None of it was billed. It was not a decision. Nobody ever sat down and decided to work for free. It arrived one favour at a time, and it turns up at year end as a write-down, next to a realization figure that a partner will describe as disappointing.

What billing reconciliation is

Reconciliation is a check that runs on every invoice, and on the time behind it, before it leaves the firm. It compares what is being billed against two documents the firm already has: the engagement letter and the rate card.

It has two halves that behave very differently, and the difference matters.

The first half is arithmetic and rules, and it is exact. Hours multiplied by the rate for that person’s role. Line totals against the printed subtotal. The total against the subtotal less the retainer. The amount in words against the amount in figures. Dates inside the billing period. The engagement start date. Time recorded in the increments the letter specifies. Third-party costs above the threshold the letter set. People billing who are not timekeepers. Payment terms that do not match the letter. None of that requires judgment. Every one of those has been an invoice that went out wrong at some firm.

The second half is judgment: is this line inside the scope the engagement letter describes? That question cannot be answered by arithmetic, and it is the one that costs the most money.

Here the design matters more than the accuracy. When the system says a line looks out of scope, it must quote the sentence in the engagement letter it relied on. The firm’s own copy of the letter is then checked for that sentence. If the sentence is not there word for word, the finding is downgraded to something a person should look at, rather than presented as a conclusion. A billing dispute built on a misquoted engagement letter is worse than no check at all.

What comes out is a list, not a gate. Each item names the line, the rule or the clause behind it, and the amount at stake. A person decides. The invoice is not held, the partner review is just shorter and better aimed.

Nothing is changed in the accounting or practice management system without someone approving it.

Does a firm your size need it

Start with one number: what did the firm write off last year, and can anyone say why?

Most firms have the first number and not the second. Write-downs are recorded as a single figure per matter, decided at the end, with the reasons living in the memory of whoever made the call. A firm in that position cannot tell scope creep from bad estimating from a client who was always going to be difficult, so it cannot fix any of them.

The strongest signal is fixed-fee and capped work. Hourly billing at least records the extra time even when it is later discounted. Fixed fees absorb it silently, and the firm finds out at the end.

The second signal is disputes that arrive after the invoice rather than before. A dispute is a check the client ran that the firm did not.

The third is realization that varies by partner rather than by matter type. That means the discipline lives in individuals, which means it leaves when they do.

Across the systems we run for firms and service businesses, this check runs at volume: 27,167 service requests taken and 20,871 matched to a provider, with every job’s charge checked against its approved scope before the invoice and before the contractor was paid. The check runs in the seconds before money moves, which is the only moment it is cheap.

What it costs

Reconciliation is built as a fixed-price phase inside a plan. The price depends on how many fee arrangements the firm uses, which systems hold the time and the invoices, and how consistent the engagement letters are. A firm billing hourly against one letter template is a smaller build than one running fixed fees, retainers, contingency and recurring work side by side.

The audit prices it as a phase before anything is built, and credits its own fee against that phase if the firm goes ahead.

Running costs are low. Most of the rules are arithmetic against the firm’s own data, and only the scope judgment needs a model at all.

The cost that is easy to miss is the state of the time entries. A check can only be as good as what it reads, and “attended to matter” tells it nothing. Most firms find that entry quality improves during the first phase without anyone being lectured about it, because people see their own gaps in the same week they made them rather than at the end of the quarter.

What the first week looks like

The first week is spent on invoices that already went out.

We take a quarter of them, with a bias toward the ones that were disputed, discounted or written down, and reconstruct what happened. What was billed, what the letter said, where the two parted company. That exercise usually produces the first real finding of the engagement, which is that the same two or three things go wrong repeatedly.

The rules are written from the firm’s own engagement letters and rate card, not from a generic template. A rule that does not trace back to something the firm wrote down does not go in.

Then the rules run against invoices the firm has already sent, where the outcome is known. Every finding is checked: was this a real leak, or noise? Noise gets tuned out before anyone sees the system in a live billing run. A reconciliation tool that cries wolf in its first month is ignored by its third.

By the end, the check is running ahead of partner review, and the people who prepare invoices have agreed the findings are worth reading. That is the condition for go-live, in this build and every other one.

The full scope is on the service page: /services/billing-pipeline-capacity. To read a fictional firm’s invoice review, then change a rate or a date and watch findings appear and disappear, the demo is here: /demos/scope-check.

