Law, accounting, clinical, financial advice

The judgement is billable.The paperwork around it is not.

In regulated work the constraint is evidence: who decided, on what basis, and can you show it two years later. That rules out most of what people mean by AI — and leaves a surprising amount that it does not.

Three days on site · three days of analysis · the report is yours either way
12

Open source projects

Public on github.com/JIGGAI

4

AI plugins

Published on npm, free to install

65.4K

AI plugin installs

65,377 to date, per npm

2.8K

AI monthly plugin installs

2,811 in August 2026, per npm

Sound familiar

What we hear in the first hour.

Not a generic list. These are the ones that come up in businesses shaped like yours.

01Fee earners

Fee earners spend their afternoons on file notes rather than files.

The record has to exist, but it does not have to be typed by the person who is billed at the highest rate.

02Intake

Intake takes a week because three people have to check the same things.

Conflict checks, identity, and engagement terms are rules, not judgement. Rules automate cleanly and leave a better trail than people do.

03Compliance

We cannot use AI. We are regulated.

You cannot use it to decide. You can use it to prepare, to check completeness, and to make the audit trail better than the current one.

04Repeatability

Every matter is bespoke, so nothing is repeatable.

The advice is bespoke. The steps around it — opening, evidencing, closing, archiving — are close to identical, and nobody has looked at them in years.

05Approval

Compliance says no before anyone has said what the thing is.

Usually because previous proposals could not answer what data went where. That is a scoping problem, and it is answerable in writing.

The rule

Regulation does not forbid automation. It forbids automation you cannot account for.

Which is the same standard we would apply anyway. Everything we recommend leaves a record of what it did and why — that is what makes it defensible, and it is also what makes it worth having.

What we expect to find

The software already in the building.

We do not arrive with a platform to sell you. We arrive expecting these, and the work is usually in the gaps between them.

01

Practice management

Clio, Actionstep, Karbon, IRIS. The system of record, and usually the one people work around rather than in.

02

Document management

iManage, NetDocuments, SharePoint. Where the evidence lives, and where naming conventions go to die.

03

Time and billing

Where under-recording quietly costs more than any inefficiency we are likely to find elsewhere.

04

Identity and conflict checking

Often a subscription service plus a spreadsheet plus somebody remembering. The spreadsheet is the risk.

05

The client portal, if there is one

Frequently bypassed by email, which is where the compliant record stops being the real one.

06

Retention and archive

What has to be kept, for how long, and whether anyone could actually retrieve it under pressure.

A worked example

Taking the file note off the fee earner.

Roughly what one entry in your report looks like — a real shape, with the numbers changed.

What we saw
Senior staff writing up attendance notes from memory at the end of the day, long after the conversation, in a format that varied by author and was never checked for completeness.
What it costs today
Fee-earner time at the highest charge-out rate in the building, producing a record whose quality depended on how tired the author was.
What we would build
A drafting step that takes the practitioner’s own notes and produces the file note in the house format, with the required fields flagged when missing. The practitioner reviews and signs; nothing is filed without that signature, and the draft and the final are both retained.
How you would know it worked
Time between the attendance and the note being signed, the share of notes missing a required field, and recorded time recovered. If completeness does not improve, the format is wrong rather than the tool.
And what we would not automate

The advice, obviously — but also the decision about what goes in the note. What a practitioner chooses to record is itself a professional judgement, sometimes a defensive one, and a system that decided for them would be producing evidence nobody stands behind.

The arithmetic

What coordination work costs you.

Three numbers you already half-know. Move them until they look like your business.

12

re-keying, chasing status, producing the same document again

6h

on that work specifically, not their whole job

$65

salary, tax, benefits, desk

Coordination work, per year
3,312hours
Plausibly retired by agentic automation
2,484hours

A 75% capture rate. The other twenty-five per cent is judgement, exceptions, and not wanting to look greedy.

1.4
FTE equivalent
$161,460
Per year

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 →
What happens

What the six days look like.

The most common question we get is not about AI. It is what these people will actually do in my building.

Assessment · 3 days on site · 3 days analysisNo prep required
  1. Day 1Walk the floorWhoever is on shift

    We start where the work happens, not in a meeting room. Nobody prepares anything, and the first day is mostly watching.

  2. Day 2Sit with the people doing itOps, admin, finance

    Conversations with the roles that touch the work most, and a real task followed end to end — including the parts that happen in a group chat.

  3. Day 3Systems, then a read-backWhoever holds the logins

    What you run, what talks to what, and where a person is currently the integration. We tell you what we saw before we leave, while it is still cheap to correct.

  4. Days 4–6Research and discoveryOur desks, not yours

    Away from your building. We cost the work we watched, model the alternatives, and test the shortlist against your own numbers rather than a framework.

  5. +1 weekOne recommendationPresented in person

    The single change worth making first, specified precisely enough to build — with the ranked analysis behind it and the list of what we would not automate.

Before you ask

The questions this raises.

Where does our client data go?

That is the first question the report answers, per recommendation, in writing. Options that would send client data to a third-party model are identified as such so you can rule them out — or approve them deliberately. Several of the recommendations we make in this sector run entirely on your own infrastructure.

Our regulator has published guidance. Do you work to it?

We read it before the visit and write the report against it. We are not compliance advisers and we do not pretend to sign anything off, but a recommendation that ignores your regulator is not a recommendation, it is a liability.

What about privilege and confidentiality?

We sign your engagement terms or NDA before the visit, we do not read matter files, and the systems inventory is structural rather than substantive. We need to know what runs where, not what it says.

Our partners will not change how they work.

Then the recommendations should not require them to. The best findings in this sector remove steps from support staff and leave the fee earner’s day nearly untouched — which is also why they actually get adopted.

Start with half an hour.

We'll tell you whether an assessment would pay for itself in an operation like yours. Sometimes the answer is no.