TL;DR
- Legal work is personal. Two firms drafting the same contract will produce meaningfully different documents. Style, etiquette, perspective, and judgment vary from practitioner to practitioner.
- Generic AI tools standardize what shouldn't be standardized. They impose uniform workflows on a profession where differentiation is the product.
- 90% of AI use cases never leave pilot. McKinsey found that about 90% of function-specific AI use cases remain stuck in pilot mode, and nearly 8 in 10 companies report no significant bottom-line impact.
- It's the workflow, not the agent. McKinsey's analysis of 50+ agentic AI deployments found that efforts focused on reimagining entire workflows are more likely to deliver positive outcomes than those focused on the tool itself.
- Build AI around how your team actually works. The firms seeing real efficiency gains are encoding their own style, processes, and standards into AI, not adapting their practice to fit a tool.
The Standardization Trap
From the outside, legal work looks like it should be easy to automate. There are rules, statutes, precedents, regulatory frameworks. A contract is a contract. A compliance review follows a checklist. It all seems standardized.
Spend time with actual legal practitioners, and that assumption falls apart quickly.
One firm's approach to drafting a services agreement reflects decades of accumulated preferences: how they handle limitation of liability clauses, the tone they strike in indemnification language, the level of detail they include in scope definitions. A legal counsel at a different firm, drafting the same type of agreement, makes different choices. Not because one is right and the other wrong. Because legal practice involves judgment, style, and perspective that are deeply personal to the practitioner and the firm.
There is emotion in how a seasoned attorney structures an argument. There is etiquette in how they communicate with opposing counsel. There is an institutional voice that distinguishes one firm's work product from another's.
This is not a flaw in legal practice. It is the practice. The differentiation between legal professionals is precisely what clients pay for.
And this is where generic AI tools run into trouble.
Why Generic AI Disappoints Legal Teams
Most legal AI products are built on a reasonable-sounding premise: standardize common legal tasks, and everyone becomes more efficient.
The problem is that standardization conflicts with how legal professionals actually work. When a tool generates a "standard" first draft of a contract, it reflects an average of many approaches rather than the specific approach your firm has refined over years. The practitioner then spends time undoing the tool's assumptions and reshaping the output to match their own standards.
That is not efficiency. That is extra work with a technological middleman.
The numbers confirm the pattern. McKinsey reports that nearly eight in ten companies using gen AI report no significant bottom-line impact, and about 90% of function-specific AI use cases remain stuck in pilot mode. Legal teams are not immune to this. A tool that generates generic output your attorneys then rewrite is a tool stuck in permanent pilot.
The gap between "generic output" and "our firm's standard" is where time savings evaporate. Every edit to align AI-generated text with your firm's voice, every restructured clause, every reworked paragraph is time that was supposed to be saved.
Gartner's Hype Cycle for Legal, Risk, Compliance and Audit Technologies highlights this shift, noting that as legal teams gain hands-on experience with gen AI, expectations are moving from hype to a pragmatic focus on measurable outcomes and sustainable adoption. The honeymoon phase is over. Legal leaders want results, not demos.
It's Not About the Agent. It's About the Workflow.
Here is the data that should change how legal firms think about AI investment.
McKinsey's analysis of more than 50 agentic AI deployments found a clear pattern: efforts that focus on reimagining entire workflows, including people, processes, and technology, are more likely to deliver positive outcomes than those focused on deploying a single tool. Their conclusion is direct: "It's not about the agent; it's about the workflow."
The same research reveals a critical insight for legal teams: high-variance, low-standardization workflows benefit most from AI agents, while low-variance, high-standardization workflows where agents based on nondeterministic LLMs could add more complexity than value. Legal work sits in a unique position. The tasks look standardized (contract drafting, compliance review, due diligence), but the execution is highly personal and variable.
What does this mean for your firm? A generic tool treats legal drafting as a low-variance task and applies a standardized approach. But your practitioners know that every engagement, every client, every opposing counsel brings variation that demands judgment. The AI needs to account for that variation, not flatten it.
McKinsey's research on custom-built agents reinforces this: realizing the full potential of agentic AI requires agents deeply aligned with the company's logic, data flows, and value creation levers, making them difficult to replicate and uniquely powerful. For legal teams, your "logic" is how your practitioners think. Your "data flows" are your precedent libraries, clause banks, and client histories. Your "value creation" is the distinct quality of your work product.
What Custom AI Looks Like for Legal
Custom AI for legal is not about building a model from scratch. It is about configuring AI to work the way your team works.
The practitioner is not replaced by the tool. The practitioner's expertise is encoded into the tool.
What this looks like in practice:
- Firm-specific drafting standards baked in. Your preferred clause structures, your tone, your formatting conventions are built into the AI's workflow, not applied as afterthought edits.
- Practitioner-level preferences captured. Senior attorneys' judgment about how to handle specific scenarios becomes part of the system, so junior team members produce work that reflects firm standards from the first draft.
- Workflow sequences, not single prompts. Instead of "generate a contract," the process becomes a series of steps: intake, clause selection based on deal type, risk flagging against your firm's thresholds, review checkpoints aligned to your approval process.
- Continuous refinement through use. As practitioners interact with the system, their feedback sharpens it. The AI gets better at producing work that matches your firm's specific standards over time.
Gartner reports that 36% of General Counsel are now focused on adopting AI, building AI skills, or improving AI risk management. The investment is happening. The question is whether that investment goes toward another generic tool that your team works around, or toward AI that genuinely encodes how your firm practices.
62% of organizations are at least experimenting with AI agents, and 23% are already scaling agentic AI systems. But scaling without workflow alignment is how you end up in the 90% that never leave pilot.
Before You Invest in Legal AI
The next time a vendor pitches you a legal AI product, ask these questions:
- Does this tool adapt to our firm's style, or do we adapt to it? If the answer is "you'll need to adjust your processes," you are buying someone else's workflow.
- Can it encode our senior practitioners' judgment? The value of your firm lives in how your best people approach problems. If the AI cannot capture that, it is replacing your differentiation with a generic average.
- Does it redesign our workflow or just add a step? A tool that generates a first draft you then rewrite is not saving time. A tool that produces output aligned with your standards from the start is.
- What does the pilot period look like? If the vendor expects plug-and-play deployment, they are not accounting for the customization that legal work demands.
Legal AI has genuine potential to reduce the time practitioners spend on repetitive, low-value tasks and free them for the judgment-intensive work that clients actually value. But that potential is only realized when the AI is built around how your team works, not when your team reshapes their practice to fit the tool.
The firms that get this right will not just be more efficient. They will be more distinctly themselves, with AI amplifying the expertise and style that makes their practice worth hiring.