We Build AIthat pays for itself.
We spend three days inside your business, find the work that should not be done by a person, and hand you the plan to fix it.
Re-keyed by hand into two other systems
Correct, and bypassed whenever something is urgent
Undocumented, load-bearing, maintained by one person
Chased manually because nothing tracks who is waiting
The actual escalation path, written down nowhere
Three weeks stale by the time anyone decides from it
Open source projects
Public on github.com/JIGGAI
AI plugins
Published on npm, free to install
AI plugin installs
66,095 to date, per npm
AI monthly plugin installs
2,811 in August 2026, per npm
The things we hear in the first hour.
Every assessment starts with a version of one of these. None of them are AI problems yet — they are process problems that AI happens to be good at.
Two of my systems don’t talk, so someone re-types everything between them.
The cheapest kind of finding. It shows up on the first walk-through and has an obvious owner.
Month-end eats a person for two days and nobody can say exactly why.
Usually three or four small jobs wearing one trench coat. They come apart cleanly once separated.
We tried an AI pilot. Everyone was impressed. Nobody uses it now.
The pilot was never scoped against a process, so there was no way to tell whether it worked.
Status updates are a full-time job that nobody was hired to do.
Coordination work — the most automatable category, and the least visible on any org chart.
I can’t tell which of this would even benefit from AI, so we’ve started nothing.
The normal position. Deciding what not to automate is most of the value of an assessment.
AI is not the bottleneck. Deciding what to do with it is.
Which is why we sell analysis before we sell software — and why the analysis is useful even if you never buy the software.
The work nobody invoices you for is the work eating your margin.
Coordination is not a small tax on the real work — in most operations of this size it is a full-time job spread thinly across everybody, and it is paid for every week whether or not anyone notices.
It is already paid for
Re-keying, chasing and reformatting are done by people you hired for something else, out of salaries you already committed to. There is no invoice to query and no line to cut, so it survives every review that a piece of software would not.
The spreadsheet became infrastructure
Something was missing, so somebody built the gap out of a spreadsheet. It worked, so it stayed. Now three teams depend on a file one person maintains by hand, and nobody replaces it because nobody owns it.
It scales with the business
More work means more of the same coordination, so the answer is another hire. The admin grows in step with revenue instead of staying fixed, which is why the margin on your newest customers is thinner than on your oldest.
None of it shows up anywhere you would look for it, because measuring it means adding up hours nobody logs. So add them up.
What coordination work costs you.
Three numbers you already half-know. Move them until they look like your business.
re-keying, chasing status, producing the same document again
on that work specifically, not their whole job
salary, tax, benefits, desk
A 75% capture rate. The other twenty-five per cent is judgement, exceptions, and not wanting to look greedy.
This is arithmetic, not a finding. It rests on three numbers you guessed. The assessment replaces all three with numbers we observed — and tells you which of those hours are actually worth automating.
See what a real finding looks like →An AI company that consults — not a consultancy with an AI page.
Most firms selling agentic AI have never had to keep an agent running past the demo. We ship the tooling, in public, and other people's teams depend on it. That is where the opinions in your plan come from.
Anyone can tell you what to build. Rather fewer have had to answer for an agent that broke at 2am in month four — which is the difference between a recommendation and an opinion.
Six ways to start.
Not everything begins with an audit. Three agent teams are ready to configure and run in weeks; the rest is built for you or worked out first. Each one is scoped on a free call and quoted in writing before it starts.
Your own AI team, without hiring one.
The part nobody mentions until after the build: somebody has to keep it working, keep it current as the models change, and answer the next question. That is the arrangement, and its size is entirely yours to set.
- A search measured in months, for a role you cannot interview for confidently
- One person’s opinions, and one person’s holiday cover
- Salary before a single thing ships
- A team that has already built this, from the day you start
- Sized by you, changeable whenever, at whatever level you need
- On call for anything AI — including the things we did not build
And you can leave with all of it. The software is yours, in your repositories, and we transfer it cleanly whenever you ask — how the relationship works.
Sorted by how the work moves, not by industry.
What changes an assessment is how many systems a job crosses and who carries it between them — not what you sell. Six shapes, each with the software we expect to find and one finding worked end to end.
See the six shapes →The five questions we always get.
Are you a consultancy or a software company?
Both, and the two are not separable. We sell analysis, and we build the tooling that analysis usually points at. Running our own plugins in production is where the opinions come from — it is the difference between recommending something and having had to keep it working in month four.
If the plugins are free, how do you make money?
Assessments, prototypes and implementation. The plugins are free and public because they are the proof, not the product — they are also the fastest way for you to find out whether we know what we are doing without paying us anything.
Do we have to use your tooling?
No. Plenty of recommendations have nothing to do with our plugins, and some engagements end with us telling you to switch on a module you already pay for. A report that only ever recommended our own software would not be worth paying for.
We already tried AI and it didn’t stick. Why would this be different?
Because pilots usually die of scoping, not technology. Nobody decided which process was worth pointing it at, and nobody defined how you would know it had worked. The assessment produces both of those before anything is built.
What is the smallest way to start?
A free 30-minute call, on which the honest answer is sometimes that an assessment would not pay for itself in a business your size. Failing that, install one of the plugins — it costs nothing and tells you more about how we work than any page here.
Start with three days on-site.
Tell us about your business and we'll come see it. Our assistant will help you write the first message.
Free 30-minute call · no charge, no pitch · quoted in writing before anything is booked