Seven percent.
That is the share of law firms with a documented, active AI roadmap for marketing and business development, according to the 2026 State of the Industry Report that InnovAItion Partners and Gladstone Growth Strategies published in October.
The most-quoted number in the same report is AI use for pitch, proposal and RFP support, which went from 42% of firms in 2025 to 79% this year. Yet only 32% of firms report an AI-assisted workflow for that same pitch work. Just 17% run any multi-step workflow, either as a pilot or in regular use.
Nearly every department now has the tools. Very few have changed how the whole department works with them, beyond the few people who figured it out on their own.
I am running AI transformation programs with marketing and BD departments at several law firms, so I have a stake in this argument.
The department learned less than its people did
You probably have someone on your team who has figured it out. They get a usable first draft of a directory submission in a fraction of the time it used to take. They have a prompt for partner bios that follows the style guide better than half the bios on the website. They built it on a Sunday, it lives in their chat history, and when they take a week off, the department goes back to doing it the old way.
The survey shows where that kind of individual adoption stops. In 46% of firms, the person using the tool is the one who reviews the work. Only 28% offer role-specific AI training for marketing and BD. Just 18% offer training tied to specific workflows. At 38% of firms, the content people need is scattered across systems. And 60% have no measurement of results, or only anecdotes.
When one person finds an approach that works, the department has no way to turn it into team practice. Review stays personal and knowledge stays in private folders. Training stops at the introduction, so the next person starts from scratch. Leadership can't show the firm whether any of it is paying off, because no one ever standardized it enough to measure.
I don't count that as anyone's failure. The tools arrived faster than any department could reorganize around them. Firms handed MBD teams a license and a policy and asked them to find the value while still turning around the same volume of pitches.
The knowledge has to belong to the department
When an approach lives in one person's chat history, the department loses it when that person leaves. Holly Chen, VP of Growth Marketing at Samsara, described exactly this at MAICON, the Marketing AI Conference, in October: an agent that stopped working when the colleague who built it went on leave.
The fix is a single body of knowledge the whole department works from, readable by people and by every AI tool the firm licenses. What goes in is material the department already depends on, from approved bios and matter descriptions to the firm's style rules and the prompts teams reuse. Nothing goes in until someone has approved it, because a knowledge base people can't trust is one they stop opening.
Two design choices decide whether anyone uses it.
The first is ownership. The firm owns the knowledge and licenses the tools. Tool choices are still far from settled across firms: Copilot at 78% of firms, legal-grade platforms at 48%, Claude at 38%, ChatGPT at 36%. Some firms open their legal-grade platform to marketing and BD. Others reserve it for lawyers. A department that keeps its approved language inside one vendor's workspace rebuilds it every time the firm changes vendors. A department that stores its approved language where every licensed tool can read it can run each job on whichever platform it has access to, then file the finished work back to the document management system as the record.
The second is a path from one person's idea to department practice. Someone has to decide what the team adopts and keep it working when the person who built it is out. That owner is usually marketing operations. The survey ranks marketing operations the second most common owner of MBD AI overall. At firms with more than 500 lawyers, it ties with senior MBD leaders.
Rules people can follow without asking permission
Shared knowledge brings a question of control with it. Once everyone can reach the department's best prompts and its approved client language, someone has to decide what can be used where.
Firms have answered part of that. Ninety-two percent now have an AI use policy; 87% apply one policy to every employee. That is sensible governance. It also means a coordinator summarizing internal meeting notes works under the same rules as a lawyer drafting a brief. Under rules written for lawyers, every new use needs a permission request, and people either stop or switch to a tool nobody approved. Either way, leadership loses sight of what is in use.
Claire Potter-Schneider, who led marketing AI transformation at Motorola Solutions and now leads it at Dynatrace, laid out the alternative at MAICON: published guardrails instead of approval gates. Leadership writes the rules once and tiers the work by risk. Translated to a law firm marketing and BD department, it looks roughly like this:
- Low risk. Internal work such as meeting summaries, research digests and first drafts for internal review. No approval required, and routine use encouraged.
- Medium risk. Anything clients or prospects will see, including client alerts, pitch materials, directory submissions and web content. Department review for accuracy, brand and quality.
- High risk. Confidential client information, and anything that acts without human review. Escalated to the firm's AI governance, with Legal and IT.
Tiering makes the existing policy usable for marketing and BD work without asking a CMO to rewrite firm policy, which few are in a position to do. From there, each documented workflow states where a person leads, where a person reviews and what can run on its own. Seventeen percent of firms report no formal review process for AI-assisted MBD work. A review line in each workflow names who checks the output and what they check it against.
The drafting that trained junior staff is the drafting the tools now do
Once low-risk drafting runs without approval and the department's best material is one search away, the tools now do most of the work that used to teach the craft. First drafts of bios, RFP responses and directory submissions. Research ahead of a pitch. A coordinator learned what a strong submission looks like by writing a great many weak ones and getting them back covered in a senior colleague's edits.
Paul Roetzer , founder and CEO of SmarterX and the Marketing AI Institute, described this at MAICON as the bottom rungs of the career ladder disappearing, and proposed three roles in their place. Adapted for a law firm department:
The architect, usually marketing operations or a senior MBD leader, decides which workflows to standardize first, owns the knowledge base and its approval path, and writes the specifications the firm's innovation or IT team builds against.
The orchestrator, typically a proposal manager, practice-group BD manager or communications lead, runs each workflow from intake to delivery, checks output against firm standards and decides which steps need a lawyer's review.
The apprentice, a coordinator or specialist in their first years, reviews drafts against firm standards and works beside senior colleagues on live pitches and submissions. Apprentices are assessed on the quality of their review rather than the volume of their drafting, so reviewing drafts is how they learn the craft.
People need a way to grow into these roles. I would base advancement on what someone has built that colleagues now use, rather than hours spent in a course.
The training should run on the department's own live work, so each session leaves something reusable in the shared knowledge base. I wrote last month about teaching judgment on purpose by putting an AI draft beside the version a senior colleague would have sent, with the senior person explaining the difference in a channel the whole team can read.
Holly Chen's advice on hiring was simple: don't go looking for AI-native people; make the team you already have AI native. Her marketing organization went from about 3% of staff using AI tools to more than 90% building their own workflows and agents in roughly a year, supported by departmental champions, a shared list of the agents people had built, and monthly demonstrations in existing meetings.
A pilot ends on a date
Everything above takes longer than a pilot. A pilot tests a tool and stops on a date. This changes how a department does its daily work, so from the first week the plan has to include handing ownership to the department's own marketing operations team. The goal is for the department to run it without outside help.
Claire Potter-Schneider's program at Motorola shows what running it this way can produce. There, 30 marketers saved 12,000 hours and generated $1.5 million in pipeline in the first six months, and adoption was still at 80% two years later. She was clear that those results came from a program with no end date. A law firm marketing department is not Motorola, and I would not promise anyone those numbers. The order carries over, starting with a survey of the team before anyone designs the training.
At one large global law firm where I am running this, the department began with a skills survey before any training was designed. It is early, so I would rather describe the approach than claim results it hasn't had time to produce.
If you lead a marketing and BD department and want one thing to do before the end of the year, survey your team before you design anything. Find out who is already building on their own, often in tools nobody approved, and who hasn't started yet.
The results also give you a starting point for a written roadmap, which only 7% of firms have today.