The Inferred Blog
Plain-English briefings on business software and AI
When a change in software, AI or security matters to a small business, we explain what happened, what it changes and what is actually worth doing about it.
Written in HorshamUseful to businesses anywhere
Sources linkedClaims you can inspect for yourself
No panic buyingAction only where the evidence supports it
Recent briefings
The signal, without the performance
Reporting and practical interpretation for owners and operations teams deciding what AI means for their business.
Before AI answers “is it in stock?”, your system needs to know.
AI can help answer questions about stock and orders. A useful answer starts with reservations, damaged items and current records. What this week’s UiPath and Snowflake announcement means for a smaller business.
Read the briefing →Barclays is using AI to sort emails. Start with one handover.
Barclays is using AI to help process incoming emails. For a smaller business, the useful question is what happens to the next enquiry: who checks it, what gets copied, and whether the next person has enough information to act.
Read the briefing →Your Access database won't stop working on 13 October. That's the problem.
Office 2021 and Access LTSC 2021 lose Microsoft support on 13 October. Your database may open the next morning as if nothing happened, which is why this is easy to ignore. What the date changes, and when to replace the database rather than just upgrade Office.
Read the briefing →Notes from International Security Expo 2026: the perimeter is the whole conversation
We spent a day at International Security Expo and International Cyber Expo at Olympia. Almost every stand was about deciding where the boundary is and watching it, and many now carry an AI layer that was not there before. Two plain questions to ask any supplier about it.
Read the briefing →The AI can see your business data now. Start with permissions.
The next AI risk is not a chatbot saying the wrong thing. It is a useful tool being given access to the right data and the wrong authority. Before you connect an AI to your inbox, files or accounts system, check what it can read, what it can change and who can see what it did.
Read the briefing →The Real Cost of the Microsoft 365 Packaging Update
Microsoft folded a year of AI and security features into new M365 suites, and renewals will carry bigger bills. The trick isn't just the price hike: Copilot costs scale with headcount, not usage, locking businesses into a recurring expense with no ceiling. A plain-English look at what to check before you sign for another term.
Read the briefing →Google just put another AI tool inside your business
Google Pics is rolling into Workspace as a new AI image creation and editing tool. Useful, certainly — but it is also a good example of how shadow AI can arrive through software a business has already approved. Five checks to make before people start feeding it company material.
Read the briefing →Three AI services switch off this month — and nobody emailed you
Three AI services stop working before the end of August and a much bigger group follows in October. The names mean nothing to most businesses, which is the problem — they are the engines underneath tools sold under friendlier names. One email to whoever built yours.
Read the briefing →AI just got 80% cheaper — but probably not for you
OpenAI cut one of its model prices by 80% three weeks after launching it. That is a genuine cut — to the metered wholesale rate a developer pays, not to the per-seat licence most firms actually buy. What changed, what didn’t, and when renting is still the right call.
Read the briefing →The AI didn’t escape. The fence wasn’t there.
Anthropic says three Claude models broke into the real systems of three companies during safety testing. The headlines went with “AI escapes”. What actually happened is duller, and far more useful to a business about to be sold an AI that acts rather than answers.
Read the briefing →The AI deadline on Sunday — and why you probably think it was cancelled
Europe's AI rules were delayed — except the transparency layer, which starts on 2 August. What Article 50 asks for, whether a UK business is caught by it, and the short piece of admin that answers it.
Read the briefing →The rogue AI had an owner — and a detail nearly everyone missed
We now know whose models broke into Hugging Face. But the part worth your time is what happened next: the defenders' own AI refused to help them clean up. What that does and doesn't mean for a normal business.
Read the briefing →The AI chats that turned up on Google — what actually happened
Shared AI conversations became findable in search results this weekend. Not a hack — a share button. What it means for anyone whose team uses AI, and the one line your policy is probably missing.
Read the briefing →The AI that broke out of its box — what actually happened
An AI “went rogue” and hacked another company. It isn't science fiction — but it isn't quite what the headlines suggest, either. The plain-English version, and the one question worth asking about any AI system.
Read the briefing →Occasional email
Get the useful ones in your inbox
No more than monthly. Unsubscribe at any time, or use RSS without sharing an email address.
Used only for the briefing. Nothing else, and never shared.