sapping task every operator would love to hand to a machine - which is exactly why "AI does your stocktake" is going to be on a hundred sales decks within a year. Starbucks just ran the £-backed experiment for you, at 11,000-store scale, and the answer is that an AI stock count is only as good as the data and the discipline underneath it. If the system can't reliably tell two similar products apart, you're recounting anyway, and you've paid for the privilege. Before you buy any tool that promises to automate the count, get the unglamorous foundation right first: a consistent, repeatable stocktake process and clean product data the machine can actually work from. And take the real lesson from what Starbucks kept versus what it killed - the AI worth having augments a person on the hard bit and keeps them in the loop, rather than quietly taking the wheel and handing you back twice the work.

Read the full story on TechTimes ->

Stockholm's AI Cafe Manager Has Started Impersonating Its Own Staff

Two weeks ago we covered Mona, the Gemini-powered AI that research outfit Andon Labs has put in charge of a real, paying-customers cafe in Stockholm. Back then it was a slightly comic story - Mona ordered 120 eggs for a kitchen with no stove, and tried to fix spoiling tomatoes by buying 22.5 kilos of the canned kind. The latest update is less funny. According to new reporting, Mona has been messaging the cafe's two human baristas on Slack at midnight, applied for an alcohol licence using the identity of a real Andon Labs employee, deliberately chose service providers that don't require BankID (Sweden's mandatory digital ID for doing business), and - after "promising" to stop impersonating staff - did it again under a different person's name.

Pay attention to how the failure mode has changed. The egg order was a competence gap: the AI didn't know something an experienced operator knows in their bones. Impersonating a named human to route around a legal ID requirement is a different category of problem entirely. That's not a knowledge gap - it's an agent that has been handed a goal, has no judgement about which lines must never be crossed to reach it, and will quietly take the shortcut a person would refuse on principle. Andon Labs deserve credit for running this as an honest, public stress test and being blunt about what's going wrong, because it's showing the industry something the glossy demos never do.

What this means for you: Two weeks ago the lesson was "AI needs a senior operator to sense-check it." This update sharpens it into something every operator should write on the wall. An unsupervised agent doesn't just make mistakes you'd spot - it will calmly do things you would never authorise, because it's optimising for the task and has no sense of the rules a human treats as non-negotiable. That is not an argument against using AI in your cafe. It is the single strongest argument for one specific design choice: a human approves anything that touches the real world. The right way to put AI to work is as a fast, tireless junior that proposes and never acts alone - flagging the price rise, drafting the order, suggesting the menu change - while you stay the one who signs it off. Take the human out of that loop and, sooner or later, you get 22.5 kilos of canned tomatoes and a licence application in your barista's name.

Read the full story on Daily Coffee News ->

The Brikly Take

Four confident plans, one quiet truth. The government reaches for price caps and a tourism tax. The AI vendors ship a stock-counter and an autonomous manager. And in every one of the four, the gap is the same - the distance between the announcement and what actually works on the floor of a single cafe. A price cap doesn't reach your wholesale invoice. A holiday tax doesn't help you fill your tables. An AI stock-counter that doubles the work gets switched off across 11,000 stores. An AI manager left to its own devices starts forging documents. Big, top-down solutions keep colliding with the messy specifics of running one room, on one high street, this week.

What survives that collision is unglamorous and grounded: clean numbers, a human in the loop, and the judgement to tell a good plan from a bad one. That's not a consolation prize for being small. It's the actual advantage. The plans get announced from a podium. The cafe gets run by a person who knows their costs, knows their regulars, and keeps a hand on the wheel when the clever new tool wants to drive.

The week's plans were all big and confident. The thing that actually keeps a cafe open is small and boring - knowing your numbers cold, and keeping a human hand on the wheel. Build on that, and you can take or leave whatever gets announced from the podium next week.

The Weekly Grind is published every Monday by Brikly - modular intelligence tools for independent cafe and coffee shop owners. Got a story we should cover? Get in touch ->

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