During this year's review cycle I've talked with five CMOs at midsize and large law firms. Each told me the same thing: Nearly every self-evaluation that crossed their desk listed AI training as a need.
I read that as the first visible sign that the people in the middle of the marketing and BD department are about to change what they do all day, from producing the tactical work to directing AI that produces it. By the middle, I mean the specialists and coordinators who build the pitches and RFP responses, along with the managers who supervise them. They are asking for help before their firms have planned for the change.
The request is coming from the people doing the work
The people being reviewed write their own self-evaluations, so these requests came from team members setting their own development goals for next year.
They have reason to ask. In the forthcoming 2026 State of the Industry Report: AI in Law Firm Marketing and Business Development, which InnovAItion Partners produced with Gladstone Growth Strategies , 47% of the 87 firms surveyed have provided broad introductory AI training. Fewer, 28%, have offered training designed for marketing and BD roles. Only 18% have built training around specific marketing and BD workflows. A quarter have offered none.
Introductory training explains what AI is. Team members are asking for something harder to deliver: how to do their actual jobs with it.
The middle of the team starts directing the work
Look at what fills a law firm marketing and BD team's week: award submissions, RFP responses, pitches, directory listings and competitive intelligence research. Much of that work is headed to AI agents, meaning software that carries a multistep task from start to finish while a person checks the result.
One CMO put it more bluntly than I would: The middle layer of the department disappears, along with the manager layer. I agree it gets smaller. I expect the team that remains to direct agents through the tactical work and review their output, where today people do every step by hand.
That is a forecast. Few law firms use agents yet, because the risk is still too high while security and confidentiality questions stay open.
Even so, I expect agents to be in regular use at law firms within about a year, once firms settle those concerns. A year is one more review cycle.
Judgment is whats left, and most of it lives in peoples heads
When agents do the drafting, the human work is directing them: telling the agent what the firm wants from this piece of work, then having the taste to decide whether the draft that comes back is good enough to send. Directing an agent is closer to managing a colleague than using a chatbot: You hand over the assignment, then review the work.
Take an RFP response. An experienced person knows where the firm is strongest for this kind of work and which attorneys and matters to put forward as proof. An agent can assemble a clean submission but cannot make that call, because most of what informs it was never written down.
That judgment sits with the people now in the middle layer. I don't know yet how the next generation learns it once agents take over the years of hand-drafting that used to teach it.
Write down the reasons behind the work
Start by writing down what good work looks like. Few of the law firm marketing and BD teams I've worked with keep formal standard operating procedures (SOPs). Today that knowledge sits with the manager. With no written standard to measure against, checking falls to whoever did the work, or to no one.
An SOP records the steps. An agent, or a new hire, also needs the reasoning: why this matter went into the pitch instead of that one, and when the firm made an exception to the rule because someone pushed back.
Zapier offers a model from outside legal. On The Artificial Intelligence Show, Episode 231, Dan Slagen, Zapier's chief marketing and AI transformation officer, described the AI "marketing brain" Zapier built for its marketing team: a shared AI system that pulls in the team's public Slack channels and documents. It can only learn from what it can see, so a one-on-one call or a private message never reaches it. Slagen started tracking how much of his own communication happened in private and, over three months, took it from about half to 4%. As he put it, "most things can and should live out in public."
At many law firms, Microsoft Teams does the job Slack does at Zapier. The conversations about which matter goes in the pitch and why already happen in one-on-one emails and calls. Shift them into Teams channels the marketing and BD department or the whole firm can see. Public here means visible inside the firm only. Those threads become a written record of why the team made each decision, exceptions included, that the next person in the seat can learn from right away. As firms settle the confidentiality questions and connect their AI tools to those channels, I expect AI to draw on the same record.
You can teach judgment on purpose
Then teach the judgment itself. A junior team member puts an AI draft beside the version a senior colleague would have sent, then hears from them why one is stronger. Make that comparison a standing part of the work so the team learns from every draft it reviews.
The training also has to be specific to law firms. Legal is a regulated industry where confidentiality touches nearly every decision a marketing or BD team makes, so the examples have to come from that work, like an RFP response or a directory submission.
Collaborating with AI changes a team's daily routine, so the training has to keep going well past the first course. Zapier treats it that way. Slagen described a training push that grouped Zapier's marketers into three tiers of builders, meaning people who build their own AI tools and workflows. Each tier now gets new problems to solve, matched to its level.
Yes, I train law firm marketing and BD teams, so weigh my enthusiasm for ongoing training with that in mind.
The training has to come before the agents
Firms haven't planned for this yet, which is the problem. Their own people are already asking for help: Nearly every self-evaluation I heard about this cycle requested AI training.
If agents arrive within a year, the training has to come before them. Teaching judgment after agents are already doing the work means teaching it in a hurry.
The people writing those self-evaluations know that working with AI is where their careers are headed. These skills make them more valuable and more marketable. They will go where they can build them, to a firm willing to invest in its professional staff as well as its lawyers.
Their request is sitting in this year's reviews.