AI Doesn't Eliminate Inefficiency. It Can Solidify It.

Does someone clearly own it?

Then:

Is it operating on the right software platform?

Only after those questions are answered do I want to start thinking seriously about embedding AI into the workflow.

Because AI layered on top of inefficiency doesn't eliminate the inefficiency.

It can solidify it.

A Faster Bad Process Is Still a Bad Process

This isn't a new technology problem.

Law firms have been doing this for years.

A process is inefficient.

Instead of redesigning the process, the firm buys software.

Then the software gets configured around the inefficient process.

Now the firm has an inefficient process inside an expensive technology platform.

A few years later, everyone complains about the software.

Sometimes the software is the problem.

Sometimes it isn't.

Sometimes the firm simply digitized a process that never should have existed in that form.

AI creates the potential to take this one step further.

Now we can automate the inefficient process too.

That's not necessarily progress.

The Goal Isn't to Make Your Current Process Faster

This distinction matters.

If a process takes ten steps and AI can help you complete those ten steps faster, that may be valuable.

But what if the process should only have five steps?

What if two approvals are unnecessary?

What if the work is being done by the wrong person?

What if information is being entered twice because two systems aren't integrated?

What if the process exists because of a software limitation?

What if nobody remembers why one of the steps exists?

What if the entire workflow was built around a problem the firm no longer has?

Now making the ten-step process faster isn't necessarily the best answer.

The better answer may be:

Redesign the process.

Then determine where AI belongs in the five-step version.

The goal isn't to make your current process faster. The goal is to determine whether it's the right process in the first place.

Before AI: Process → Ownership → Platform

When I'm looking at where AI should fit into a law firm's operations, I think the order matters.

1. Process

What is the most efficient reasonable way this work should happen?

Not:

"How have we always done it?"

Not:

"How does Susan do it?"

Not:

"How does our current software force us to do it?"

How should the process work if we were designing it intelligently today?

2. Ownership

Who owns the process and the outcome?

Who notices when it stops working?

Who has authority to improve it?

Who monitors performance?

Who is responsible for making sure the process evolves as the firm changes?

3. Platform

What software should support that process?

Does the current system actually fit the firm's needs?

Are we using the capabilities we already pay for?

Are systems integrated appropriately?

Are we entering the same information multiple times?

Is the platform scalable?

Is the team actually using it consistently?

4. AI

Now ask:

Where can AI reduce friction?

Where can it eliminate manual work?

Where can it improve speed?

Where can it improve consistency?

Where can it create capacity?

Where can it improve the client experience?

Where can it help us make better decisions?

That's a very different conversation.

Process first. Ownership second. Platform third. AI fourth.

Don't Automate Around a Process You Already Know Is Broken

Let's use intake as an example.

Suppose your firm has an inconsistent intake process.

Calls are routed differently depending on the time of day.

Follow-up isn't standardized.

Some leads are entered into the CRM.

Some aren't.

Different employees qualify leads differently.

Attorneys handle consultations differently.

Nobody clearly owns conversion.

Reporting is unreliable.

The CRM isn't configured well.

Now leadership decides:

"We should add AI to intake."

Why?

What exactly are we automating?

Before adding another layer of technology, I want to fix the foundation.

Define the stages.

Define qualification.

Determine who owns conversion.

Build the right workflow.

Configure the CRM appropriately.

Automate what the existing platform can already automate.

Establish reporting.

Train the team.

Then ask where AI can make the process substantially better.

Otherwise, we risk automating chaos.

Don't Use AI to Make the Wrong Software Tolerable

I think this is going to become increasingly important.

Law firms already have enormous sunk-cost attachment to software.

"We've used it for eight years."

"Everyone already knows it."

"All our data is there."

"We've customized it."

"I don't want to go through another implementation."

Those are understandable concerns.

But none of them answer:

Is this still the right platform for the firm?

Now imagine adding a significant AI layer on top of it.

Custom workflows.

Automations.

Integrations.

Prompts.

Data connections.

AI assistants.

Maybe third-party tools built specifically around the existing platform.

Suddenly switching software doesn't only mean migrating data and retraining employees.

It means rebuilding an entire AI ecosystem too.

The firm has increased its switching cost.

And if the underlying platform wasn't right to begin with, it has now become even harder to leave.

Don't use AI to make the wrong software tolerable.

Fix the platform first.

Then invest in making the right platform dramatically better.

AI Can Turn Technical Debt Into Operational Debt

Technology teams have talked about technical debt for years.

Law firms can accumulate a similar kind of operational debt.

Every workaround adds complexity.

