According to data sourced from Harbor’s 2025 Law Department Survey on how top-performing teams are using AI to orchestrate legal work:
45% of legal departments shrank in size last year.
That’s nearly double the proportion from the year before.
And while headcount has shrunk, the workload hasn’t. The business still needs contracts reviewed, questions answered, and risks flagged, just with fewer people and tighter timelines. So, when AI enters the conversation as the solution, it’s hard to argue against it. More output, less manual effort, and faster turnaround sounds like a great deal. But there’s a problem most teams aren’t addressing before they invest, and it’s sitting right at the beginning of all legal operations.
Legal Departments Are Shrinking. Workloads Aren’t.
Last year, more legal departments shrank than grew. 45% reduced headcount compared to 41% that expanded, and that 45% figure is nearly double what it was the year before. Looking ahead, most companies expect lawyer headcount to stay flat, with only 32% planning to add lawyers and 10% planning to cut further. On the surface that sounds stable, but in practice, that means the majority of legal teams are being asked to absorb more work with the same or fewer people.
In terms of market response, departments are being pushed to re-assess internal resources versus outside counsel, prioritize automation, and re-evaluate how work is distributed across the team. These are necessary operational duties and as Tarryn Puzsar, Director of Process Automation at Harbor, put it in a recent webinar on operationalizing AI in legal: “You’re always being asked to do more with less.”
The immediate instinct for most legal departments is to look for an AI solution. But what is often overlooked is the order of operations, and this arguably matters more choosing which AI tool to invest in.
What Is the Legal Front Door, and Why Is It Broken?
Before any legal automation or AI conversation can take place, legal professionals need to ask a more fundamental question: “How does work actually arrive in your legal team?”
For most in-house legal functions, the answer is: however it can. A Slack message here, an email chain there, a calendar invite from a business partner who needs a quick contract review, a verbal request picked up in passing. There’s no standardized intake, triage logic, consistent way to capture what’s coming in, assess its priority or risk, or route it to the right person or process.
This means priorities get set by whoever is loudest or most persistent, work that matters gets delayed, and visibility into legal department metrics is limited. Legal becomes a function that businesses learn to work around rather than with.
This is the front door problem. And it exists in most legal teams regardless of size, industry, or maturity. The absence of a structured entry point doesn’t just create inefficiency, it creates the conditions where every solution you layer on top will underperform.
Why AI Fails When Legal Intake Is Unstructured
Here’s where most legal AI investments go wrong. A legal team under pressure, with a growing backlog and a shrinking headcount, makes the decision to invest in AI tooling. The tool gets deployed, expectations are set, and a few months later, the team discovers that the results are underwhelming. Requests are still getting missed and neglected, response times haven’t drastically improved, and the business is still frustrated.
It wasn’t the AI that failed, it was the process underneath it.
“
If your process is bad, your AI experience will be bad. You can have a terrible AI process layered on a terrible workflow foundation, and that is going to give you poor results.
Tarryn Puzsar Director of Process Automation at Harbor
Legal AI needs structured inputs in order to produce useful outputs. It needs to know what type of request is coming in, what information is relevant, what the priority level is, and where it should be routed. When the front door is broken, none of that exists. Requests are ambiguous, context is missing, and routing is inconsistent.
Essentially, when AI has nothing clean to work with, it produces outputs that are unreliable or incomplete, meaning humans still have to step in to fix, interpret, or redo the work.
The end result isn’t a failed AI tool. It’s a faster, more automated version of the same dysfunction that existed before.
How to Operationalize AI in Legal: Start With Repetitive Requests
In Checkbox’s recent webinar on Operationalizing AI at the Front Door of Legal, attendees were asked to identify their top priority. 42% said handling repetitive requests more efficiently. That result isn’t surprising to anyone who has spent time inside a legal team as routine questions, standard NDAs, low-complexity approvals, and recurring vendor agreements consume a disproportionate share of attorney time. This is typically because there’s no system in place to handle them without pulling in a lawyer every time.
This is exactly where a structured Legal Front Door delivers its clearest return. When requests come in through a centralized intake point, integrated into the communication tools your business already uses, whether that’s Microsoft Teams, Slack, or email, the right information gets captured upfront. Requests get triaged by risk and priority automatically. Routine work gets routed to the right process without a lawyer needing to touch it. The team gets visibility into what’s coming in, what’s in progress, and where the bottlenecks are.
And critically, this is where AI has the most to work with. High volume, predictable patterns, repeatable outcomes. The Legal Front Door creates the structured data layer that makes downstream AI tools useful rather than theoretical.
“
The lowest hanging fruit is to get a good automation tool.
Tarryn Puzsar Director of Process Automation at Harbor
The solution isn’t a more sophisticated AI model or a bigger tech stack, it’s a solid automation foundation that brings order to how work enters the function.
Get that right, and everything that follows (i.e. contract management, workflow automation, reporting, and AI-assisted drafting and review), can actually function as intended.
Key Takeaways
The teams getting the most from AI aren’t necessarily the ones with the most advanced tools. They’re the ones who built the right foundation before they deployed them.
Workflow isn’t the boring prerequisite you get through before the interesting stuff starts. It’s the reason the interesting stuff works at all.
“
It is an investment. But your processes and your department will fail without workflow.
Tarryn Puzsar Director of Process Automation at Harbor
AI is an amplifier that makes good processes faster and more scalable. On the contrary, it can also make broken ones more visibly, more efficiently broken. The Legal Front Door is where that distinction gets made.
If you want to see what the right foundation looks like in practice, and how other legal teams are operationalizing AI in a way that actually holds, book a demo and see it for yourself.
.png)
