The Weekly Grind: AI & Tech News for Cafe Owners - 24 August 2026
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 number underneath the headline. Barclays' card data says the line that includes cafes grew in July; the transaction count underneath says there were fewer transactions, not more, and the CPIH rate Barclays prints alongside it says that growth was not growth. The ONS says food inflation has fallen to 1.3%; the same table says restaurants and hotels are still rising at 4.0%, and your customers only read the first line. A forecaster says business energy costs are up 25%; its footnote says that is a site using 2,334MWh of electricity and 15,275MWh of gas a year, and the fact that does transfer is a near four-year high in wholesale prices with an October renewal coming. A robot company reports $3.9m of revenue for nine months; the note says $861,000 of it was robots sold. And a trade headline says a salmonella outbreak has been traced to hospitality; the health agency's own report says it is still tracing, and adds a line about two of the venues that the headline left out.
Restaurants, Cafes and Bakeries Took 2.1% More From 0.3% Fewer Transactions in July - Below Inflation, While the Pub Next Door Took 7.4% More on 10% More Transactions
On Tuesday 11 August Barclays published its Consumer Spend Report for July; we are working from the retail trade site Insight DIY, which reproduced the release and its full category table, credited to Barclays. Card spending across Barclays' card base - one bank's cardholders, not the whole economy - grew 2.0% year on year in July, after 1.9% in June - which is all "highest for 12 months" means here, one tenth of a point - and, in Barclays' own words, "remained below the latest CPIH inflation rate of 2.8 per cent". Essential spending grew 2.9% and non-essential 1.6%, with England's World Cup run, the July heatwave and recovering consumer confidence all credited. Rohan Kumar, Head of Spend Insights at Barclays: "Card spending strengthened further in July, reaching its highest level for 12 months, with both essential and non-essential spend remaining in growth."
The table is where the story is, because Barclays prints two columns for every category, spend growth and transaction growth, and the eating-out rows point in different directions. Eating and drinking as a whole: spend up 3.1%, transactions up 0.1%. Inside it, restaurants, cafes and bakeries - one pooled line, so this is not cafes alone - spend up 2.1%, transactions down 0.3%. Bars, pubs and clubs: spend up 7.4%, which Barclays says is its highest since December 2023, on 10.0% more transactions, credited to the World Cup. Takeaways and fast food: spend up 1.8%, transactions down 5.0%. So the whole eating-out category was flat on volume, the pubs carried it, and the line that includes cafes took slightly more money from slightly fewer transactions, with all of the increase, and then some, coming from what each transaction was worth - roughly 2.4% more per transaction, on our arithmetic. Against the CPIH rate Barclays used, 2.8%, that 2.1% is a real-terms fall. Confidence is improving from a low base: GfK's August barometer, published on Friday 21 August, rose three points to -14, its highest for two years, and still a minus. And one more number from the same Barclays release, the consumer-AI story of the week: 48% of UK adults have now used AI for planning or budgeting, up from 35% in July 2025, and 78% of Gen Z. Nearly half the people deciding whether the £4 flat white is in this month's budget now use AI somewhere in their planning.
What this means for you: Do the five-minute version on your own July. Our break-even guide already tells you to pull last month's takings and transaction count from the till; put July 2026 next to July 2025 and you have the two columns Barclays prints. If takings rose about 2% on flat or falling transactions, you matched the sector - and, against 2.8% inflation, went backwards. That is not a comfortable reading, but it separates two problems that feel the same from behind the counter. Fewer transactions is footfall, and the average-spend guide is right that footfall is largely out of your hands; July is the national version of that claim, and Barclays credits the World Cup for the pubs taking 10% more transactions in the same month. More per transaction is the lever you do hold, and that guide's point is that attach rate and add-ons move it without a price rise. The price rise is the other half: the pricing guide's premise that raising prices is not optional for most independents in 2026 is now in Barclays' own comparison, because a rise that merely matches inflation is standing still. And the AI number is a reminder that the customer weighing that £4 is, increasingly, planning with help.
