Marketing Engineering for Mass Torts: What Changes When Agents Do the Grind
Marketing engineering means building systems where agents handle research and production while marketers own the judgment, the compliance, and the final call. Mass tort advertising may be the best proving ground for it anywhere.
By Jacob Malherbe · September 10, 2026
Marketing engineering is the practice of building systems that do marketing work: agents research and produce, connected data tells them what happened last time, and the marketer decides what they need to know, where their output gets checked, and what stays human. In mass tort advertising, that means faster reads on new torts, more creative tested per dollar, and a compliance gate that checks every asset the same way, with a person approving everything that runs under a firm's name.
I have been running Meta campaigns for mass tort firms since 2015. I started in legal marketing during the 2010 BP oil spill, building landing pages for lawyers who needed claimants. I have watched a lot of "this changes everything" moments come and go.
This one is real.
Our team is becoming ten times the marketers we used to be. Not because we hired ten times the people, and not because a chatbot writes our ads. It is because we started building systems that do marketing work, and we moved ourselves into the role of shaping, reviewing, and steering what those systems produce.
There is a name for this now. Shann Holmberg (@shannholmberg) published a strong walkthrough of the discipline this week, How to Become a Marketing Engineer, and it is worth your time if you work in any marketing role. His piece covers the discipline in general: workspaces, knowledge and data layers, loops, scheduled jobs. This post is about what marketing engineering looks like when the product is a signed mass tort retainer, the rules come from state bars, and a new tort can go from docket filing to crowded ad auction in a matter of weeks.
What a marketing engineer actually is
A marketing engineer builds the machinery behind the marketing. Agents do research and production. Connected data tells them what happened last time. Business context tells them who you are targeting, what you can claim, and what you have already tried. The marketer decides what the agents need to know, where their output gets checked, and which decisions stay human.
The core idea is borrowed from software teams: write the process down, make it repeatable, test it, version it, and improve it every time it runs. Marketers have always had processes. We just kept most of them in our heads and in someone's inbox.
When you move that knowledge into a system, two things happen. First, the grind stops eating your week. Second, your judgment starts compounding, because every correction you make gets saved and applied to the next run.
Why mass torts are the ideal proving ground
Plenty of industries can use agent workflows. Mass torts might be the best fit of all, for four reasons.
The qualification criteria are dense and specific. A tort campaign lives or dies on the details: the product, the exposure window, the injury list, the diagnosis timing, the statute of limitations by state. That is exactly the kind of structured context agents handle well when you give it to them, and exactly what they get wrong when you do not.
Compliance is not optional. Attorney advertising is governed by rules like ABA Model Rule 7.1 and each state's version of it, plus platform policies on top. Every claim in every ad has to be supportable. That sounds like a reason to keep agents away. In practice it is a reason to build them properly, because a well-defined compliance gate checks every single asset the same way, every time, at three in the morning.
The litigation moves constantly. New filings, MDL orders, bellwether dates, settlement news. A human team can only track so much. A research workflow that reads the dockets and the news every morning and flags what changed gives your strategists a head start on every tort.
The outcome is measurable. We do not optimize for clicks. We optimize for signed retainers, tracked per tort, and we publish the live cost per signed retainer by tort. When the data layer connects ad IDs to leads to signed cases, an agent can compare creative angles against the only number that matters to a firm, and flag where tracking is incomplete instead of guessing.
What we actually run at MTAA
This is not theory for us. Here is some of what our agent fleet, which we call The Stable, handles today.
Creative production with a compliance gate. Our ad pipeline pulls the approved tort brief, extracts the visual direction from creative that has already performed, and builds variations. Before anything reaches a human, every render goes through automated checks: does the text fit, is anything clipped, does the claim match what the tort brief says we can support, is the injury language inside the approved list. Assets that fail go back with specific feedback and get revised. Assets that pass land in front of our team for the decision that matters: is this good, and would we put a firm's name behind it.
Creative forensics. A separate agent studies what is running in the market for a given tort and reports back on angles, hooks, and formats, clearly separating what it observed from what it is assuming. That research feeds the next brief instead of living in a screenshot folder.
Outreach drafting that never posts on its own. One of our agents runs on a weekly schedule, finds forum threads and discussions where our expertise is relevant, and drafts a reply. It never posts anything. I read the draft, edit it, and decide whether it goes up. That is the model for anything public-facing: the agent does the finding and the drafting, the human does the publishing.
Litigation monitoring. Scheduled workflows read litigation news daily and turn it into briefs our team and our clients can use. When something moves on a tort we are running, we know before the morning call.
