An AI-Native Law Firm Is Not a Law Firm That Uses AI

For general counsel, founders, and brand teams choosing how to run a multi-jurisdiction trademark portfolio.

Every law firm uses AI now. Drafting tools, research tools, document review. Enough firms have claimed the term "AI-native" that it has stopped meaning anything on a website.

It still describes something real. The difference is not how much technology a firm uses. It is where that technology sits relative to how the firm makes money.

The test

When a firm gets faster, who keeps the saving?

Under hourly billing, the firm's revenue is time. A tool that saves a paralegal four hours takes four billable hours off the invoice. Firms can adopt that tool, and many do, but every efficiency it creates works against their own revenue. So they resolve it the way any business would. They adopt what makes the work more profitable per hour, and go slowly on what shrinks the bill.

The record bears that out across legal spend generally. Thomson Reuters' Legal Department Operations Index tracks the share of corporate legal spend going to traditional law firms. In 2007 it was 92 percent. By 2024, after seventeen years of legal technology, it was 86 percent. Six points. Roughly 90 percent of law firm spend is still billed hourly, and fixed fees account for around 5 percent.

Now the rates. In 2024, worked rates across US law firms grew 6.5 percent, the fastest pace since the financial crisis, per Thomson Reuters and Georgetown Law's annual analysis of 183 firms. Citi's separate survey of large firms put billing rate increases at 9 percent over the first nine months of the same year. And Thomson Reuters notes client pushback was minimal, with realization rates holding steady, meaning firms collected those increases rather than discounting them back out.

Here is the part that matters. The same report puts average law firm expense growth above 5 percent over that period, against a pre-pandemic norm of around 3 percent, and attributes the increase primarily to technology and knowledge-management spend rather than to salaries or general inflation.

So firms invested in technology, their costs went up, and the increase reached clients as higher rates. The technology cost was passed through. The saving was not.

One caveat worth naming, because a reader who follows this data will know it. Thomson Reuters flags that part of the measured growth reflects more senior lateral hires lifting the average rather than price rises on identical work, so the like-for-like increase is probably smaller than the headline number.

The direction is not in question, though. Nominal rates rose through this period, and nominal is the only unit a budget is denominated in. A trademark budget set in kronor or euros or dollars does not get adjusted for whether the firm charging it beat inflation. It buys what it buys, and each year it has been buying less.

Trademark-specific rate data barely exists, and it is worth saying so rather than implying otherwise. The main US IP survey puts the median private-firm IP attorney rate at $450 an hour in 2022, unchanged from 2020, against $295 in 2004. Nobody publishes a trademark-only series. Nobody publishes European or Nordic rate data at all.

On fee structure, though, trademark work is moving faster than the general market. INTA's 2020 benchmarking found 86 percent of in-house trademark practitioners using hourly rates with outside firms. In INTA's 2025 survey of 291 in-house practitioners, 42 percent had moved to fixed, capped or blended fees, against 2 percent who reported using more hourly billing than before. Two percent. One respondent said they still value their law firm partners but can no longer justify open-ended hourly billing to management.

So clients have already decided what they want. What has not been settled is which firms can actually do it. INTA's research is explicit that in-house teams now expect their costs to come down as their external partners adopt AI. That expectation is where a firm's own cost structure stops being its internal business and starts deciding whether it can hold the price it quoted.

A firm that offers a capped fee while still paying people to do the tracking is taking on a cost it cannot control. That works until it doesn't. Then the price goes up, or the cap quietly starts covering less.

An AI-native firm is built the other way around. The price is fixed, so a saving lands on the firm's costs instead of the client's invoice. When the software gets better at tracking renewals, the firm's margin improves and the client's price does not move. Over time the client gets more for the same number.

That is the whole difference. It is structural, not a matter of how enthusiastic a firm is about technology.

The second half of it is ownership. A firm that licenses generic docketing software can only automate what that software allows, and pays a third party for the privilege. That licence sits in the firm's costs, and the firm's costs sit in your rate, whether anyone names it or not. A firm that built its own can shape the automation around the legal work instead of shaping the work around someone else's tool. It also controls the costs underneath its own pricing, which is what makes a fixed price possible at all.

What it does not mean

It does not mean AI makes the legal decisions. Worth being blunt about, because plenty of firms have made claims here that did not hold up.

Trademark work has a clean line running through it. On one side, measurement. Comparing a mark against registrations across jurisdictions, scoring how similar two marks sound and look and mean, screening for descriptiveness and other absolute grounds, pulling deadlines out of an office action, keeping a renewal calendar. This work is mechanical, high in volume, and the kind of work where people miss things. Software is better at it.