It takes most of an afternoon. It happens every month, or every week, forever. And the one person who knows which columns to fix is the reason it happens at all.

What automated reporting is

Reporting on autopilot has two parts, and firms usually want the first one and end up relying on the second.

The first part is the pack itself, compiled on schedule from the systems where the numbers live. Practice management, accounting, time recording, and the ad platforms if the firm spends on acquisition. Same layout every month, delivered where the partners already read it. No exports, no pasting, no retyping.

The connections are read-only. Nothing is written back to the accounting system, and the accountants keep their tools exactly as they are. What changes is that the numbers stop being assembled by hand.

Every figure links back to the rows that produced it. A partner who thinks realization looks wrong clicks the number and sees the matters and time entries behind it, without asking anyone. That single property ends most of the arguments that reports normally start, because the disagreement moves from “is this number right” to “what should we do about it”, which is the conversation the meeting was for.

The second part is asking questions between reports. How is realization on that matter. What is aged over ninety days. Which practice area grew last quarter. Cost per new matter this year against last.

How that answer is produced matters more than that it exists. The question is turned into a query against the firm’s own data, the system runs it, and the system renders the figures. The numbers on the screen are computed from the firm’s rows, not typed out in a sentence by a model. That distinction sounds pedantic until the first time a partner quotes a figure to a client. A model writing prose about numbers will eventually produce a confident, wrong, well-formatted total. A model that writes a query and hands the arithmetic to the database will not.

When an answer needs a caveat, it says so. When the data cannot answer the question, it says that instead of estimating.

Does a firm your size need it

Two questions. How many hours a month go into producing the reports the firm already reads, and who could produce them if that person left tomorrow?

If the answer to the second is “nobody without a lot of digging”, the firm has a continuity problem that has nothing to do with AI and will be solved by the same work.

The strongest signal is that the report is late, or arrives after the decision it was meant to inform. Monthly numbers that land on the fifteenth are history, not management.

The second signal is meetings spent establishing the numbers rather than discussing them. If the first twenty minutes go to reconciling two versions of the same figure, the firm is paying its most expensive people to do data entry out loud.

The third is the spreadsheet that only one person understands. Every firm has one. It usually holds something genuinely important, and it usually has a formula in it that nobody has checked in three years.

The pattern at volume: 160,896 ledger rows parsed from monthly spreadsheet packages into 48 versioned client reports, 51,867 daily ad-spend rows reconciled against the leads they produced, and more than 765,000 app events and 216,000 calls and texts consolidated into one store feeding 19 dashboards from four source systems. That is a reporting layer nobody rebuilds by hand each month.

There is an honest exception. If nobody actually reads the report, do not automate it. Stop making it. That is question one of the audit, and reporting is where it gets the most use.

What it costs

Reporting is built as a fixed-price phase in a plan. The price depends on how many source systems are involved, how clean they are, and how many reports are in scope. One report from two systems is a small build. Client-facing statements plus a partner pack plus practice-level reporting, drawn from five systems with a decade of inconsistent data behind them, is a larger one.

The audit prices it as a phase before any work starts, and credits its own fee against that phase.

Running costs are modest: storage for the firm’s own data, and model usage only for the question-asking layer, which is a small share of the total.

The cost that surprises people is definitions. What counts as realization here. When is a matter open. Which entity does this client belong to when they appear three times under slightly different names. Does a written-off hour count as utilized. These are firm decisions, not technical ones, and they have to be made once and written down. Most firms discover during this phase that two departments have been using the same word for different things, which explains a number of past arguments.

What the first week looks like

The first week rebuilds the report the firm already has, exactly as it is. Not an improved version, not a redesign. The existing one, produced by the system instead of by a person.

That constraint is deliberate. A report the partners already know how to read can be checked figure by figure against the last three months of hand-built versions. Every difference gets explained before anyone is asked to trust it, and about a third of the differences turn out to be errors in the manual version that had been quietly carried forward.

The definitions get written down as the rebuild happens, in the report itself, next to the figures they govern. That document outlives the project.

Once it reconciles, it goes on a schedule and the person who used to build it stops building it. Their Friday afternoon comes back. That time goes to the work only they can do, which is usually reading the report rather than assembling it.