A registry for the whole system. Once you have more than a handful of agents and workflows, you need to know what exists, what it touches, and what depends on it. We keep a control plane that catalogs every piece of our infrastructure, so any agent (or any person) can ask what a workflow does before changing it. Software teams would call this basic hygiene. Most marketing departments have never needed it before. They will.
None of this replaced our people. It changed what our people spend their day doing.
Where the humans stay in the loop
The fastest way to get burned with agents in legal marketing is to let them publish. The second fastest is to let them decide what you are allowed to say.
Here is where we keep human hands on the wheel:
- What goes live. Every ad, landing page, and public reply gets approved by a person. Agents can prepare and schedule. They do not launch.
- What we can claim. The approved claims for each tort are written by people, saved as a file the agents read, and updated when the litigation changes. An agent that invents a claim fails the gate.
- Budget changes. Reporting on spend and recommending shifts are agent jobs. Moving money is a human job.
- Creative direction. Passing automated checks does not mean the creative is good. Taste, empathy for claimants, and knowing what a grieving family will respond to are still ours.
Every time we reject something, we save the reason. "This implies a diagnosis the criteria do not require" is a correction the next run can check for. That is how the system gets sharper instead of just faster.
What we learned the hard way
Generic output means missing context. When an agent hands back bland, interchangeable ad concepts, the problem is almost never the model. It is the brief. Was the injury criteria file attached? Did it have last round's results? Did it know which angles already fatigued? Fixing the inputs fixes the output.
Define "done" before you automate. "Make better creative" is not a task. "Produce six square and six vertical variations from the approved brief, all passing the compliance checklist, with the offer matching the landing page" is a task. The more precise the finish line, the more you can let a loop run toward it.
Schedules do not fix unfinished workflows. Before a job runs every morning, it needs to run correctly by hand. A reporting job that quietly uses stale numbers is worse than no report at all. Ours are built to flag a failed data pull instead of writing a confident summary on top of it.
Start with the work you already know. You do not need a bot for every channel on day one. We started with the task we understood best, creative production for Meta, and built outward from there.
What this means for law firms
If you are a firm buying mass tort advertising, marketing engineering changes what you should expect from an agency.
You should expect faster reads on emerging torts, because research no longer waits for someone to have a free afternoon. You should expect more creative tested per dollar, because production is no longer the bottleneck. You should expect cleaner reporting tied to signed retainers, per tort, with gaps in tracking called out honestly. And you should expect a clear answer to a simple question: where does a human review the work before it runs under your firm's name?
If your agency cannot answer that last question, that tells you something. If you want to see how we answer it, here is how our tort campaigns work.
We are still early
I know I live in a bubble. My feed is full of people building agent workflows, and it is easy to assume everyone is already doing this.
They are not. Especially not in legal marketing.
Most of the plaintiff bar is still running campaigns the way they ran in 2019. That is not a criticism. It is an opportunity. Over the next few years I expect marketing engineering to become one of the most valuable roles inside any firm or agency that runs its own tort campaigns. As building things gets easier, the scarce skill becomes knowing how to put them in front of the right claimant at the right moment, in a way the bar and the platforms will stand behind.
It sounds a little silly until you do it yourself and see how much gets done.
See it live at MTMP
I will be at Mass Torts Made Perfect at the Bellagio in Las Vegas, October 13 to 15. If you want to see what marketing engineering looks like on a live mass tort campaign, from the research feed to the compliance gate to the signed-retainer reporting, I am taking meetings on Tuesday, October 13.
Frequently asked questions
What is marketing engineering?
Marketing engineering is the practice of building systems that do marketing work. Agents handle research and production, connected data supplies performance history, and business context guides decisions. The marketer designs the system, sets the checks, and keeps final judgment.
Is AI-generated mass tort advertising compliant with bar rules?
It can be, but only if a person approves what runs. The tool does not change the rules. Every claim still has to be truthful and supportable under the applicable state rules, and the responsible attorney still owns the communication. Automated compliance checks help catch problems consistently, but they supplement attorney and marketer review rather than replace it.
Does marketing engineering replace a mass tort marketing team?
No. It changes what the team does. Research, drafting, variation building, and routine checks move to agents. Strategy, creative judgment, claim approval, and budget decisions stay with people.
How should a law firm evaluate an agency that uses AI agents?
Ask where human review happens before anything goes live, how approved claims are maintained per tort, whether reporting is tied to signed retainers rather than clicks or leads, and how the agency flags incomplete tracking.
Where can I learn more about marketing engineering in general?
Shann Holmberg's article How to Become a Marketing Engineer is a solid starting point for any marketer who wants to build a first agent workflow. Follow him at @shannholmberg.
