In 2025, most law firms reached the same conclusion: a firmwide legal AI platform is no longer optional – it’s essential, writes Joe Rayment.
It’s a sensible productivity investment. Drafting is faster, summaries are cleaner, and first passes happen earlier.
So, why do so many firms still feel like they’ve got AI without quite banking the benefit?
The answer is straightforward: general AI improves writing. Specialist tools improve outcomes.
The multi-player reality
A law firm is a collection of micro-professions under one brand, each with different outputs, risks, and definitions of value. Take a single matter and look at the people involved. A corporate partner lives in negotiated language, precedent position, and pragmatic risk trade-offs. A litigation heavyweight lives in theory-of-the-case, evidentiary discipline, and the brutal cost of a factual error.
Both can benefit from base productivity tools. But expecting one general platform to deliver the same return on investment (ROI) across both is like expecting one practice group to run the whole firm.
That’s a view shared by Harvey’s co-founder, Winston Weinberg. In a recent Reddit post, he noted the legal AI market will not be winner-takes-all.
“I don’t think a single player is going to capture all of the pretty enormous amount of value that will be created in the next ten years in this space,” he noted.
When asked whether lawyers will use Harvey alongside other AI products, he answered yes, explicitly pointing to integrations, and concluding: “We can’t build everything.”
Those comments tell you what your firm’s AI reality is likely heading in 2026 and beyond: a base productivity layer across the firm, plus specialist tools where the work is deeper, riskier, or more operationally complex.
Where firmwide AI platforms shine (and don’t)
Global legal AI platforms excel when the task is text-shaped, iterative, low-risk to verify, and often single-player. That maps naturally to corporate work, where transactions move quickly and the lawyer driving the work can verify the result themselves.
Litigation is different. Matters run for months or years. The evidentiary record grows and shifts. Multiple people touch the same analysis at different times. A chat transcript from three months ago doesn’t help the senior associate who’s just been staffed, and it won’t help when the other side serves new documents that change the picture.
When litigation partners talk about ROI, they don’t mean nicer paragraphs. They mean fewer write-offs, reduced rework, faster turnaround, better leverage, and confidence that work can withstand scrutiny. That gap between individual productivity and institutional performance is where the next wave of legal AI value sits. Bridging it is mostly workflow design.
Why firms hit the adoption ceiling
Many firms have also hit an adoption ceiling with their AI tools. A small group becomes power users who build prompts (essentially instructing AI) and show impressive results. But most lawyers don’t want a second job as prompt engineers. They want AI tools that behave like legal systems: predictable, guided, and embedded into how work is actually delivered.
The next step isn’t telling everyone to prompt better. It’s shifting from open-ended chat to role-specific workflows with pre-built tasks, structured outputs, and controls that make the right way the easy way.
A distinct category for litigators
If legal work inside a firm is diverse, specialist tools won’t be one thing either. Transactional teams may need systems that standardise positions. Knowledge teams may need systems that govern precedents. But for litigation and investigations, a distinct category is emerging: matter analysis platforms.
These platforms are designed around the reality that litigators don’t win by producing more text. They win by controlling the record. That means helping teams create and maintain matter artefacts – structured, evidence-linked, persistent work products that are tied back to source material, repeatable, reviewable, and able to survive handovers.
Litigation is high-stakes and multi-player by nature: partners, seniors, juniors, counsel, experts, and clients. A chat transcript doesn’t coordinate that. A matter artefact does. And the risk profile flips: in disputes, close enough can be catastrophic if it becomes a submission or affidavit. The requirement is defensibility – provenance, auditability, and clear review lanes.
What to do next
If you already have a major platform investment, the next 90 days are about converting access to AI into outcomes for your firm.
Segment by role and risk, not enthusiasm. Pick two workflows where the cost of rework is obvious. Build the review lane before you scale – who checks what, when, and what triggers escalation.
Decide where you need evidence-linked, auditable outputs. Measure the outcomes partners actually care about: cycle time, write-offs, rework loops, and confidence under scrutiny.
None of this requires a dramatic leap in model capability. It requires accepting a simple truth about law firms: the work is diverse, and your AI approach needs to be as well.
Global legal AI platforms are a smart step – arguably the necessary step – because they raise the floor across the firm. But the next gains come from raising the ceiling in the areas where legal work is most specialised, where outcomes depend on workflow, coordination, repeatability, and defensibility.
Joe Rayment is CEO of Automatise, which develops Cicero, a matter analysis platform for litigation and investigations.