The ask-a-question layer comes after, never before. A firm that cannot yet agree on what realization means will not be helped by being able to ask about it faster.

The full scope is on the service page: /services/firm-numbers. To ask a fictional firm’s ledger about realization, aged receivables or last quarter’s growth, the demo is here: /demos/numbers.

The same question will be asked again in about six weeks. The firm knows the answer. It cannot get to it.

What a firm knowledge base is

A knowledge base is the firm’s own written material, indexed so that a question in plain language comes back with the answer and the document it came from.

The material is what already exists: procedures, checklists, precedent documents, prior memos, filing guides, the notes a partner wrote for a training session, the email explaining why the firm stopped doing something a particular way. Most of it is written already. It is spread across a document management system, a shared drive, three inboxes and one person’s desktop.

Four properties do the work.

It answers only from what the firm put in it. Ask about a filing deadline and the answer comes from the firm’s procedure for that filing, not from what a general model has read on the subject. Where the firm’s practice differs from the textbook, and it usually does, the firm’s practice is what comes back.

Every answer carries its source. The answer is a starting point. The document is the authority. Anyone who needs to rely on it opens the source and reads the paragraph themselves. Nobody is asked to trust a paraphrase.

It says when it does not know. A question the material does not cover returns nothing rather than a plausible guess. That behaviour is the whole system, and it is the first thing worth testing: ask it something the firm has never written down and watch what happens. If it invents an answer, it is not usable in a professional practice.

Access follows the folder. A document nobody outside the tax team can open does not become an answer for the rest of the firm. Permissions come across with the material rather than being rebuilt by hand.

Nothing leaves the firm’s environment, and client files are never used to train anything.

Does a firm your size need it

The test is repetition. How often does someone answer a question they have answered before, and how long does it take the asker to find it alone?

A firm where everyone has been there eight years does not need this. The knowledge is distributed and the answers are quick. The picture changes with hiring. Every new person spends their first months asking questions that have documented answers, and every question costs two people’s attention rather than one.

The second signal is a single point of failure. If one person is the firm’s institutional memory, the firm has a risk it has never priced. That person goes on holiday and work slows. That person leaves and twenty years of practice goes with them.

The third is silent divergence. Two associates producing the same filing differently, because each learned it from a different colleague, is the same problem as the missing memo. The knowledge exists. The routing is broken.

The pattern holds up at volume. In one archive project, 103,382 files on a firm’s drive were indexed and cross-linked to their case files, and 4,517 appearing parties and 2,182 act types were extracted into structured records. That was a back catalogue nobody could search, turned into one that answers questions.

There is one case where the answer is honestly no. If the firm has never written its procedures down, there is nothing to index. That is not a failed project, it is a different first phase: writing down what the firm does. Indexing an empty shelf produces a system that confidently knows nothing.

What it costs

The knowledge base is built as a phase in a plan, at a fixed price agreed before work starts. The number depends on how much material there is, what state it is in, and how many places it lives in. A firm with a document management system and a written procedures manual is a smaller build than one with fifteen years of shared drive.

That is why this work starts with the audit. The audit maps what the firm has, prices the build as a fixed phase, and credits its own fee against that phase if the firm goes ahead. The plan belongs to the firm either way.

Running costs are model usage, which is low for question answering, and storage, which is negligible.

The cost that gets underestimated is curation. Not everything on the drive should be in the knowledge base. Superseded procedures, drafts that were never adopted and the version of a checklist from before the rules changed are all worse than nothing, because the system will answer from them in good faith. Someone at the firm has to decide what is current. That work is unglamorous and it determines whether the answers are right.

What the first week looks like

The first week starts with questions, not documents.

We collect the questions the firm actually asks. The group chat history is the best source. So is the person who gets interrupted most, who can usually list the top twenty from memory. Those questions become the acceptance test, written down before anything is built.

Then the material is traced backwards from the questions. Which document answers this one? Is it current? Where does it live? This is where the gaps surface, and they are useful findings on their own. A question that no document answers is either undocumented practice or a procedure that needs writing.

A narrow slice gets indexed first, usually one practice area or one process end to end. Narrow and correct beats broad and unreliable, and it gives the firm something to judge in days rather than months.