Read the Barclays July report as reproduced by Insight DIY ->
Read GfK's August consumer confidence barometer on NIQ ->
Wholesale Energy Is at a Near Four-Year High and October Is a Big Renewal Window - and a Cafe Has No Price Cap
On Thursday 20 August the energy analysts Cornwall Insight published an update to their Business Energy Cost Forecast, and the headline number is one you should not use. Business energy costs, it says, have risen 25% since February, "as the conflict in the Middle East continues to push wholesale prices higher": an annual electricity contract now costing £638,500, gas for the same site £1.15m. Read the footnote first. The 25% is measured on Cornwall Insight's archetypal "small industrial or commercial" site, which by its own definition uses 2,334MWh of electricity and 15,275MWh of gas a year. That is, in Cornwall Insight's own words, a larger retail and leisure site or a small manufacturer, not a cafe, and on our reading the percentage does not transfer either: a bill that size is dominated by the wholesale commodity, while our energy guide puts network charges, levies and standing costs at roughly 40-50% of a typical business electricity bill, so the same wholesale move lands as a smaller percentage of a small site's all-in bill. Nobody has published the cafe-scale version, and we are not going to invent one.
What does transfer is the market. "The conflict in the Middle East has seen wholesale prices rise to their highest level in almost four years, with a lasting ceasefire still not in sight," the release says, and then the line that matters: "Many businesses have contracts renewing in October, and our forecasts suggest they'll be locking in at a worse rate than they'd hoped." Jacob Briggs, Cornwall Insight's Energy Users Lead: "Gas prices have been climbing over the past six months, largely driven by the ongoing conflict between the US and Iran, and those costs are landing on business energy bills." Government support is narrow - two schemes, around 500 energy-intensive businesses now and around 10,000 in the Industrial Strategy's eight growth sectors from April 2027 - and "nearly 90% of business energy consumption is from businesses that have seen no support at all on their policy costs". Households have a price cap; a cafe does not, and Cornwall Insight said so plainly when it ran this warning in March: "Unlike households, businesses have no price cap to insulate them from volatility in wholesale energy costs." That March release matters because this is an update rather than an event: it warned then that electricity bills had risen 10% to 30% since late February, into what it called one of the busiest renewal periods of the year, and August's news is a firmer number and a later window. The Climate Change Levy, the levy line on a business energy bill, rose on 1 April from 0.775p to 0.801p a kilowatt-hour, and GOV.UK already publishes 0.827p for April 2027; that is 0.026p a unit, which on a cafe using 50,000kWh a year is about £13 on our arithmetic, which is to say noise. The wholesale line is the story.
What this means for you: If your contract ends between now and Christmas, you are about to be quoted off the dearest wholesale market since 2022, and it lands on top of the September squeeze - rent quarter, VAT bill, post-summer dip - that our guide told you to plan for in August. Our energy guide covers the half you can act on any week, using less and buying smarter, and already tells you to set a reminder eight to twelve weeks before the end date, to get quotes from at least three business energy brokers, and to be wary of a broker who will not show you the commission. Two things it does not cover, because the market has turned since it was written. It was written in April on the premise that wholesale prices had come down from the 2022 peak; they are still below that peak, but they have since risen to a near four-year high, and we have added a note to it this week, because that is not the market you are renewing into. And it says nothing about how long to fix for, which this autumn is the decision. A long fix buys certainty at a price that contains a conflict premium, the part most likely to unwind; a short fix buys time at a high price and puts you back in the market next year. Neither is free, and the mistake is not deciding: Briggs's own advice is to plan ahead, "whether that's locking in contracts early or looking at on-site generation". For cafe scale, use the only cafe-scale energy number we have: the flat white P&L prices energy at 40p a kilowatt-hour, 3p to 5p a cup for the machine and grinder and 10p to 12p once lights, fridges, dishwasher and heating are in. On those numbers a 10% move in your unit rate is about a penny a cup - not a crisis, but a known step that a 13-week cash flow absorbs if you put it in now and does not if it arrives as a surprise in November.
Read the August forecast on Cornwall Insight ->
Read Cornwall Insight's March warning, including the price cap line ->
Read the Climate Change Levy rates on GOV.UK ->
207 Cases, One Death, and Eggs the Strongest Signal So Far - and Possible Cross-Contamination Risk Events Reported in Two of the Five Venues
On Tuesday 18 August the UK Health Security Agency's Health Protection Report carried a note on a national outbreak of Salmonella Enteritidis, and on Wednesday 19 August The Caterer ran it under the headline that it had been "traced to hospitality businesses". The report is more careful than the headline, and the careful version is more useful. As of 13 August there were 207 confirmed cases between 11 August 2025 and 1 August 2026, all in one genetic cluster "indicating a common source of contamination"; 206 of the 207 were reported in 2026, and UKHSA first spotted the rise on 26 May through whole genome sequencing. One death is associated with the outbreak. Most people had a self-limiting stomach illness, but 38% of those with information available were admitted to hospital and two had bloodstream infections. Only 6 of the 207, or 3%, reported travel outside the UK, so UKHSA does not think this is a travel-related outbreak.