Every unnecessary handoff adds complexity.

Every manual spreadsheet built because two systems don't communicate adds complexity.

Every exception adds complexity.

Every employee-specific process adds complexity.

Then technology gets layered around those workarounds.

The temporary solution becomes permanent infrastructure.

AI can accelerate that if firms aren't careful.

You don't just have a workaround anymore.

You have an automated workaround.

And automated workarounds can feel sophisticated enough that nobody goes back and asks whether the workaround should exist.

That's how inefficiency becomes institutionalized.

Your Current Software May Already Do More Than You Think

Before buying another AI tool, I frequently want firms to look at what they're already paying for.

I have worked with firms that own sophisticated practice management systems, CRMs, phone systems, accounting platforms, and marketing technology but use only a fraction of their capabilities.

Workflows aren't configured.

Automations aren't turned on.

Integrations aren't connected.

Custom fields aren't being used.

Reporting isn't set up.

Templates aren't standardized.

Employees have created manual workarounds because nobody redesigned the process when the technology was implemented.

Then someone sees a demo of a new AI product and thinks:

That's what we're missing.

Maybe.

But let's first determine whether the technology you already own can solve the problem.

The Shiny Penny Has Gotten Much Shinier

I've written before about law firms chasing shiny pennies.

AI may be the shiniest penny we've ever seen.

That's not because AI lacks value.

It has tremendous value.

It's because the possibilities are so broad that almost any operational problem can now be presented as an AI opportunity.

Slow intake?

AI.

Administrative burden?

AI.

Poor documentation?

AI.

Billing inefficiency?

AI.

Knowledge management?

AI.

Client communication?

AI.

Reporting?

AI.

Maybe.

But sometimes the answer is still:

Your process is bad.

Your software isn't configured correctly.

Nobody owns the outcome.

The wrong person is doing the work.

The team hasn't been trained.

You don't have clean data.

AI doesn't make those problems disappear.

Technology Cannot Replace Operational Leadership

This is one of the biggest mistakes I think firms can make in the next several years.

AI can perform work.

It can assist decisions.

It can identify patterns.

It can summarize information.

It can automate repetitive activity.

It can dramatically increase individual productivity.

But it doesn't eliminate the need for someone to decide:

How should this process work?

Who owns it?

What does good performance look like?

What technology should support it?

What are the guardrails?

What happens when something goes wrong?

How will we measure whether the change worked?

Those are management decisions.

Technology can enable good operations.

It cannot replace operational leadership.

AI Is Leverage. Be Careful What You Leverage.

Leverage is powerful.

That's exactly why the underlying system matters.

If you apply leverage to a strong process, you can get an extraordinary result.

If you apply leverage to a weak process, you can scale the weakness.

That's how I think law firms should approach AI.

The question isn't simply:

"Where can we use AI?"

It's:

"Where do we have a strong enough foundation that AI can create meaningful additional leverage?"

Those aren't the same question.

The Firms Best Positioned for AI May Be the Ones That Need It Least

There's an interesting irony here.

The firms most prepared to capture significant value from AI may already have:

clear workflows.

documented processes.

clean data.

defined ownership.

appropriate technology.

consistent adoption.

measurable outcomes.

They know how work moves through the organization.

They know where bottlenecks exist.

They know which activities consume the most labor.

They know what things cost.

That makes it much easier to identify where AI could create real economic value.

Meanwhile, a firm with operational chaos may look at AI as the solution to the chaos.

But without understanding the existing process, how will it know what should be automated?

Without reliable data, what will the AI use?

Without clear ownership, who is responsible for the result?

Without measurable outcomes, how will leadership know whether AI improved anything?

Foundation matters.

That Does Not Mean You Should Wait to Start Using AI

This is where I want to be very clear.

I am not suggesting law firms spend the next three years perfecting every process before anyone is allowed to touch AI.

Quite the opposite.

I think firms should be learning how to use it now.

The distinction I would make is between:

using AI as a tool

and

embedding AI into your infrastructure.

Those have very different levels of commitment and risk.

Start With One-Off Administrative Work

One of the easiest places to begin is individual productivity.

Use AI for work that doesn't require redesigning your entire operating system.

For example:

Turn messy meeting notes into organized action items.

Create the first draft of an internal email.

Summarize a lengthy document for internal review.

Organize information from several sources.

Analyze a spreadsheet.

Create a meeting agenda.

Turn rough notes into a checklist.

Draft training materials.

Create a first version of an SOP.

Compare information.

Brainstorm questions.

Help organize a project plan.

Take a large amount of information and identify themes.