Then the staff who asked the questions ask them again, of the system. Every answer is checked against the source. A thin or wrong answer is almost always a gap in the material rather than a fault in the search, and the fix is to correct the document, which improves the firm whether or not the system exists.

Go-live waits for the people who answer these questions today to agree the system answers them properly. That rule holds in every build. Systems that staff route around are worse than no system.

The full scope is on the service page: /services/drafting-knowledge. To ask a fictional firm’s procedures manual a question and see the sources it answers from, the demo is here: /demos/knowledge.

The matter is waiting on a document the firm requested five weeks ago. The associate assumed the paralegal was chasing it. The paralegal assumed the client had been told it was urgent. The client assumed silence meant the firm was working. Everyone behaved reasonably and the file sat still for over a month.

Nothing failed here. Nothing was supposed to catch this.

What pipeline tracking with staleness alerts is

It is one board of every open matter, deal and engagement, each with a last-activity date, and a rule about how quiet is too quiet.

The important part is where the last-activity date comes from. It is computed from the systems the firm already uses: a time entry, a filed document, a sent email on the matter, a status change in practice management. It is not a field someone has to remember to update. A pipeline that depends on people updating a status field has the same failure as the matter above, wearing a different hat. The point of the system is to notice things when nobody is paying attention, which is exactly when nobody updates fields.

Thresholds are set per stage, because quiet means different things in different places. A new enquiry that has not been answered in four hours is a problem. A matter waiting on client documents is fine for two weeks and suspicious at four. Litigation waiting on a court date can sit for months and be entirely healthy. One global rule produces either constant noise or nothing at all.

Alerts escalate rather than nag. The first one goes to whoever owns the matter, quietly. If it is still quiet after the second threshold, it goes to the partner responsible. That order matters: the associate gets a chance to say “waiting on the court, all fine” before anyone above them sees a list with their name on it.

Every flag carries its evidence. The last activity, what it was, who did it and what the matter is waiting on. A flag without evidence is a demand for a status meeting, which is the thing this is supposed to prevent.

It reads from the firm’s systems and changes nothing in them.

Does a firm your size need it

Here is the test. Without opening anything, name the three oldest open matters in the firm and say what each one is waiting on.

Most partners can name the oldest. Almost nobody can name the third, and the third is where the damage is. Very old matters are usually known and deliberate. It is the middling ones, six weeks quiet and nobody worried, that turn into a complaint or a missed limitation date.

The clearest signal is that the pipeline is a person. If the answer to “what is open right now” is a partner’s memory plus a spreadsheet they maintain, the firm has a reporting system with one point of failure and no alarm.

The second signal is handoffs. Work that moves between people stalls in the gaps between them, because the person who let go assumes the person who picked up is on it. Firms with paralegals, associates and partners on the same matter have more of these gaps than firms where one person does everything.

The third is clients asking for updates. Every “just checking in” email is the client doing the firm’s monitoring for it, and only on the matters where the client is attentive. The quiet clients are not quiet because they are happy.

The pattern shows up in the numbers across the systems we run: 16,062 leads tracked from first contact to outcome, with a response time measured on 7,912 of them, and 70 percent of inbound leads reached on the first call. None of that comes from people being more diligent. It comes from the delay being visible while it is still small.

What it costs

This is built as a fixed-price phase inside a plan, and it is usually the least expensive build in its pillar, because it reads from systems rather than writing to them. There is no migration and no change to how anyone works.

The audit prices it as a phase before anything is built, and its own fee is credited against that phase if the firm goes ahead.

Running costs are close to nothing. It is mostly scheduled queries against data the firm already holds.

The real cost is a decision, not a build. Someone at the firm has to say how long is too long for each stage of each kind of work, and then live with being told about it. Firms that skip that conversation end up with thresholds set by whoever configured the system, alerts that fire on healthy matters, and a team that turns the notifications off in the second month.

What the first week looks like

We start by reconstructing the present. Every open matter, with its last real activity pulled out of the systems, as of today.

That first list is usually uncomfortable, and it is the most valuable output of the week. Firms consistently find matters quieter and older than anyone believed, including a few that should have been closed and a few that nobody currently owns. Cleaning those up on its own often pays for the phase.

Then the thresholds get set, with the people who run the work rather than for them. This is a working session with the practice leads, one kind of matter at a time. It goes faster than people expect, because everyone already has an instinct for when a file has gone quiet. The session writes it down.

