Ideas, experiments and useful things
We tried a video toolkit published on GitHub on one idea: a small museum of abandoned projects. With it, we made a carousel, a walk-through, a poster and a looping GIF that all match. But nearly every step needed a workaround. The list of those workarounds is the part your team should keep.
We tried a tool published on GitHub that turns a topic into a short narrated film. We made three films with it. The way it works held up well. Six of the settings it came with did not suit films for social media, so we changed them. Nothing was posted, and its licence needs the author's permission for any commercial use.
I had AI agents turn release notes and approved ad copy into short branded videos. The format worked. But a price counted up from an unapproved $10 before landing on the approved $28, and the checks missed it. Everything was invented, and no human audience test was run.
Some expensive decisions come from a mechanism nobody can see. I had AI agents build four small working drawings where the picture and the sums come from one model. The sums held. Whether people decide better is still untested.
An AI agent can take a whole responsibility off someone's list. It still needs a driver. Five things decide how well it drives, and one habit keeps it on the road.
Help people find answers in business documents without giving everyone access to everything. Start with current sources, permissions and useful tests.
A practical way to sort customer enquiries, gather missing details and prepare reply drafts, with a person checking before anything is sent.
A practical process for creating campaign copy, branded graphics and a product video outline from an approved brief, with clear checks before use.
The specific questions to ask before you sign with an AI agency, plus how to check ROI numbers, headcounts, testimonials, and guarantees yourself.
The 'unlimited automations, one flat fee' pitch reads well until you open the terms. Four mechanics worth checking before you sign a subscription.
The seven clauses that matter most in a UK AI policy, and the risk each one closes: tools, data, training, human review, disclosure, and incidents.
What running product through AI agents changed about the PM job: vision anchors, machine-checkable done-conditions, and gates in place of status meetings.
What happened during an unattended overnight build: a vision anchor, a gate on every commit, and a morning audit of what survived. No demo, just the log.
Making Tax Digital turns record chasing from a January scramble into four deadlines a year. Map the process, automate the repeatable part, check the rest.
Making Tax Digital for Income Tax turns one annual return into four quarterly deadlines a year. The capacity maths for a 300-client practice, in full.
A plain-English timeline of who Making Tax Digital for Income Tax pulls in, when, and what the quarterly rhythm actually requires once you're in scope.
The real 15-minute flow behind our free AI assessment: function mapping, scoring for severity and proof, and what you leave with if you never hire us.
The method we run on every engagement: a vision anchor stating what a project is not, done-conditions a machine can check, and a gate on every commit.
A ten-question checklist covering ownership, the production path, verification and the exit, to run against any AI vendor before you sign, including us.
The exact arithmetic behind our free AI ROI calculator, laid out in full. Plus the three things the formula cannot see, adoption and upkeep among them.
We read the live pricing pages of UK AI automation agencies instead of their marketing claims. The exact numbers, the sources, and the dates.
How AI is quietly changing businesses around the world: saving time, saving money, and opening new room for growth.
