Why your firm's AI mandate is about to flood the market with slop. And the four-part fix every law firm CMO should implement.
A managing partner I know recently rolled out a firm-wide AI mandate. Every attorney has to log AI usage. Every marketing person has to "show AI productivity." There's a leaderboard.
I told him what I'm telling you.
He's about to mass-produce the worst work of his firm's last decade.
Forbes ran an article this month by Genpact's Sanjay Srivastava that put a name to what's happening across enterprise. He calls it "token-maxing." Companies measuring AI adoption by how much compute their employees burn through.
Meta is tying AI usage to performance reviews. Nvidia is talking about token budgets worth half an engineer's base salary. The message is everywhere from boardrooms to all-hands decks. Use AI. Use it now. Use a lot of it.
And here's what's showing up in inboxes as a result: slop.
Stanford and others surveyed more than 1,000 US workers, and 40% had received what researchers called "work-slop" in the past month: AI-generated content that looks polished but says nothing. Emails that sound authoritative and contain no information. Decks that look organized on a glance and fall apart on a read.
Now imagine an alert like that landing in the inbox of a corporate general counsel during her morning triage.
Why this hits law firms harder than tech
In legal, the cost of bad work is brutal, and it's about to get worse.
Tech firms can absorb a flood of mediocre internal memos. The cost is wasted hours and one annoyed colleague.
Law firms can't. Reputation is the product.
The GC reading your client alert is the most cynical content reader in any industry. She has six AmLaw 100 firms sending her near-identical alerts on the same case before lunch. If yours reads like a Wikipedia summary with a partner photo on top, she deleted it.
Pitch decks are worse. RFPs are worse.
Bylined articles are the worst, because the GC really knows the partner. She knows how he thinks. When the byline doesn't sound like the man who tried her last matter, the disconnect registers as carelessness and trains her to discount the next thing she sees from your firm.
Every artifact you put in front of a buyer that doesn't pass the smell test costs you future credibility, even when the work itself was sound.
So why are law firm leaders racing toward the same trap as Meta? Volume is easy to measure. Voice is much harder to put on a dashboard.
The four-part adaptation
Srivastava's piece offers a four-part framework for getting AI adoption right. It maps onto law firm marketing and business development with a few law firm-specific upgrades.
1. Define acceptable use. Permission alone is too low a bar.
Most law firm AI policies right now are permission slips drafted by the GC. They say things like "don't paste client data into the public model" and "use only the firm-approved tool."
Those rules describe what's legal. Compliance is a different question from quality.
The next move for law firm CMOs and BD heads is to map the marketing function by judgment intensity:
- AI augments judgment. This bucket holds first drafts of client alerts, RFP scaffolding, and brainstorming sessions for the next quarter's content calendar. It also covers competitive intelligence summaries, first cuts at awards submissions, and transcript-to-byline conversions in cases where the partner did the real thinking out loud during the original recording.
- AI replaces judgment. This bucket is the danger zone. It includes voice-of-the-partner bylines published without rigorous partner review, pitch narratives generated from a template the partner never read, and anything else where the partner's deep thinking is the actual product the client thinks she is buying.
- AI shouldn't touch. This bucket covers crisis communications, lateral departure announcements, farewell letters from a partner to her client, and notices on sensitive matters. The artifacts where trust is the entire payload should never be machine-generated, even in a first draft.
A real AI strategy starts the moment you replace "is this allowed?" with two specific diagnostic questions for every artifact: where does AI sharpen the partner's brain, and where does it bypass the partner's brain?
2. Assign a named human to every AI-assisted artifact.
This is the area where law firms should be best in class. Most are getting it backwards.
Lawyers already live with named accountability across every part of their day. They sign briefs and opinions with their own name. Their name is on the door of the firm itself.
Marketing and BD have gotten loose. A client alert drafted in Harvey gets reviewed by the practice group's marketing manager and goes out under a partner's byline. The partner often hasn't read it.
That's the wrong human in the accountability seat. The accountable name has to be the person whose reputation rides on the artifact.
The test: if a sophisticated client called the partner tomorrow and asked "did you really write this?", could she answer yes without flinching?
If the truthful answer is no, you have a slop pipeline. Fix it before a GC asks the question first.
3. Measure the outcome. Activity metrics will mislead you.
Law firms run on activity metrics across every function. They count hours billed, pages drafted, and CLE hours completed.
The instinct when AI shows up is to ask: how many partners are using the tool weekly? What's our seat utilization rate, and how does it compare to peer firms? How many prompts went into Harvey last quarter? Those questions will steer you wrong because they say nothing about whether the work got better.
Outcome metrics worth tracking inside legal marketing and BD:
- Pitch cycle time. Measure RFP receipt to send.
- Win rate on AI-assisted versus non-AI-assisted pitches. A lower rate on the AI side means the AI is hurting you.
- Open and reply rates on client alerts. If AI freed up partner time for sharper takes, the open rate should climb. A flat rate means your alerts read like everyone else's.
- Time-to-publication on bylines. Partners are the rate-limiting step in this pipeline. AI should compress publication from months to weeks.
- Voice Fidelity Score. This is the percentage of bylined partner content that a sophisticated reader could correctly attribute to that specific partner with the byline removed.
That last one is my anti-slop metric. Adopt it if it's useful.
4. Investing in the 100 trees pays the highest return.
Srivastava coined this term, which is the most important one to act on this quarter.
Every firm has 5 to 10 people getting more out of AI than the rest of the firm combined.
The rainmaker who ghostwrites her LinkedIn posts in 20 minutes with a custom project trained on her past articles. The BD director who built a private model on three years of the firm's pitches and uses it to scaffold every RFP. The associate who turned six-hour competitive intel into a 45-minute exercise.
Find them.
Then redirect your training budget away from vendor-led "intro to Copilot" sessions for the whole firm. Those sessions deliver exposure to the tools, which is a different thing from skill transfer.
Effective training documents what your power users do across an entire week of normal work. Their prompts get captured by someone watching them work. Their workflows get written down step by step. Their review rituals get codified into checklists that another partner can follow on a Monday morning. Then you package the whole thing as the firm's playbook.
Pay the rainmakers for their workflow IP. Their time costs the firm money and their methods are worth more than the time.
A quick disclosure
I am the Cofounder of InnovAItion Partners. We license workflows and run AI training for marketing, BD, and communications teams at law firms. So yes, I make money when firms get serious about this.
I want to be upfront about that, because the article above tells you to fire your generic AI training vendor. If I sound like one of those vendors, ignore me.
Here's the difference. We don't run "intro to Copilot" sessions. The first thing we deliver isn't a tool demo. It's a judgment-intensity map of your marketing function, Section 1 of this piece, applied to your specific firm. The training that follows is built around the use cases your power users are already running, plus a library of marketing, communications and business development workflows created specifically for professional services firms we've refined across the firms we work with.
Some firms hire us to do the mapping. Some read articles like this and do it themselves. Both work. The point isn't who runs the exercise. The point is that somebody runs it before your firm-wide AI mandate produces six months of slop.
The bottom line
Your firm's AI strategy this year will be judged by your clients. Your IT department doesn't get a vote.
If a GC can still tell your partner apart from the algorithm twelve months from now and prefers reading your partner, you won the AI shift.
If she can't, no amount of token consumption will save the relationship.
Stop counting tokens. Start counting decisions.