These uses allow your team to develop AI literacy while potentially saving meaningful administrative time.

And if one experiment doesn't work?

Stop doing it.

You haven't rebuilt your core infrastructure around it.

Use AI to Help Build the Foundation

This is the part I think firms may be overlooking.

You don't have to wait until your operational foundation is complete to use AI.

Use AI to help you build the foundation.

Your process isn't documented?

Describe what happens today and use AI to help organize it into an SOP.

Your workflow feels clunky?

Map the current steps and use AI to help identify redundancies, unnecessary handoffs, and questions leadership should investigate.

You're evaluating software?

Use AI to help develop requirements.

Have it organize your must-haves versus nice-to-haves.

Use it to develop questions for software demonstrations.

Use it to compare functionality based on the information you've gathered.

Need to formalize a role?

Use AI to help turn scattered responsibilities into a clearer job description or accountability framework.

Need to train employees on a new process?

Use AI to help create training materials, checklists, scenarios, and FAQs.

Trying to improve reporting?

Use AI to help think through what information might answer a particular business question.

That's an excellent use of AI while the foundation is still being built.

You don't have to wait to use AI while you build the foundation. Use AI to help you build the foundation.

AI Can Help You Ask Better Questions About Software

This is another area where I think it can be useful.

Law firms frequently buy software based on demos.

The vendor shows an impressive platform.

Leadership gets excited.

Then implementation begins and someone realizes:

"Oh. It doesn't actually do this thing that's incredibly important to us."

Part of the problem is that the firm didn't clearly define its requirements before shopping.

AI can help with that preparation.

You can use it to organize:

current workflows.

pain points.

integration requirements.

reporting needs.

user roles.

automation requirements.

client-facing needs.

security questions.

implementation concerns.

migration issues.

Then walk into software evaluations with a much better list of questions.

AI doesn't make the software decision for you.

But it can make you a much more informed buyer.

Use AI to Challenge Your Existing Processes

Here's another practical exercise.

Before automating a workflow, document the current process.

Then interrogate it.

Why does this step exist?

Why does this person touch it?

Why does this require approval?

Could this information be entered once instead of twice?

Could the software trigger this automatically?

Could this decision happen earlier?

Could this work be moved to a different role?

Could two steps be combined?

Could one be eliminated?

Does the client need to be involved here?

Is this exception still necessary?

What would happen if we removed this step entirely?

AI can help generate those questions.

The answers still require human judgment.

But simply forcing the organization to examine the workflow before automating it can reveal substantial inefficiency.

Then Move From AI Tool to AI Infrastructure

Once the foundation is stronger, the opportunity gets much more interesting.

Now you can begin asking where AI should be embedded into recurring workflows.

Perhaps it supports intake.

Perhaps it assists with internal knowledge retrieval.

Perhaps it helps prepare matter summaries.

Perhaps it reduces administrative steps in a recurring process.

Perhaps it assists with data analysis.

Perhaps it helps standardize internal communication.

Perhaps it eliminates repetitive manual work.

The specific opportunity will vary by firm and practice area.

But now AI is sitting on top of something intentionally designed.

That's the difference.

Think About AI Adoption in Stages

I would think about it roughly this way.

Stage 1: AI as an Individual Productivity Tool

Let people learn.

Experiment with appropriate one-off tasks.

Save administrative time.

Build familiarity.

Understand strengths and limitations.

Stage 2: AI as an Operational Design Tool

Use AI to help document processes.

Identify inefficiencies.

Formalize SOPs.

Evaluate software requirements.

Create training.

Organize data.

Improve the foundation.

Stage 3: AI Embedded Into Workflows

Once the process, ownership, software, data, and desired outcome are reasonably sound, begin incorporating AI into recurring operational workflows.

Stage 4: Measure the Economics

This is the stage I don't want firms to skip.

Did it actually work?

"We Implemented AI" Is Not a Business Outcome

I don't care whether your firm can say it uses AI.

I care what happened because you used it.

Did you reduce administrative labor?

Did you increase attorney capacity?

Did turnaround time improve?

Did conversion improve?

Did errors decrease?

Did client experience improve?

Did employees spend less time on low-value work?

Did the firm avoid a hire?

Did realization improve?

Did profitability improve?

Did attorneys get more time for higher-value work?

Technology adoption isn't the outcome.

The business result is the outcome.

Measure the Before and After

This is why implementing AI on a defined process is so much more valuable.

You can measure it.

Before AI:

This process required four hours per matter.

After AI:

It requires two.

Now we know something.

Before AI:

The team took 24 hours to complete a particular administrative step.

After AI:

Four hours.

Useful.