Alerts run to a test channel first, for a couple of weeks, so the firm can see what it would have received without anyone being pinged. The false positives get tuned out there. Thresholds nearly always move at least once.

Then it goes live, when the people who will receive the alerts agree they are worth receiving. As in every build, the people who do the work decide when it is ready.

The full scope is on the service page: /services/billing-pipeline-capacity. The live demos are here: /demos.

That is intake at most firms. It is not anyone’s fault. It is what happens when the first contact depends on a person being free at the moment the client writes.

What an intake agent is

An intake agent is a system that handles the first conversation with a prospective client and everything that follows from it, up to the point where a person has to make a decision.

It answers the first message within moments, in the client’s language, at any hour, on whichever channel they used: the website, WhatsApp or email. It asks the questions the matter type needs. For a property matter, that means the parties, the address, the dates and the documents the client already holds. For a tax engagement, it is the entity type, the year in question and what has already been filed. Urgency is asked about early, because a client with a deadline next week should not be booked into the same slot as one who is planning ahead.

When it has enough to work with, it books the consult directly into the right calendar. Routing follows rules the firm sets: by practice area, by who has capacity this week, by language.

Before a lawyer or accountant spends any time on the matter, the agent runs a conflict check against the firm’s matter history. A conflict caught before the consult costs nothing. One caught after costs the consult and the goodwill.

If the agent is unsure about anything, it hands the conversation to a person with the full transcript attached. It does not guess. It does not give legal or tax advice, and it is built so that it cannot be talked into it.

Then one approval from the responsible partner does the rest. The matter is opened in practice management. The document folder is created. The billing record exists. The client has portal access. An engagement letter, drafted from the intake data, is waiting for the partner’s signature. Every step is logged, so anyone can see what was asked, what was answered and who approved what.

It runs on the phone number, inbox and calendar the firm already has. Clients do not notice a new system. They notice that someone replied.

The design is reviewed against the relevant bar or licensing-body rules before it goes live, because the rules on solicitation, confidentiality and who may speak for a firm vary and they matter.

Does a firm your size need it

The honest test is a count. How many new enquiries came in last month, and how many became consults? If the firm does not know the second number, that is the first finding.

A firm with a handful of enquiries a month, all from referrals who will wait for a callback, can run intake by hand. The agent will not change much there.

The picture changes when enquiries arrive from the website, from directories or from ads, when they arrive outside office hours, or when they arrive in more than one language. Those clients are comparing. Reply time decides where they go. A firm in that position is usually losing matters it never knew it had, because the enquiry that went unanswered for a day does not show up in any report.

The other signal is the receptionist’s day. If a large share of it is repeating the same intake questions and chasing the same missing documents, that is work the agent takes off the desk. The person stays. The paperwork goes.

Across the systems we run for firms and service businesses, the pattern holds up at volume. Those systems have logged, transcribed and reported on more than 216,000 calls, texts and forms. They have sent more than 50,000 texts automatically and routed 6,271 leads to the right person without anyone forwarding an email. Seventy percent of inbound leads were reached on the first call. That last figure is the one that matters for a firm: the client is still on the line when the firm answers.

What it costs

Intake is built as a phase in a plan, and each phase carries a fixed price agreed before work starts. The number depends on how many matter types the firm handles, how many channels are in play and what the existing systems look like. A firm with one practice area and a modern practice management tool is a smaller build than one with five practice areas and a shared inbox.

That is why intake work begins with the audit. The audit maps the current process, prices the build as a fixed phase and credits its own fee against that phase. A partner sees the full cost before committing to anything, and the plan belongs to the firm whether or not the build goes ahead.

Ongoing costs are the messaging and model usage, which are modest, and the time the firm spends reviewing the transcripts it wants to review. There is no per-seat licence.

What the first week looks like

The build starts from the intake process as it runs today. The first days are spent with the receptionist and whoever handles new matters, walking through real enquiries from the last month: what was asked, what was missing, where it stalled.

From that, the question set for each matter type is written and reviewed by the partner responsible for that area. The conflict check is connected to the matter history. Calendar rules and routing rules are agreed.

By the end of the week the agent is running in a test channel, not the live number. Staff send it enquiries themselves, including bad ones: vague messages, wrong language, someone asking for advice, someone hostile. Every handoff is checked. Every transcript is read.