On the other side, judgment. Whether a similarity score amounts to likelihood of confusion. Whether to oppose or coexist. Whether a hundred-year-old registration is vulnerable to non-use. Those are legal calls, not scores, and they belong to an admitted attorney who puts their name to them.

AI-native means the first half runs on software and the second half runs on lawyers, with a record of both. It does not mean the line moves.

The market agrees on where that line sits. In INTA's 2025 follow-up survey, 74 percent of in-house trademark practitioners were actively using AI tools, 40 percent of them daily. Their main concerns were accuracy, at 86 percent, and confidentiality, at 84 percent. Seventy-seven percent said their employers now have formal AI policies. Practitioners in that research describe AI as indispensable but not ready to replace human judgment. One put their strategy as human first, human last.

Same line. Take the measurement, keep the judgment.

The split matters because of how the work divides. INTA's 2020 benchmarking found routine work, prosecution and registration, maintenance and clearance, taking 37 percent of in-house team time on average, against 16 percent for enforcement, litigation and opposition. Outside firms handled 82 percent of the prosecution and registration work, 64 percent of clearance, and 61 percent of maintenance.

Most trademark work is routine, and most of the routine work is bought from law firms by the hour. That is the part software changes. The judgment stays where it was.

How to tell

The claim is easy to make and easy to check. Five questions do it.

Does the firm own its software, or license it? Ask what is built and what is bought. A firm running on a generic platform with legal services on top is a traditional firm with a subscription.

Can a saving reach you through the fee structure? If the firm bills by the hour, a faster system cannot lower your cost. It can only lower the firm's cost of producing the hour. Ask what happens to the price when the technology improves. If nothing happens to it, ask why.

Where does the machine stop and the attorney start? A firm that can draw that line precisely, for a specific deliverable, has thought about it. A firm that calls software output legal advice, or cannot say where its own tools end, has not.

Can you see the record without asking anyone? The point of automating this work is that the state of the portfolio is always current. If getting your live mark count still takes an email, the firm automated its own job and kept the benefit.

Who built it, and can they still change it? Software the firm owns gets fixed when your portfolio needs something specific. Software the firm rents gets a support ticket.

Three costs, one invoice

A general counsel running trademarks pays for it in three places. Only one of them arrives with the word trademark on it.

Outside counsel spend. Visible. Filing, prosecution, office action responses, oppositions, renewals, and the hours spent on all of it.

Software. Mostly invisible. ACC's benchmarking puts IP management software adoption at around 20 percent of legal departments, with legal technology accounting for roughly 3 to 5 percent of a legal budget. Read quickly, that says software is a small problem. Read properly, it says something more useful. Most companies never buy trademark software because their law firm already did. Docketing systems, watch platforms, portfolio databases. The firm licenses them, the licence sits in the firm's costs, and those costs sit in the rate. You have been paying for that software for years. You have never seen it as a line item.

Internal time. Also mostly invisible. Nobody's title says trademark coordinator. Somebody still instructs counsel, chases status across firms, reconciles invoices in four currencies, and keeps the spreadsheet that holds the only complete view of the portfolio.

So two of the three costs are real and neither one is visible. One sits inside your counsel rate. One sits inside your team's week.

That is why the saving works in both directions at once. When outside counsel spend falls by 30 to 50 percent, the software cost falls with it, because that is where the software was. Not as a separate discount, but because a rate that used to carry someone else's licence fee no longer does. We built the system we run on, so there is no vendor margin inside our price, and the cost of running it does not grow with your portfolio the way paralegal hours do.

That is the same pass-through as before, and it is not only true of software. A firm's costs reach you through its rate. INTA's 2025 report makes the general point directly: law firms have absorbed higher costs in recent years and passed them on as higher billable rates to corporate clients. Software is one of those costs.

If you are in the minority that licenses IP management software directly, the arithmetic is starker. You are paying for two systems. Yours, and the one inside your firm's rate. They do not share a record, so somebody reconciles them.

The internal cost moves for the same reason. Mechanical work that used to land on your team runs on software instead, so the coordination work stops growing every time the portfolio does.

Worth stating plainly, because it is the claim we are making: we typically reduce outside counsel spend by 30 to 50 percent while internal efficiency goes up at the same time. Those are not competing outcomes, and the fact that they arrive together is the tell that something structural changed rather than somebody discounting.

What this changes about centralized and decentralized

The two axes have not moved. Where you sit on them still decides how the relationship works.

What changes is that the bottom-right box used to be mostly empty for a serious portfolio.