Before AI:

An attorney spent five hours a week on a repetitive administrative task.

After AI:

One hour.

Now we can calculate the capacity created.

Without a baseline, firms can end up spending significant money on AI without knowing whether it actually improved anything.

Don't Confuse Activity With ROI

AI tools are going to produce impressive usage statistics.

Number of prompts.

Documents processed.

Hours supposedly saved.

Automations run.

That's interesting.

But eventually, I want the economic question answered.

What changed in the business?

If AI theoretically saves 500 hours but nobody's workload changes, no additional capacity is used, no headcount is avoided, no revenue increases, no turnaround improves, and no client outcome changes, what did those 500 hours actually create?

That's the conversation law firms need to have.

Don't Use AI to Avoid a Management Problem

There will also be situations where AI becomes an attractive workaround for something leadership doesn't want to address.

An employee is inefficient.

Let's automate around them.

Attorneys won't follow the process.

Let's build more technology around the exceptions.

Nobody wants to own intake.

Let's buy an AI intake tool.

Partners won't delegate.

Let's give them AI.

Sometimes technology helps.

Sometimes you're spending money to avoid managing people.

Know the difference.

AI Should Make Good Employees More Valuable

One of the things that excites me most about AI isn't necessarily replacing people.

It's increasing the leverage of strong people.

An excellent operations employee who can use AI to analyze information, document processes, prepare projects, organize data, and eliminate repetitive administrative work can potentially create far more value.

A strong attorney who can reduce low-value administrative work can spend more time on substantive legal work, client relationships, or business development.

A great manager who spends less time producing reports can spend more time actually managing.

That's where I see enormous opportunity.

But again, you have to know what you want people doing with the capacity you create.

If AI Saves Time, Decide Where the Time Goes

This sounds obvious.

It isn't.

Suppose AI saves each attorney three hours per week.

Fantastic.

Now what?

More billable work?

Business development?

Client relationship work?

Training?

Higher matter capacity?

Reduced headcount needs?

Better work-life balance?

There isn't one right answer.

But leadership should know what it's trying to accomplish.

Otherwise "time savings" becomes another vanity metric.

Don't Wait for Perfect. Do Be Intentional.

Your processes will never be perfect.

Your software stack will continue evolving.

AI will certainly continue evolving.

So don't interpret foundation-first as:

"We'll get to AI when everything else is finished."

You'll never get there.

Experiment now.

Learn now.

Let your people develop fluency.

Use AI for low-risk administrative work.

Use it to help document and improve your operations.

Use it to become a better buyer of technology.

Use it to identify opportunities.

Just be more disciplined before you deeply embed AI into a recurring core process.

There's a big difference between experimentation and institutionalization.

Ask This Before You Automate Anything

Before integrating AI into a recurring workflow, I'd ask:

Is the current process actually efficient?

Does every step still need to exist?

Is the work being performed by the right person?

Does someone clearly own the outcome?

Does that person have appropriate authority?

Are we using the right underlying software platform?

Are we using that platform effectively?

Is the underlying data reasonably reliable?

What specific problem are we asking AI to solve?

How will we know whether it worked?

If you can't answer those questions, you may not be ready to automate the process.

You may be ready to redesign it.

The Foundation Still Matters

AI changes a lot.

It doesn't change the fundamentals of running a good business.

You still need efficient processes.

You still need clear ownership.

You still need appropriate technology.

You still need reliable information.

You still need accountability.

You still need leadership.

AI can make a strong operating model dramatically more powerful.

But it doesn't eliminate the need for the operating model.

That's why I don't think the right question for law firms is:

"How quickly can we adopt AI?"

I think it's:

"Where are we operationally ready to leverage AI—and where should we use AI to help us get ready?"

Use it.

Experiment.

Learn.

Let it eliminate tedious work.

Let it help you build processes.

Let it help you evaluate technology.

Let it help your best people become more productive.

But before you wire AI deeply into the way your firm operates, make sure the foundation underneath it is one you actually want to keep.

Because AI layered on top of inefficiency doesn't eliminate the inefficiency.

It can solidify it.

Get the foundation right.

Then use AI to make a good operating model exponentially better—not to make a bad one harder to unwind.

If your law firm is exploring AI but your processes, ownership structure, software platforms, or reporting still need work, don't treat those as separate projects.

They're connected.

At ING Collaborations, I help law firms build the operational foundation first: efficient processes, clear ownership, appropriate technology, useful reporting, and accountability.

From there, technology—including AI—can create real leverage instead of adding another layer of complexity.

The goal isn't to implement more technology.

It's to build a better law firm.