Here is what the investigation has actually found. Of 114 people interviewed, 95, or 83%, had eaten food prepared outside the home in the week before they fell ill. Interviews identified five food businesses that two or more unconnected cases had visited, linking 24 cases, and 11 of those 24 reported eating dishes made with eggs. "The signal for egg consumption remains the strongest in this investigation to date," UKHSA says; the five premises are "supplied by eggs from multiple sources, including egg importers", and "at this time there is no identified link to UK-produced eggs or poultry". But the source is not identified - "extensive epidemiological and food chain investigations are ongoing" - and the line the headline left out is this: "Local investigations have also identified possible cross-contamination risk events within two of these settings." Eggs are the strongest signal so far; in two of the five venues, local investigators also identified what UKHSA calls possible cross-contamination risk events, inside the premises rather than upstream of them. It is not the first: UKHSA says the strain is genetically linked, within 25 SNPs, to a 2025 outbreak that "was also linked to imported eggs" and stood at 123 confirmed cases as at 11 November 2025, in which UKHSA said at the time it had identified a number of food businesses, several of them "linked to the same distributor of imported eggs". None of the businesses is named, and UKHSA says "food businesses" throughout, not cafes.
Four days before the report, on Friday 14 August, the Food Standards Agency published a new page of egg guidance written specifically for caterers. It is guidance rather than law, and it is the clearest one-page version of the FSA's egg advice - published, as it happens, while the outbreak investigation was live. Buy from suppliers "who can demonstrate the traceability and provenance of their products"; UK hen eggs are best practice; British Lion and Laid in Britain eggs present "a very low risk of Salmonella", which the page is careful to say "is different to the risk being 'negligible'", and "no egg can be guaranteed to be free from Salmonella, whatever the source or brand". The fallback is one sentence: "If you cannot verify that eggs are produced under either assurance scheme, you should ensure that egg dishes are thoroughly cooked or prepared using pasteurised egg products." It names the dishes needing most care - runny yolks, fresh mayonnaise, mousses and souffles - and asks for a spot check on delivery, rejecting damaged or dirty eggs. The NFU's poultry board chair Will Raw, quoted by The Caterer, said the outbreak "underlines exactly why sourcing British matters", which is a farming lobby making the case it exists to make; the FSA's version is the one to work from.
What this means for you: Keep the proportions in view: 24 of the 207 cases link to the five businesses, 11 of those 24 reported an egg dish, and the source is still open. This is reporting, not a reason to take the eggs Benedict off. It is a reason to do two free things this week. First, look at the egg line on your last invoice and see whether it says Lion or Laid in Britain; if it says nothing, ask, because the FSA's advice is that an egg you cannot verify as British Lion or Laid in Britain is one to cook through, or to replace with pasteurised egg in anything raw or lightly cooked. Our goods-in guide is a sixty-second back-door routine for matching what arrived against your order and the delivery note; the FSA's page tells you exactly what to look for on the egg tray while you are already standing there - reject any that are damaged, dirty or not to the required standard. Second, and this is the half you control, walk the step after the delivery: where the eggs sit, what cracks them, what wipes the surface, what touches the salad after. Local investigators identified possible cross-contamination risk events inside two of the five implicated venues; eggs are the strongest signal so far, and that step is the one your kitchen owns. The brunch-rush post already put a number on your exposure, 624 eggs across a weekend at 2.4 a plate on its worked example; you know the volume, and this week's question is whose they are and what they touch. Our hygiene-rating guide makes the point that inspectors score confidence in management rather than a single fridge, and being able to show the egg spec on your invoice and the handling step in your kitchen is exactly what that looks like.
