Time to Look Beyond Finding Problems Your AI Can Solve
Law firms are buying generalist AI platforms, then scrambling to find problems they can solve. We heard a story. An AmLaw 100 firm spent 12 months trying to make their $500,000 AI platform review contracts. Meanwhile, their $25,000 contract analysis tool was left unused — even though it was purpose-built for exactly that task.
This is backwards.
This way of experimenting with Gen AI platforms is taking a solution and looking for a problem. And we know from the last few decades of working with technologies how that story ends.
So, over the last 12-18 months, we have watched a swath existing legal tech products — the ones that solve specific problems with more “conventional” methods or technologies — get pushed to the back of the queue behind generalist AI platforms.
The pressure is real. Partners are breathing down necks of innovation and IT teams demanding AI adoption. Everyone needs to "be seen doing something" with generative AI.
Here's our unsolicited advice: Pick a generalist AI platform. Pay for it. Get that out of the way. Then, treat the generalist AI platform as a training tool. The tools will change faster than your people.
The Last Mile Problem
Generalist AI platforms are incredible. They can do things that were literally impossible before 2022:
Ideation and creativity from scratch
Drafting from vague instructions
Getting "more than halfway there" when starting from nothing
These capabilities should absolutely be put in the hands of lawyers. But that doesn't mean these generalist AI platforms solve every problem. It doesn't mean they replace specialist solutions that fit precisely into specific workflows.
Generalist platforms are, by definition, not tailored for specific problems. So, users end up with the last mile problem — the gap between what the the tool delivers and what users actually need. Every time there is a last mile problem, humans have to exert labor to close that gaps. Always.
What we're seeing now — and what we expect to continue seeing through 2026 — is firms that have paid for these Gen AI platforms trying to make them work everywhere for everything. Finding where they excel. Finding where they come up short. Finding where, frankly, a purpose-built tool would do the job better.
This is the necessary experimentation phase. But these experiments are happening at the cost of ignoring the tools that actually solve specific problems well.
The Holistic Legal Tech Solution
So what does a complete legal tech system actually look like? Think of it less like a brain in a jar, and more like a complete body.
AI works like a brain — replacing work that could have been done by human thinking. The AI brain is connected to a nervous system with receptors that receive inputs and detect what's happening in the world. These receptors ping that central AI brain to trigger responses. Just like a human nervous system, those responses happen at three different speeds:
Reflex responses: You touch something hot, you pull your hand back. Immediately. Automatically. No thinking required. In legal tech, these are the systems that just handle routine matters instantly, keeping you safe without any human intervention.
System 1 thinking: Fast, trained responses. Your brain has learned to handle these situations through experience. In legal tech, the system are powered by language models and RAG workflows to provide first-cut answers that are quick and “good enough”.
System 2 thinking: Slow, deliberate analysis. This is where you actually think hard about consequences. In legal tech, this is where human expertise matters — any outputs from machines are either initiated or reviewed by humans experts.
AI has skewed how some people perceived legal tech. Legal tech products that are more than just brains in jars. Full systems are like bodies, and bodies have hands. Hands need tools.
The fourth category of legal tech that is largely being ignored in all the AI hype is deterministic tools. The tools are like calculators — specialized to operate reliably. Even though AI may brute force their way to answers, there are many situations where using AI would not make sense. Often, we are better off with a calculator. In legal tech, these are tasks like project management, search, data management, comparisons, template creation, proofreading, etc.
So, four categories of tools will exist:
Reflex systems: Always on, passive, operating in the background
System 1 tools: Fast thinking, can be always-on or activated on demand
System 2 tools: Slow thinking, deliberative, requiring human-AI collaboration
Deterministic tools: Deterministic, reliable, always ready when you need them
Each category needs completely different user experiences.
Reflex systems probably don't even need a user interface beyond a dashboard.
System 1 and System 2 tools need rich interfaces for human interaction and receptors to detect triggering events in the world.
Deterministic tools need to be perpetually ready for activation by humans or AI agents.
Generative AI is transformative, but it's not the complete answer. The future isn't just AI that thinks. It's AI that thinks, tools that calculate reliably, and systems that combine the different categories and know when to use each category. The hard thing will be to build systems that seamlessly integrate all those different categories of technologies in a manner that will actually solve real problems… and the ultimate builders of those holistic systems may not be the legal tech vendors, but they may be the legal service providers.
Who Will Be Here in the Future?
Fast forward three to five years (maybe ten), what does the legal tech landscape actually look like?
The answers depends on the broader legal services market. What tools are positioned to serve the lawyer of tomorrow? What tools are positioned to serve the clients of tomorrow?
We expect to see:
massive consolidation of legal tech vendors, big and small
comprehensive ecosystems around sources of data
tech giants competing in this narrow vertical
(a possible maybe) artificial general intelligence directly delivering legal services to end consumers
The market is changing fast. Gen AI has made it possible for machines to solve problems that previously were reserved only for human subject matter experts. For example, look at Clio. They are not just building practice management software — they're integrating the business of law with the practice of law into one ecosystem. Matter management, project management, billing, research, drafting, everything a firm needs. Who else is positioned to do this? Maybe Harvey and Legora — the newer players who can move fast to keep expanding while the incumbents can't pivot quickly.
The future winners isn't determined by who has the best AI. It's about identifying problems that matter to clients, and orchestrating reflex systems, fast thinking, slow thinking, and deterministic tools into a coherent whole to solve those problems.
The future belongs to lawyers and legal service providers that move past the AI gold rush and get back to the fundamental work of understanding their clients' problems.
The playing field is flatter than ever. Everyone has access to the same underlying LLMs. The question isn't who has the technology, but it's who knows what to do with it.
What This Actually Means for You
If you are a law firm leader or an in-house GC, here are our thoughts on what you need to do:
Stop treating your generalist AI platform as something that needs to solve every problem. Pick one. Pay for it. Accept that it is a part of your IT infrastructure now, like email or document management. Then move on to focus on the problems faced by your clients.
Start mapping your actual workflows. Where do you need reflex systems? Where do you need fast thinking? Where do you need slow, deliberate analysis? Where do you need deterministic tools that work the same way every time? You may end up having to build these systems.
Invest in training across your organization: Lawyers will need to know how to make the most of Gen AI tools. It has enabled capabilities previously left only to sci fi, but understand that your competitive advantages will not be in having access to the same AI platforms as everyone else.
If you're building legal tech, here are our thoughts on what you need to understand:
Know which category you're building for. Are you building reflex systems, System 1 tools, System 2 tools, or deterministic tools? Because each one needs completely different features, user experiences, and interfaces.
Don't try to be everything. The generalist platforms are spending billions to claim that space. That fight will be won by marketing dollars. Your opportunity is in solving specific problems really, really well.
Deterministic tools matter. Maybe even more than they did before, because both humans and AI will make use of deterministic tools to deliver reliable, repeatable results every single time. Integrations will become expected features.
