opendata
Open menu

AI Is Not Just Replacing Legal Tasks. It Is Compressing the Invisible Workflow Around Them.

7 min read
AI Is Not Just Replacing Legal Tasks. It Is Compressing the Invisible Workflow Around Them.

By the time law firms change the org chart, the workflow, pricing logic, and role boundaries have usually changed first.

The most misleading way to talk about AI in legal and professional services is to ask whether it is “replacing lawyers.”

That framing is too shallow.

The first real impact is happening somewhere else:

not in the title,
not in the official headcount,
but in the invisible workflow around legal work.

Preparation gets compressed.
Research gets accelerated.
RFP work gets standardized.
Client intelligence gets scored.
Document review gets re-sequenced.
Routine drafting gets absorbed into secure internal workspaces.
Support work, handoffs, and synthesis layers start collapsing into fewer steps and fewer people.

In other words:

AI is not only automating legal tasks.
It is compressing the workflow architecture that used to sit around those tasks.

That distinction matters, because workflow compression arrives before formal restructuring. By the time a firm rewrites job titles, revises staffing ratios, or changes compensation logic, the economics of the work may already have moved.

And the public evidence now points in exactly that direction.

Thomson Reuters’ 2026 AI in Professional Services Report says the era of early AI adoption has passed. The sector is now entering a strategic phase in which firms are redefining workflows, reshaping value, and building AI into the foundation of business strategy.

That is not a theory anymore. It is operational reality.

The shift is already visible in law firm workflows

The most revealing public examples are not the flashy “AI lawyer” headlines.
They are the quieter workflow cases.

At Baker Botts, the data science team built an internal AI-assisted client and opportunity intelligence tool using Gemini plus public, financial, and firm data. A process that previously took about 24 hours of research was cut to a workflow where lawyers and business developers can get a scored readout in minutes.

At Ballard Spahr, AI was used to turn fragmented RFP history into a searchable, reusable repository. Partners estimated roughly two hours saved per RFP once tagging and reuse were built into the system.

At Orrick, the firm’s GenAI-driven eDiscovery approach was publicized as reducing review time by 88%, while the same announcement explicitly linked AI not just to eDiscovery but to lawyer training and staffing.

At Foley, the winning AI implementation was not presented as a sidecar chatbot. It was described as a secure internal workspace used for legal strategy development research, contract review, and day-to-day legal workflows.

These are not isolated examples of “people using AI.”

They show something more important:

the sequence of work is changing.

The work that used to pass through multiple human layers — collecting, preparing, synthesizing, routing, reviewing, summarizing, reusing — is being compressed into fewer steps, fewer touchpoints, and fewer role transitions.

That is why workflow compression is a better lens than replacement.

The first roles affected are often not the ones people expect

When people imagine legal AI disruption, they often imagine one of two extremes:

either AI fully replaces junior lawyers,
or it barely changes anything because legal judgment remains human.

Both views miss the middle.

The first roles and functions under pressure are often the ones built around:

  • preparation
  • synthesis
  • coordination
  • handoff
  • support
  • reuse
  • workflow management

These are not always the most junior roles, and they are not always the least skilled.
They are often the least visible layers of knowledge work: the work that makes expert work flow.

In professional services, a huge amount of organizational value is embedded in these layers:

  • getting the right material in front of the right lawyer quickly
  • standardizing repeatable responses
  • accelerating client readiness
  • reducing friction between matter teams, research functions, business development, practice support, and operations

That is exactly the kind of terrain where AI does damage first.

Not because it can own final responsibility,
but because it can absorb or compress much of the surrounding support structure.

This is why role boundaries blur before headcount visibly moves.

Morgan Lewis’ Chief AI Officer put it plainly in March 2026: staffing models will evolve to reflect hybrid skill sets, with greater emphasis on judgment and strategic advisory work.

That one sentence explains a lot.

The firm of the near future is not simply “lawyers plus software.”
It is a different bundle of roles:

  • legal + data
  • legal + workflow design
  • legal + product
  • legal + knowledge systems
  • legal + governance

The hybrid layer grows.
The invisible support layer compresses.
The pure title taxonomy starts to lag reality.

Governance is not keeping up with usage

Another reason this matters: adoption is moving faster than structure.

The ABA summary of the 2026 8am Legal Industry Report says 69% of legal professionals personally use GenAI tools such as ChatGPT, Gemini, or Claude for work-related purposes.

But the same summary says 54% of firms have provided no training on responsible GenAI use and have no current plans to do so.

That gap is not trivial.