Read the outbreak report on GOV.UK ->
Read the FSA's egg guidance for caterers on GOV.UK ->
Read UKHSA's November 2025 report on the earlier outbreak ->
Read The Caterer's coverage ->
The Robot Company With $302m in the Bank Sold $861,000 of Robots in Nine Months - and Made More Renting Them Out for Events
On Friday 14 August Richtech Robotics, the Nasdaq-listed Las Vegas company behind the ADAM robotic drinks system and, through a subsidiary, its own robot-operated restaurant sites, told the SEC it could not file its quarterly report on time. On Wednesday 19 August it filed it, for the nine months to 30 June, and the accounts were worth the wait. Total revenue was $3.9m, up 9.6% on the same period last year; that is the headline number. The note underneath breaks it into four lines. Product sales - robotic systems sold outright, with their installation - were $861,000, down 45.1% from $1.57m, which the company calls a deliberate shift to subscription. Robotics-as-a-Service, the subscription, was $1.08m, up 189%. "Others", the experience store, repairs and data services, was $484,000. And the largest line, at $1.52m, was "event services", which the filing defines as "Short-term rental of robotic equipment for specific events", on contracts that "typically have a duration of one to three days". The single biggest thing a listed hospitality-robotics company did in nine months was hire its robots out for the weekend. Against that, cash of $302.0m, up from $185.6m at the start of the financial year - $110.3m of it came from issuing shares in the period, and the company credits "the successful execution of our capital-raising initiatives" for the rise in its equity - and total liabilities of just $4.1m. The nine-month loss was $20.4m, narrower than the $31.8m a year earlier - though the loss from operations nearly tripled, from $10.8m to $30.7m, and the filing credits the narrowing to $9.8m of investment income on the cash pile and a swing in the value of its warrants from a $22.1m loss to a $0.6m gain. This is not a company in trouble. It is a company that has raised a great deal more money than it has sold product, which is a different fact and the one that matters to a customer.
Then the write-off. In the June quarter Richtech impaired $9.5m of long-lived assets, leaving its intangible assets at $371,000 against $10.7m at the start of the financial year, and the reason is in the notes: "In January 2026, we launched a strategic transformation initiative referred to internally as 'AI Across All' ('AAA'), under which we are transitioning our operations, software platforms, and technology infrastructure to an AI-native architecture," and consequently "we determined that our existing software platforms would have no future economic benefit upon completion of the transition to the AI-native environment". The charge was non-cash and "did not impact our operating cash flows". Read that as a customer rather than a shareholder: within six months of launching the plan, a vendor with $302m in the bank wrote the software platforms its business had been running on down to nothing, and says it completed the transition of "substantially all affected software platforms" by 30 June, with those legacy platforms "fully replaced by AI-native systems". The subscriptions on its RaaS side are, per the same filing, "typically structured with an initial non-cancelable term ranging from 12 to 36 months", with "an automatic renewal provision" that rolls the contract on unless either side gives written notice at least 30 days before the term ends. One more detail for the diligence file: the report for the quarter to March was filed on 7 August, twelve days before this one. And for contrast, the other automation number of the week: on Thursday 20 August a Miso Robotics press release carried by QSR Web said "multiple operators who automated their fry stations are already seeing $6,400 or more in new monthly profits per location according to Miso Robotics", with a return on investment that "can land in under three months" and, from the only operator named anywhere in it, a White Castle regional director's LinkedIn post about a busy lunch shift that says nothing about profit. What the release does not contain, anywhere, is a price, a subscription cost, a site count, or which operators saw the $6,400.
What this means for you: Our AI guide told you in April that a robot barista costs more than your annual turnover and that automation suits high-volume, low-personalisation sites, which is not what an independent is. This filing is the receipt: a listed vendor sold less than a million dollars of the actual product in nine months and made more hiring them out for events. The new lesson is not whether to buy a robot but what to read before you subscribe to anything that runs on somebody else's cloud, and it applies to the till and the ordering screen as much as to the arm. Our EPOS guide warns that the wrong choice locks you in for years; this filing adds the case where the vendor writes off and replaces substantially all of its own software platforms while its subscribers sit on 12-to-36-month non-cancellable terms. The lease-or-buy guide covers how to pay for physical kit, outright, on hire purchase or on a lease; it does not cover kit whose software the vendor can retire, and that is the term sheet to read twice, because as the downtime post puts it an outage is lost turnover, not a repair bill, whether the cause is a group head or a platform decision. Two habits, then. For any vendor asking for a multi-year subscription, read the accounts - Companies House for a UK one, the SEC's site for a US one - and if product revenue is a fraction of the money raised, price the exit before you sign. And for every autumn pitch that leads with the profit it will make you, ask for the number it left out. Miso's release is the template: the figure you are given is the return, the figure you are not given is the cost, and the second decides whether the first is true.
