Blog
What we are building, what broke, and what we would do differently. Written by the people doing the work.
Most restaurants do not have a software problem — they have five, and none of them talk to each other. The story of building a single restaurant operating system, now serving 20,000+ customers across 20+ UK restaurants.
The ranges in every custom software cost guide are real and useless. Here is how the number actually gets built — and the contract clause that matters more than the price.
A channel manager has never persuaded anyone to book anywhere. What each system in your stack actually does — and six symptoms that tell you which box is broken.
Anyone can build the ordering site in a fortnight. Whether a direct channel survives comes down to six decisions before it — allergens, payments, drivers, the kitchen, delivery pricing and who owns the customer list — and they come in an order.
Ask a hotelier what they pay Booking.com and most say 15%. Here is what actually leaves the account — and the arithmetic that decides whether a direct booking channel is worth building.
Just Eat and Uber Eats publish their rates; Deliveroo does not. Here is what UK restaurants and takeaways actually pay once VAT and the fees underneath are counted — and how to work out whether your own ordering channel is worth building.
54% of UK SMEs now use AI, yet MIT found 95% of GenAI projects return nothing measurable. Here is what actually pays for itself in small business software, what doesn't, and how to test one use case in 30 days.
Half of enterprises have shipped an AI agent that passed every internal evaluation and then failed in front of a customer. The tests weren't skipped - they were the wrong shape.
Every vendor selection guide tells you to look for a proven track record and deep industry expertise. Neither is something you can check before you sign — which is precisely why both survive. Here are five questions that fail properly.
Humans are extraordinary middleware. They reconcile broken data silently, without a ticket. Automating the workflow removes them from exactly that position — and the data stays broken.
Enterprise agent sprawl is real research aimed at the wrong reader. At three agents, the risk runs the other way: one component, six jobs, no boundary. What actually protects a small portfolio — and the ten-field register that does most of the work.
