What 44 Lawyers Said About AI: The Access vs. Workflow Gap

The Survey Nobody Commissioned — And Why That Makes It More Valuable

Earlier this year, 44 practicing attorneys — mostly in-house counsel and small firm lawyers — answered three questions before walking into a legal AI meetup: What AI products are you using right now? Is AI actually saving you time? What do you think the future looks like?

The answers were more honest than anything you will find in a commissioned industry report. Commissioned surveys are designed to produce headlines. This one was designed to get people into a room. That difference matters.

But here is the part most coverage of legal AI gets wrong. The story is not about which tool is winning. It is about why the same tool produces radically different outcomes for different lawyers — and what that gap reveals about where most firms are actually stuck.

Copilot Is Everywhere. That Is Precisely the Problem.

The most-used AI tool in the room was not Harvey, CoCounsel, or even ChatGPT. It was Microsoft Copilot — by a wide margin.

That result is not surprising. Copilot is bundled into Microsoft 365, which most firms and legal departments already pay for. The adoption barrier is essentially zero. IT flips a switch, and suddenly every attorney in the organization has "AI."

Here is the problem with that framing. The "not really saving time" answers in the survey came almost entirely from Copilot users. When you examine the pattern across responses, a clear line emerges: lawyers who deliberately chose a tool reported time savings at a significantly higher rate than those who simply had Copilot enabled by their IT department.

Flipping a switch to enable Copilot is not an AI strategy. It is a license. The distinction matters enormously — and most firms have not made it yet.

The Real Divide: Passive Access vs. Deliberate Integration

This is the distinction that the headlines about legal AI consistently miss.

Passive access means the tool is available. It shows up in your toolbar. You can open it. You might use it occasionally to summarize an email or draft a quick paragraph. You have not changed how you work. You have added a feature.

Deliberate integration means something different entirely. The lawyers in this survey who reported genuine time savings — those using Harvey), CoCounsel, or Claude as their primary tools — shared a common profile:

• They had invested time in prompt engineering specific to their practice area
• They had built project-based workspaces or matter-specific contexts
• They had developed personal or team workflows around the tool, not alongside it

This is the AI in legal practice lens that matters right now. The question is not which tool your firm has purchased. The question is whether anyone has designed a workflow around it.

The gap between having access and having a workflow is where most firms are stuck. And that gap does not close on its own.

Time Savings Are Real — But Only for the Intentional Minority

Roughly half of the 44 respondents reported that AI is genuinely saving them time. That number sounds encouraging until you look at who is in that group.

The time savings are concentrated among lawyers who made an active choice. They selected a tool based on their specific work — document review, contract analysis, legal research — and then built a practice around using it well. CoCounsel, built by Casetext and now part of Thomson Reuters, is purpose-designed for document review and legal research. Harvey), developed by Counsel AI Corporation, is built specifically for legal professionals. These are not general-purpose chatbots with a legal skin. They are tools designed around legal workflows.

The lawyers who are not saving time are largely those who received AI passively — through a Microsoft 365 bundle — and were never given a reason to change how they work. That is not a technology failure. It is a workflow design failure.

The real question is not whether AI saves time. It is whether your firm has invested in the workflow design that makes time savings possible.

What Intentional AI Adoption Actually Looks Like

If the survey data points to one actionable conclusion, it is this: the firms and legal departments getting real value from AI are treating it as a practice design problem, not a procurement problem.

Here is what that looks like in practice:

Step 1: Choose tools based on your actual work, not your existing licenses

If your practice involves heavy document review or contract analysis, a purpose-built tool like CoCounsel will outperform a general assistant every time. If you are doing legal research and drafting, Harvey or Claude may be the better fit. The tool should follow the workflow, not the other way around.

Step 2: Invest in prompt engineering before you invest in more tools

The lawyers reporting the highest time savings are not using more tools. They are using fewer tools better. Prompt engineering — learning how to give an AI model the context, constraints, and output format it needs — is a skill that compounds. It is also one that most firms have not formally developed.

Step 3: Build team-level workflows, not individual habits

Individual power users do not scale. The firms that will pull ahead are the ones that document their AI workflows, share them across the team, and treat them as institutional knowledge. A workflow that lives in one attorney's head is not a competitive advantage. It is a single point of failure.

The bottom line: AI adoption in legal practice is not a technology problem. It is a legal engineering problem. And legal engineering requires intentional design.

Key Takeaways

• Passive AI access produces passive results. The survey data shows that Copilot users — who received AI through a Microsoft 365 bundle rather than by deliberate choice — reported time savings at a significantly lower rate than lawyers who actively selected their tools.
• Tool selection should follow workflow design, not precede it. Lawyers using purpose-built tools like Harvey and CoCounsel reported better outcomes because those tools were chosen to match specific legal tasks, not because they were available by default.
• Prompt engineering is a professional skill, not a technical curiosity. The attorneys getting the most value from AI have invested time in learning how to communicate with these models effectively. That investment pays compounding returns.
• Team-level workflows are the real competitive advantage. Individual AI power users create fragile systems. Firms that document, share, and institutionalize AI workflows build durable capacity.
• The access-vs-workflow gap is where most firms are stuck right now. Closing that gap requires treating AI adoption as a practice design problem — not a procurement decision or an IT rollout.