Read the Richtech Robotics 10-Q on the SEC's site ->
Read the Miso Robotics release on QSR Web ->
Shelf Food Inflation Fell to 1.3% - Restaurants and Hotels Are Still at 4.0%, and Your Customers Only Read the First Number
On Wednesday 19 August the ONS published July's inflation figures. The headline went the wrong way: CPI rose 2.9% in the 12 months to July, up from 2.6%, "the first time since March 2026 that the 12-month rate had increased". Food went the other way again. Food and non-alcoholic beverage prices rose 1.3% in the year to July, down from 1.7% in June, and in the ONS's words, "The annual rate was last lower in September 2021, when it was 0.8%". That is the fourth fall in a row - 3.0% in April, 2.2% in May, 1.7% in June, 1.3% in July - and on the month prices were "little changed". The retailers' trade body, the British Retail Consortium, through its lead economist Harvir Dhillon: "Retailers have competed fiercely and managed to keep the cost of food and essentials down for shoppers, but the rising cost of government policies puts this in jeopardy" - pasta, olive oil and fresh fruit all fell on the month. IGD's chief economist James Walton, writing later in the week, called the fall a surprise given "energy market disruption and extreme weather", and, among the six contributing factors he lists, offered the one that matters here: "Retailers appear to be absorbing some cost increases, while plentiful supply earlier in the year may have provided a temporary cushion."
Now the other row in the same table. Restaurants and hotels - the ONS division covering the price of eating and drinking out plus hotel rooms - rose 4.0% in the year to July, down from 4.4% in June, and was flat on the month. So the number your customer read on Wednesday's news is 1.3%, the number for the division that includes the menu they are looking at is 4.0%, and the gap is 2.7 points. It is worth squaring this with a post of ours that several of our other guides link to: in April we ran the Food and Drink Federation's forecast that food inflation could reach 9% by the end of the year. Both are still live, because they are the same index at different dates: the FDF, the food manufacturers' trade body, was forecasting food and non-alcoholic drink inflation out to December 2026, and 1.3% is July's outturn. One of Walton's readings is that retailers are absorbing some of the cost increases in between. The forecast is still the number that shapes your supplier's next letter. July's print is the one that shapes your customer's mood.
What this means for you: Thursday's staleness audit quoted the June figure and made the argument we will not repeat: headline food inflation is a supermarket shelf index built for households, not your wholesale account, and it averages your rises away against someone else's falls. The number moved again the day before that post went out - we have added the July figure to it - which is rather the point. The live problem this week is on the other side of the counter. The next 20p on the flat white is now arguing against a headline, and the pricing guide's rule not to blame suppliers in front of customers gets a second clause: do not argue the index either. What you can say is true and specific: the ONS's restaurants and hotels index is rising at 4.0% nationally and the shelf price of food at 1.3%, and the difference is not mostly the beans. The ONS does not break that gap down, but the costs that moved for operators this year are the employer NI rise, the April wage floor, rent, and this week's energy story. But you can only make that case if you know what your own food cost did - and the honest converse is that if pasta, olive oil and fruit fell on the shelf, some line on your invoices may have moved too, a margin-recovery window you can only see with current costs. Recost the dozen dishes that matter, with the free recipe costing tool if you have nothing better, and walk into the price conversation with your number rather than the ONS's.
Read the July inflation bulletin on the ONS ->
Read IGD on why food inflation is falling ->
The Brikly Take
Five stories, and in every one the useful number was underneath the headline. Growth of 2.1% that was a fall once you read the inflation line. Food inflation of 1.3% that describes the supermarket, in the same table as the 4.0% for restaurants and hotels that describes the row you trade in. A 25% energy rise measured on a site that uses 2,334MWh of electricity a year, with a footnote that says so. Nine months of revenue at $3.9m, of which $861,000 was robots sold outright. An outbreak "traced to hospitality" that UKHSA is still tracing, with a line about possible cross-contamination risk events in two venues that the headline dropped. None of it is hidden; it is all in the release, the table, the note or the footnote. The chains have someone whose job is to read the second line. The independent reads the headline on a phone between orders, and the gap between the two is where the bad decision gets made: a price rise that stands still, a long fix at a conflict premium, a subscription signed with a vendor that has just written its own software platform down to nothing and replaced it, an egg dish whose provenance you cannot state when someone asks.
The headlines will keep coming, and next Monday will bring five more. The habit worth building is the one every story this week rewarded: find the second number, and price from that.
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.