It means firms are already living through workflow change without having fully defined:

  • where AI belongs
  • where it does not
  • what must be checked
  • what can be delegated
  • what must be disclosed
  • how quality, confidentiality, and privilege are protected

Thomson Reuters’ 2026 report reinforces the same point from a management angle. It notes that 82% of respondents either say their organizations are not collecting ROI metrics around AI or are unsure whether such metrics are being collected. The report’s own conclusion is sharp: the operational impact of AI remains largely divorced from the business impact of AI.

This is one of the most important facts in the entire discussion.

Many firms are no longer asking “Can we use AI?”

They are already using it.

The real issue is that they still do not have strong enough language, metrics, and governance for the workflow changes already underway.

So the risk is not only reckless AI use.
It is unmanaged workflow change.

Client pressure is forcing the redesign

The legal profession is not changing in a vacuum.
Clients are forcing the issue.

Deloitte’s 2026 legal AI material reports that clients are increasingly:

  • performing simple legal tasks themselves via self-service or AI tools
  • demanding faster turnaround enabled by AI
  • demanding more transparency about AI risks and due diligence
  • expecting lower rates when AI is used

Only 3% of respondents in that material said they had observed no client-driven change.

Thomson Reuters’ 2026 AI in Professional Services Report adds another layer: around 40% of respondents say their firms have been told both to use AI and not to use AI, depending on the client and the project.

This is a defining feature of the next phase.

AI is not creating one new norm.
It is creating segmentation.

Some clients will demand AI-enabled speed and cost compression.
Some clients will demand tighter control, explicit limits, and human review.
Many will demand both at once:
be faster, be cheaper, and also be more accountable.

That combination breaks the old comfort zone.

It means the future of legal service delivery is not just about productivity.
It is about redesigning work in a way that still preserves trust, explainability, professional duty, and client confidence.

Training matters more than most firms admit

There is also a deeper misconception in the legal AI conversation:

that once a capable model is available, productivity will naturally follow.

A 2026 randomized study on generative AI in legal analysis suggests otherwise.

Untrained LLM access proved counterproductive. Participants with untrained access wrote shorter answers, made more case misstatements, and performed worse than the trained group. A brief training intervention increased adoption and improved output quality.

That matters because it points to a deeper organizational truth:

AI gains do not come from access alone.
They come from trained use inside redesigned workflows.

This is why simplistic “AI will replace X” narratives are so weak.
The real dividing line is not between users and non-users.
It is between organizations that re-architect work and those that simply layer tools onto old habits.

The real restructuring is economic before it is organizational

The reason workflow compression matters so much is that it changes economics before it changes titles.

When a 24-hour preparation process becomes a two-minute scored readout,
when review time drops by 88%,
when routine drafting moves into an approved secure workspace,
when clients expect lower fees for AI-assisted work,
the question is no longer just:

How much time does this save?

The question becomes:

Which work remains billable?
Which work remains differentiating?
Which work remains human because it creates trust, judgment, defensibility, and accountability?

This is where the next real split in professional services appears.

Some work becomes cheaper and faster.
Some work becomes more strategic and more expensive.
Some work becomes internalized by clients.
Some work becomes re-bundled into hybrid operational roles.

That is why the org chart changes late.

The workflow changes first.
The economics change second.
The role design changes third.
The title catches up last.

The most valuable human layer is narrowing upward

What becomes more valuable in this environment?

Not generic task execution.

The human layer that gains value is the layer tied to:

  • judgment
  • accountability
  • strategic framing
  • client trust
  • defensible review
  • exception handling
  • governance
  • cross-functional orchestration

This does not mean junior or support roles disappear overnight.
It means the old path through those roles becomes less stable as a training and value ladder.

And that has long-term consequences.

If AI compresses the support and preparation layers too aggressively, firms may discover that they have improved throughput while weakening capability formation.

That is the deeper structural risk.

Not simply job loss.
But the erosion of the developmental layers that used to produce future experts.

Final point

The cleanest way to describe what is happening in legal and professional services is this:

AI is not just replacing legal tasks.
It is compressing the invisible workflow around them.

And that compression is forcing firms to confront harder questions about pricing, governance, staffing, training, responsibility, and client trust.

By the time the org chart changes, the workflow has usually already changed.

That is the shift serious firms need to understand.

Sources

Ready to try the data?

Start exploring public data APIs with a free search

Start with a free search
Back to blog
AI Is Not Just Replacing Legal Tasks. It Is Compressing the Invisible Workflow Around Them. · opendata