If AI Is Making Lawyers More Productive, When Do Lawyers Actually Benefit?

In my last post I argued that legal work is not really a question answering problem. Most legal AI tools still assume the workflow begins with a document and a question: upload the contract, ask for a summary, extract a clause, identify a risk. That approach is useful and often impressive, but it reflects only a small part of how legal work actually unfolds.

Legal work rarely begins with a question. It begins with an objective: close the deal, respond to the regulator, complete a restructuring, resolve a dispute. From that point forward the work evolves into a sequence of tasks, dependencies, approvals and decisions that shift as new information appears. Documents are inputs into that process, sometimes very important ones, but they are not the organising structure of the work itself.

That is why I suggested the real change in legal AI will be from document intelligence to planning intelligence, where systems begin to understand the structure of a legal matter and what needs to happen next.

There is, however, another question sitting behind this change.

Even if AI dramatically improves productivity, when do lawyers themselves actually feel the benefit?

Productivity Gains Are Already Visible

The productivity gains from the current generation of tools are real. Lawyers are clearly better off not manually extracting the same clause from fifteen contracts, in much the same way engineers are glad they no longer spend hours writing repetitive boilerplate code. Removing tedious work reduces error and allows professionals to focus on tasks that require judgement rather than patience.

Research increasingly shows how large these gains can be. Some studies comparing lawyers and language models on routine tasks such as invoice review show lawyers taking several minutes per invoice while AI systems complete the same task in seconds, often with higher accuracy.

Now even if the exact numbers vary across tasks, the general point is obvious, that small improvements in productivity compound quickly in a profession where thousands of hours of work are performed across large teams.

Yet productivity improvements do not automatically translate into a better experience of the job.

The Structure of Legal Work Has Not Changed

Despite better tools, the underlying pressures of legal work remain largely unchanged. Surveys across the profession consistently highlight patterns such as:

• long working hours and unpredictable workloads
• expectations of constant client responsiveness
• work regularly extending into evenings and weekends
• blurred boundaries between professional and personal time

Anyone who has worked in a law firm recognises all this immediately. Laptops come on holiday, emails arrive late at night and the working day rarely ends exactly when the calendar suggests it should.

Against that backdrop, AI that makes individual tasks faster only solves part of the problem. It improves efficiency within the existing system without necessarily changing the structure of that system. When a task takes half the time, the saved time rarely becomes free time, but ,ore often it becomes the next task waiting in the queue.

The Commercial Reality

There is obviously a commercial reality to legal work. Law firms exist to serve clients, and the matters they handle often involve significant financial and regulatory consequences. Responsiveness is part of the professional obligation, and the expectation that lawyers will be available when needed is unlikely to disappear.

At the same time, the profession is increasingly having another conversation. Law firms regularly talk about:

• wellbeing and mental health
• sustainable careers
• creating environments where talented people want to work

These ideas appear frequently in recruitment messaging, leadership speeches and internal strategy discussions across the industry.

If those conversations are genuine, the role of AI should eventually extend beyond making work faster and begin addressing how work is experienced.

The Next Phase of Legal AI

That shift becomes easier to imagine once legal AI starts to move beyond document questions and toward understanding the structure of work itself. A system that understands the state of a legal matter can begin to track:

• outstanding tasks
• unresolved dependencies
• upcoming deadlines

Over time those systems could also incorporate the context of the lawyer responsible for that work, including their calendar commitments and existing workload.

At that point AI stops behaving like a document assistant and starts behaving more like an operational layer for legal work.

AI That Understands Work Allocation

Once systems begin tracking the structure of legal work, they can also make better decisions about who should be doing it.

Consider a familiar situation in many firms. A large batch of documents needs reviewing and the work gets assigned without much context. The person allocating the work may not realise that the associate they have chosen is about to go on annual leave or is already managing several urgent matters.

The result is predictable.

• the work cannot realistically be completed before the leave begins
• it gets moved to someone else at the last minute
• the associate is frustrated at being asked to do something impossible
• the colleague who inherits the task is frustrated as well

A system that understands both the matter timeline and team capacity could avoid that entirely.

If Sarah is due to leave for two weeks on Friday and someone tries to assign her one hundred contracts to review on Thursday afternoon, the system should flag that immediately. Not as a passive warning, but as a planning decision.

Possible responses might include:

• suggesting another reviewer with available capacity
• splitting the work across several team members
• scheduling the task earlier in the timeline, if possible
• flagging that the work will extend beyond the planned deadline

None of this requires particularly advanced AI. It requires systems that understand how legal work actually unfolds.

AI That Spots Impossible Timelines

A similar idea applies at the matter level.

Legal matters often change (very) quickly, particularly in transactions or regulatory work. Deadlines move, new documents appear and teams adjust as they go and by the time someone realises the timeline has become unrealistic, the team is already working late to catch up.

A system that understands the structure of the matter could identify problems much earlier.

For example, it might detect that:

• due diligence requires review of 2,000 contracts
• the available team capacity supports review of roughly 1,200 before the target date
• two key reviewers are already committed to other matters during the same period

Instead of discovering the issue during a late night call, the system could surface the constraint immediately. The team can then make an informed decision about how to respond.

How Firms Can Start Approaching This

None of this requires some advanced science fiction technology.

Most of the signals needed to support this kind of intelligence already exist inside firms. They are simply fragmented across different systems.

Legal work today typically sits across:

• document management systems
• matter management tools
• email and collaboration platforms
• time recording systems
• calendar data

Each system holds a small part of the picture. Very few systems understand the whole thing.

If firms want AI to move beyond document analysis, the next step is building or adopting tools that create a system of record for legal work.

That means platforms capable of tracking:

• the structure of a matter
• the tasks within it
• the people responsible
• the deadlines and dependencies

Once that layer exists, AI becomes far more useful because it finally has the context it needs.

For years legal AI has focused on making lawyers faster at reading documents.

The next phase is likely to focus on something different by helping lawyers manage the complexity of their work. Moving from document intelligence to work intelligence

That change may be less visible than a contract summary appearing instantly on screen, but it could ultimately have a much bigger impact on the profession. Once AI understands the structure of legal work, it can start improving not only how tasks are performed, but how that work fits into the lives of the people doing it.

Who knows this may turn out to be the real transformation that is actually needed.