Key takeaways
- The dividing line is systematic versus incidental. Does AI run by default on every matter, or by whim when someone remembers?
- Seven observable markers define an AI-native practice, and every one of them is visible from outside - which makes the claim falsifiable.
- Licence count is not a marker. A firm with an enterprise AI deployment and hourly billing on manual intake is AI-augmented.
- AI-augmented is not a failure state. For structurally constrained organisations it is the honest destination, reached via AI Adoption.
- No firm hits all seven markers, and that is fine - they indicate direction, not a certification to pass.
The line: systematic versus incidental
The cleanest definition of an AI-native practice is not by the technology it runs but by what its clients experience, because that is where the operating model becomes visible. And the single distinction underneath all of it is whether AI runs by default or by whim.
A firm where three lawyers occasionally paste a document into a chatbot when they remember to is not AI-native, however many licences it holds. A firm where every matter, every time, gets an automated first pass before a human touches it - reviewed, corrected, and taken responsibility for by a lawyer - is running a different operating model entirely. The threshold question is always the same: does AI run by default, or by whim?
This matters commercially because the two produce completely different results. Incidental use produces occasional time savings for individual lawyers, invisible at practice level. Systematic use produces measurable improvement in the institution, because every matter runs through the same system and the system generates data about itself.
The seven markers, and how to check each one
| Marker | How to verify it from outside | Common false positive |
|---|---|---|
| Pricing is transparent, not open-ended | The client knows the cost before work starts - published or instantly quoted. | A published 'starting from' figure with hourly billing behind it. |
| Intake is automated end to end | A matter can be opened and substantive work begun the same day, without a scheduling call. | An online contact form that generates an email to a human who then calls. |
| The firm meets clients where they are | Communication in the client's existing channels, not a portal requiring a password reset. | A client portal nobody logs into, counted as digital engagement. |
| AI runs first, systematically | Every matter of a given type gets an automated first pass by default, not on request. | Enterprise licences issued firm-wide with adoption measured by seats, not by matters. |
| The firm improves measurably | Turnaround, accuracy, and revision rates are tracked and have moved. | Improvement asserted from partner impression rather than measured. |
| Turnaround is fast, often same-day | Work starts when it is submitted, not when it reaches the top of an inbox. | Fast on urgent matters through heroics, normal speed otherwise. |
| Technical capability sits inside | Product thinking and data literacy treated as core, not as an IT function bolted on. | An innovation committee with no operational authority. |
Why most firms land on AI-augmented, and why that is not a failure
The overwhelming majority of firms that adopt AI tools remain AI-augmented, and for large firms in particular this is structural rather than a matter of effort. A partnership whose margin depends on billing junior hours cannot fully automate junior work without dismantling its own economics. A regulated in-house function cannot re-architect intake by decision. Both adopt the tools and protect the model, which is exactly what incumbents facing a structural threat have always done.
It is worth saying clearly that AI-augmented is not a shameful outcome. For an organisation with genuine institutional constraints, the honest destination is a systematic AI layer installed inside the structure that exists - the same six primitives named and systematised, with automation deployed inside existing sign-off chains. That is AI Adoption, and pursued deliberately it delivers most of the operational benefit.
What is a failure is claiming the first while doing the second. A firm that has bought licences and describes itself as AI-native is making a claim that is externally checkable against seven markers, most of which it will fail. Sophisticated clients increasingly run that check.
The honest framing
The question is not whether you are AI-native or behind. It is whether your structure permits a full rebuild, and if it does not, whether you are running a deliberate AI Adoption programme or drifting while telling yourself otherwise.
Assessing your own practice honestly
Take one marker at a time and answer against evidence rather than intention. The most reliable test for the central marker - AI running systematically - is to pick a random matter opened last month and ask whether the automated first pass ran on it. Not whether it could have. Whether it did.
If the answer depends on who was handling the matter, you are AI-augmented. If the answer is yes regardless of who was handling it, the marker is met. That single test, applied to a random matter rather than a chosen one, cuts through more self-assessment noise than any maturity questionnaire.
Frequently asked
What is the difference between an AI-native and an AI-augmented law firm?
Systematic versus incidental. AI-native means AI runs a first pass by default on every matter of a given type, with a lawyer reviewing and taking responsibility. AI-augmented means AI is available and used when individual lawyers remember to. The licences may be identical; the operating models are not comparable, and only the first produces measurable institutional improvement.
How can I tell if a law firm is genuinely AI-native?
Check the seven markers from outside: is pricing known before work starts, can a matter be opened and worked the same day without a call, does the firm communicate in your channels, does an automated first pass run on every matter by default, are turnaround and accuracy measured and improving, is same-day turnaround normal rather than heroic, and is technical capability core rather than an IT function. Most firms claiming the label fail the majority.
Is AI-augmented a bad outcome for a law firm?
No - for organisations with genuine structural constraints it is the honest destination. A large partnership or a regulated in-house function usually cannot complete a full rebuild on any realistic timeline. Pursued deliberately as AI Adoption - the same primitives named and systematised, automation installed inside existing sign-off chains - it delivers most of the operational benefit. The failure is claiming AI-native while running AI-augmented.
How many of the seven markers does a firm need to be AI-native?
There is no threshold, because they indicate direction rather than certification. No firm hits all seven. The one that functions as a gate is systematic first-pass AI - a practice that fails that marker is AI-augmented regardless of how many others it meets, because the rest tend to follow from it rather than substitute for it.
Does buying more AI tools make a firm more AI-native?
No, and this is the most common misconception. Licence count is not a marker. A firm with a large enterprise AI deployment, hourly billing, and manual intake is AI-augmented. A solo practitioner with two well-chosen tools running systematically on every matter, transparent pricing, and automated intake is closer to AI-native. The operating model is what is being measured, not the software budget.