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
- The website your team stopped reading Last month our agent flagged two sentences on one of our own product sites. "Now on iOS and Android." There was no iOS app. The App Store button opened a waitlist. "Join thousands building..." We had counted 15 outside signups. Nobody lied. Both lines were written when they were a plan, and nobody went back. That is how most company websites age. The page is written once, approved once, then only read by visitors. Prices change, a service is dropped, a team member leaves, a "coming soon" never
- The quote request that arrives half-written Most quotes are not slow because they are hard to write. They are slow because the request arrived missing half of what you need. No site address. No quantities. No photo. No deadline. So someone replies asking. The client answers two days later, with one of the four. Someone asks again. The quote itself takes twenty minutes. Getting to the point where you can write it takes a week. This is where most teams point AI at the wrong step. They ask it to draft the quote. The quote was never
- Apollo's chief economist published a blunt note this weekend. Outside the big tech companies, he finds no AI payoff yet. Revenue per worker: flat. Profit margins: flat. And companies are starting to treat AI spend as a cost centre. The easy conclusion for a small business: "If it doesn't pay off for the S&P 500, it won't for us." I read it differently. Revenue per worker stays flat when AI is handed to everyone as a tool. Each person saves a few minutes here and there. The minutes scatter. Nobody's day changes shape, and
- Your team does not quote from the price list. They quote from memory, and correct the price list in their head. Everyone knows the day rate went up in March. Everyone knows the travel surcharge only applies outside the city. Everyone knows the payment terms in the template are the old ones. The file still says otherwise. It has for months. Nobody noticed, because a person was always standing between the file and the client. Now put an AI step in that gap to draft quotes faster. It does exactly what it is told. It reads the file. The
- The price list your AI is quoting from Your team does not quote from the price list. They quote from memory, and correct the price list in their head. Everyone knows the day rate went up in March. Everyone knows the travel surcharge only applies outside the city. Everyone knows the payment terms in the template are the old ones. The file still says otherwise. It has for months. Nobody noticed, because a person was always standing between the file and the client. Now put an AI step in that gap to draft quotes faster. It does exactly
- Stop before you submit" is an instruction. It is not a control. Last week Anthropic published a report on things its own AI models did in testing that nobody intended. One model was told to fill in a web form up to the last step, then stop. Several times, it submitted anyway. It expected a confirmation page that never came. In another test, the model had a list of things it must never do: log in, create accounts, enter personal data, buy anything. Submitting a form was not on the list. It sent an invented tip to a police
- Most quotes don't go cold. They leave without a next step. The usual sequence in a small company: Quote sent. "Let me know if you have any questions." Silence. Someone remembers it two weeks later. "Just checking in." Silence again. That follow-up failed twice. Its timing depended on someone's memory. And when it finally went out, it had nothing to say. The fix happens before you press send, not after. Every quote leaves with three things written down: When the client said they would decide. Who else has to agree. The one point
- Text from ChatGPT used in the EU will soon carry an invisible watermark. OpenAI announced it on 5 October, to meet the AI Act rule that AI-generated text must be machine-detectable. The first reaction in a small company is predictable: "So our clients will know our quotes and emails were drafted with AI." Two facts from OpenAI's own announcement, before anyone worries. The detector is not public. At launch, only approved researchers and expert organisations get access. And the signal is fragile. In OpenAI's tests, replacing a
Something wrong or missing? Send an update. Fixed within 24 hours.