Nothing goes live until the people who handle intake say it is ready. That is a rule in every build, and it is the reason the systems get used rather than routed around.

The service page covers the full scope: /services/intake-onboarding. To see an intake conversation run end to end, the demos are here: /demos.

The draft is not the hard part. The hard part is that the draft is the same work every time, and it sits between a client saying yes and the matter starting to bill.

What drafting from a form means

A drafting system takes structured input and produces a first draft in the firm’s own templates and clause library. The input can be a short form that a paralegal fills in, or the data an intake agent already collected. The output is a document that looks like the firm wrote it, because the firm did: the templates, the clauses and the house style are all the firm’s. The system assembles and fills. It does not invent language.

Engagement letters are the usual starting point, because they are frequent and structured. The same approach covers proposals, contracts, memos, deeds and tax opinions. Anything where the shape is fixed and the facts change.

Three properties matter more than the drafting itself.

The draft is marked where a person must check. Fee amounts, scope language, jurisdiction-specific clauses and anything the input did not fully cover are flagged in the document. A reviewer opens the draft and goes straight to the marks.

Every document shows which data and which template version produced it. If a letter went out with the wrong limitation clause, the firm can see whether the template was wrong, the input was wrong, or the reviewer missed the flag. That trail is what makes the system defensible when a regulator or an insurer asks.

The failure mode is a flagged blank, not a confident mistake. When the system does not have what it needs, it leaves a visible gap and says so. It does not fill the gap with something plausible. A blank is caught in review. A plausible wrong clause is caught by the client’s lawyer six months later.

The system drafts. A person signs. Nothing leaves the firm without a named human approving it, and client files are never used to train anything. The firm’s documents stay the firm’s.

Does a firm your size need it

Count how many documents of the same type the firm produces in a month. If the number is small and one person produces them all, a good template in Word is enough. Drafting systems earn their place through repetition.

The clearer signal is delay. If matters wait for an engagement letter, or proposals go out late because the person who writes them is in court, the bottleneck is the drafting queue, not the volume. A firm of any size can have that bottleneck.

The third signal is inconsistency. Two associates producing the same letter with different clauses, or an old fee schedule surviving in one template but not another. That is a sign the clause library lives in people’s heads and their personal folders. Putting it into a system is as much about tidying the library as about the drafting.

For notarial and property practices, the case is stronger still. The documents are long, the fields are many, and errors are expensive to fix after signing. Through flows we have built, 7,455 documents have been signed electronically. In one archive project, more than 103,000 notarial files were indexed and cross-linked, and 1,538 deeds were read and parsed into structured records. That is the same machinery, pointed at the back catalogue instead of the next letter.

What it costs

Drafting is built as a fixed-price phase within a plan. The price depends on how many document types are in scope, how clean the existing templates are and where the input data comes from. A single engagement letter template with a form in front of it is a small build. A clause library spanning several practice areas, fed by intake data and connected to an e-signature flow, is a larger one.

Because the size varies, drafting work starts with the audit. The audit prices the build as a phase, and its fee is credited against that phase if the firm goes ahead. The firm sees every number before committing.

Ongoing costs are model usage, which is low for document work, and the reviewer’s time. The reviewer’s time is the point: it goes down from writing to checking, and the checking is directed to the marked sections.

There is one cost that is easy to underestimate. The firm’s templates and clause library need to be in a state the system can use. If they are scattered, the first phase includes consolidating them. That is unglamorous work, and it is the work that makes everything after it reliable.

What the first week looks like

The first week is spent on the templates, not the software.

The firm hands over its current engagement letter templates and whatever passes for a clause library. Together with the person who usually drafts, we go through them and answer the same questions for each: which fields change per matter, which clauses are conditional and on what, and which sections must always be checked by a human before signing. That last list becomes the flag rules.

The input form is designed from the fields. If intake data already exists, it is mapped instead. A paralegal should be able to fill the form in a couple of minutes, and every field should be one the firm actually needs.

By the end of the week, the system produces drafts for a set of past matters. The person who normally drafts compares each one against the letter that actually went out. Differences are either template fixes or flag-rule fixes, and both get made before anyone uses the system for a live matter.

Go-live waits for that person to say the drafts are right. As with every build, the people who do the work test it, and nothing ships until they trust it.

