Overview
Presetify helps small teams automate one repetitive workflow with AI, human approval, and integrations with existing tools. The service is designed to implement workflow automation within 14 days.
In the news
- Your automation is allowed to say no. It is not allowed to say no without a reason. Zendesk's 2026 CX Trends report, 6,182 consumers across 22 countries: 95% want to know why an AI made the decision it did. Only 37% of service organisations can supply that reason — to the customer, or to their own staff. The reaction I expect: we have no AI deciding anything about customers. Then look at what already turns people away with nobody involved. The booking form that refuses a date. The postcode field that rejects a delivery address. The
- A European regulator has now classified ChatGPT as a search engine. Not a chatbot. Not an assistant. A Very Large Online Search Engine, in the Digital Services Act's most closely supervised tier. The Commission designated it on 31 August, because it answers users' queries including by searching the web, and because it reaches at least 45 million people a month in the EU. Most small businesses will read that as an OpenAI problem. It is. The obligations land on OpenAI, who have until January 2027. You have no new paperwork. But read
- Some of the emails in your inbox contain letters your screen does not draw. Microsoft published the details last week. There is a block of Unicode characters that most fonts never render. Software still reads them perfectly. In February, a phishing campaign started dropping them inside words. "Funding" went out as "fun", one invisible character, "ding". You see funding. A filter hunting for the word funding sees two fragments of nonsense. The day before, that signature fired on about 21,000 messages. The next day, more than 1.3
- Count the round trips, not the response time The quote took six days. Twenty minutes of it was work. Tuesday, the request lands. No delivery address, so you reply within the hour and ask for it. Good response time. Thursday, they answer. Now you notice there is no quantity. You ask. Monday, they reply. Now the spec is ambiguous. Nobody was slow. There were four emails and one weekend. Everyone measures how fast they answer. What decides whether you win the job is how many round trips it takes to get a complete request. Each
- The new Instagram AI rule is about the person, not the tool Instagram will now limit the reach of accounts that show an AI-generated person without saying so. Announced Monday. Not a fine, not a takedown — you stop being eligible for recommendations, and you find out by opening a status page. Most owners will read that as an AI-influencer problem and scroll on. Read the rule again. The test is not whether you used AI. Instagram states plainly that using AI tools in your creative process needs no label at all. The test is whether a
- Meta's internal code changes went up 220% this year. What reached users went up 36%. Both numbers come from Meta's own internal posts, published by Reuters on 26 August. The plan was to shrink some teams by as much as 60% and let AI agents carry the daily work. The first wave of layoffs went out in May. The second was called off hours before the first one landed. Alongside the volume: major technical and security incidents up 40%, time spent firefighting them up 70%. The reaction I expect: that is a Meta problem. I have fifteen
- Two AI assistants gave away user data last week. Neither one was breached. Security researchers sent themselves a link. One click, and Microsoft Copilot read the connected inbox — full message bodies, including a password sitting in an old email — and sent what it found to a server they controlled. No confirmation, no visible sign anything happened. Microsoft shipped the full fix on 18 August, eight months after the report. Two days later, a second team showed Grok handing over a user's name, location and entire conversation
- The AI Act deadlines just moved. Yours didn't. On 27 July, the AI Omnibus entered into force across the EU. The headline reads as a delay. It is more useful than that. What moved is the high-risk end. Annex III systems now apply from 2 December 2027. High-risk AI built into physical products — machinery, toys, lifts — from 2 August 2028. That is hiring algorithms and safety components. Not your inbox. If you use AI to triage email, prepare a quote or check an invoice before it goes out, you were never on that clock. The transparency
Something wrong or missing? Send an update. Fixed within 24 hours.