Centralized service at a genuinely fixed price meant one of two things. Either the portfolio was small enough that the fixed price was easy, or the firm was carrying a cost it could not predict and would eventually reprice. No firm can hold a flat fee across 190 jurisdictions while paying people to do the tracking. The arithmetic does not work. So for a company with a few hundred marks and no trademark department, the real options were to overpay by the hour for one firm's attention, or to run the decentralized model without the team it assumes you have.

Automating the mechanical work is what makes that box real. And it moves the threshold upward, which is the opposite of what most people assume.

The INTA benchmark from the last piece in this series, a median of around 4,500 marks and six people at one to two million dollars a year for a dedicated in-house trademark team, describes what it costs to run a portfolio when people do the tracking. Take that work off people and the portfolio size at which you are forced to build a department goes up, not down. More marks, more jurisdictions, more matters, all manageable before you have to hire.

That matters more now than it would have a few years ago, because the money is moving the other way. In INTA's 2025 survey, 31 percent of in-house teams reported budget cuts over the previous five years, against 25 percent reporting increases. Thirty-one percent reported team size reductions. Lawyer headcount was down in 43 percent of organizations and administrative support was cut in 46 percent. Only 5 percent reported real growth in lawyer numbers, mostly in high-margin sectors like luxury and pharmaceuticals.

Meanwhile 82 percent reported their work shifting significantly toward strategic, business-facing responsibilities. The job gets bigger and the support staff gets smaller. Administrative headcount is being cut hardest, and administrative work does not disappear when the person doing it does. It lands on whoever is left.

So the bar for building a department is high, the budget to reach it is flat or falling, and the people who handled the mechanical work are the first ones cut. And those budget figures are nominal, set against rates that rose over the same years.

That is a squeeze, and it is the condition most of this audience is working under.

These numbers describe the best case

One thing to hold onto about all of the figures above. INTA surveyed its own member in-house practitioners. These are people whose job is trademarks, at organizations that pay for both a trademark function and the professional membership that supports it. The report itself notes that membership cost is a barrier for teams under budget pressure, which skews the sample further in the same direction.

So a 31 percent budget cut rate, and admin support cut in 46 percent of organizations, is what happens at companies that have trademark specialists.

Most companies do not have one. Trademark work sits with a general counsel who owns everything else legal as well, or with an IP manager whose real job is patents, or with marketing. Nobody absorbs the mechanical work, because absorbing it is not anybody's role. Nobody owns the renewal calendar as a job. Nobody is checking whether a filing in one market still lines up with a mark in another.

If that is your setup, do not read these numbers as your situation. Read them as the mild version of it. Whatever strain the specialists report, you have more of it and less to absorb it with.

So for most companies the effect is not that centralized wins on principle. It is that more companies can now run a growing portfolio without building a department, and a portfolio that would have needed one five years ago may not need one now.

For the companies that already have the team, the department, the local-firm network, the effect runs the other way. The mechanical work can come out from under the decentralized model without dismantling it. Clearance, watch and matter tracking move to software. The team keeps the strategy, the local firms, and every legal decision. Decentralized stops needing six people to run well.

So the honest answer is that AI-native does not settle the centralized-versus-decentralized question. It makes both models cheaper to run, and it moves the line between them.

Where this fits

Worth looking at if:

• The portfolio is active and growing across jurisdictions, and trademark work currently sits with a general counsel or an IP manager as a side duty.

• You are paying hourly for work that is mostly tracking, docketing and status reporting rather than legal judgment.

• You cannot see the current state of your own portfolio without asking someone.

• You have a team already, and the mechanical work is eating senior time better spent on judgment.

Not worth the conversation if:

• The brand is not a strategic asset. A company pulling back from markets it once filed in, or consolidating many brand names into fewer, has a flat or shrinking mark count and no coordination problem to solve.

• The business is winding down trademark activity. Renewals on a shrinking portfolio are a maintenance task. One local firm will cost less than any model built for growth.

Where we land

We are a trademark law firm that built its own software, and every claim above is one we expect to be asked to prove.

We own the system the portfolio runs on. It handles clearance, watch, docketing, renewal tracking, portfolio and matter management, and domain management, which is why we can price a whole relationship as one fixed number instead of metering the hours. Sized to the portfolio, across more than 190 jurisdictions. It is also the same system you look at, so it is not a platform you license alongside us.

The result we typically deliver is a 30 to 50 percent reduction in outside counsel spend, with internal efficiency going up rather than down at the same time.

Admitted counsel does the legal work in every jurisdiction, ours or one of the 15 or more local counsel firms in our network. The software measures and tracks. The attorneys decide. You can see the state of the estate whenever you look, without asking anyone.

We are a law firm, not a filing platform borrowing legal language. Every step is on the record.