The full scope is on the service page: /services/drafting-knowledge. To watch a form become a marked-up draft, the demos are here: /demos.

What the audit is

The AI audit is a short engagement with one output: a ranked, priced plan of what in your office should be automated, what should be left alone, and in what order.

It follows one week of normal work at the firm. During that week we sit with the people who actually do the work: the receptionist, the paralegal, the bookkeeper, the account manager. Not the partners, at least not at first. Partners know what the firm sells. The staff know where the time goes.

Every recurring process gets mapped. For each one we record who owns it, which tools it touches, how many hours a week it takes, where it stalls, and what breaks when that person is out sick or on holiday. Intake, conflict checks, engagement letters, billing runs, filing deadlines, document requests, client updates. The list is usually longer than anyone expected.

Then each process is run through the same four questions, in order:

1. Does this need to happen at all?
2. If yes, does a person need to do it?
3. If not, can ordinary software do it?
4. If not, can AI do it reliably, with a person checking the parts that matter?

The order matters. Most processes fail at question one or two. A monthly report nobody reads. A form that is filled in twice because two systems do not talk to each other. A status email that could be a shared folder. None of that needs AI. It needs to be dropped, or handed to a calendar rule, or fixed in the practice management system you already pay for.

What survives to question four is the real candidate list. Those are the processes where the work is genuinely judgment-shaped but repetitive enough that a model can do the first pass and a person can check the result.

The candidates are ranked by return: hours recovered per week, errors avoided, and revenue that stops leaking. A conflict check that takes two days to come back is a lead that goes elsewhere. An engagement letter that sits unsigned for a week is a matter that has not started billing. Those numbers go into the ranking next to the hours.

Does a firm your size need it

If the firm has one person doing everything, probably not yet. Map your own processes on a sheet of paper and run the four questions yourself. The framework is not secret.

The audit pays for itself once there are enough people that nobody holds the whole picture. That usually happens somewhere around the point where a partner no longer knows exactly how a new matter gets opened, or an office manager is the only person who knows why a particular spreadsheet exists. At that size, processes have grown around individuals, and the cost of each one is invisible because it is spread across several desks.

There is a second signal: someone in the office is already using AI tools on their own, without a policy, on client material. That is a sign the demand is real. It is also a sign the firm needs to decide, deliberately, where these tools belong and where they do not.

The audit is not a good fit if the firm wants a general technology strategy or a software selection exercise. It is narrow on purpose. It looks at recurring work and decides what to do about it.

What it costs

The audit is a fixed fee of $1,500 USD.

That covers the working sessions, the process map, the ranked list and the phased plan. Each phase in the plan carries its own fixed price, so the firm sees the cost of every step before committing to any of it. There is no estimate that grows later.

If the firm goes ahead with a build, the audit fee is credited against the first phase. If the firm decides to stop after the audit, it keeps everything. The documents belong to the client. A firm can take the plan to another provider, or hand it to an in-house person, or sit on it for a year. That is fine.

One more cost worth naming: staff time. The sessions take a few hours from the people being interviewed, spread across the week. Their time is the most valuable input to the audit, and the firm should plan for it rather than squeezing the sessions between calls.

What the first week looks like

Before the week starts, the firm sends a short list: who does what, which systems are in use, and any process that is already known to be painful. That takes an office manager about half an hour.

Day one is a kickoff with whoever runs the office day to day. We agree on which roles to sit with and when. Nothing is installed. No access to client files is needed at this stage.

Over the next few days, the working sessions happen. Each one is a conversation at the person’s desk, watching how a matter actually moves. Where do you copy this from? What happens when the client does not reply? Who checks this before it goes out? The point is to see the process as it is, not as the procedures manual says it should be.

By the end of the week, the process map is drafted. It is reviewed with the same staff who described the work, so the map reflects reality and not our interpretation of it. Corrections get made then, not after the plan is written.

The ranked list and the phased plan follow shortly after. They are walked through with the partners in one session, with the prices attached.

Adoption is written into every phase of every plan. The people who do the work test what gets built. Nothing goes live until they trust it. That rule is not a courtesy. Systems that staff route around are worse than no system, and the audit is where that expectation is set.

If the firm is ready to see what a ranked plan looks like for its own processes, the audit page has the details and the booking form: /audit.