A managing partner told me last month that the firm had bought AI seats for everyone, run two training sessions, and eighteen months later the honest answer to “what are people using it for” was better emails. He was ready to call the whole thing a failed experiment.
The tool was not the failure. The firm had skipped the step that makes any tool worth having, and nobody had told them it existed.
That step has a name now. Jessica Aries, who trains legal marketing teams on AI and runs her own agency on a dozen AI agents, calls it a context layer. On Counsel Cast this week she described it as the foundation most firms never built before they bought the tools. This post is about what it is, what goes in it, and how a law firm builds one without a technology budget or an engineer.
What is a context layer?
A context layer is the written record of how your firm works, kept somewhere the AI reads before it does anything. Who you serve. How you sound. What a great piece of client communication looks like and what an acceptable one looks like. What never goes in a bio. Which words the firm does not use.
Some people call it a second brain. Some call it an operating system. The name matters less than the fact that it exists, because without it, every draft the AI produces is the same generic draft every other firm in your practice area is getting. The model does not know your firm. It knows the internet.
Here is the part worth sitting with. None of this is new. Ten years ago we called it a marketing strategy, and it lived in a Google Doc that went from one person to the next. If your firm never wrote that document, AI did not create the gap. It exposed it, and faster, because the generic output now arrives in seconds and someone decides marketing does not work.
Why time saving was the wrong promise
The industry sold AI to law firms as a way to save time. Jessica’s view, and I share it, is that this was the wrong promise, and it is why so many firms cannot find the return.
Measuring time saved assumes the work was already good and the only problem was speed. At most firms the work was inconsistent, the standards lived in one partner’s head, and the AI simply reproduced that inconsistency at scale. What AI does well is take good work and make it better. An A team gets A-plus capabilities. A firm with no written standards gets faster mediocrity.
That reframe changes what you build first. Not a prompt library. Not an AI employee. A clear statement of what good looks like at your firm.
The five things a law firm’s context layer needs
You can build this in a week with documents you mostly already have. The order matters less than finishing all five.
- One current description of the firm. Not the About page from 2019 and the proposal boilerplate from 2023 and the bio the managing partner rewrote last spring. One version, dated, with an owner. Jessica’s warning on the show was that connecting AI to five conflicting versions of your firm gives it five opportunities to choose the wrong one.
- A one-page ideal client. Who the firm is for, in plain language. The practice area, the client size, the problem they bring, and the two or three things they care about when choosing counsel. If three partners would write this page differently, that is the first thing to fix, and it has nothing to do with AI.
- A voice guide with examples. Three to five pieces the firm is proud of, with a short note on why. Three to five pieces that went out and should not have. The AI learns more from the contrast than from any list of adjectives.
- A list of what never goes out. The claims you cannot make. The phrases that sound like every other firm. The clichés your marketing director has spent years removing. Write them down once and the AI stops reintroducing them.
- A definition of done for one recurring task. Pick the task you run most often, a client update, a practice group newsletter, a bio refresh, and write the standard a draft has to meet before it leaves the building. This is the piece most firms skip and the one that pays back fastest, because it is what lets a person check the output instead of guessing.
Start with something low risk
The trap is connecting everything at once. Jessica’s example on the show was a trademark practice that built a database of its own past applications and the feedback each one received from the USPTO, then codified the patterns into a standard every lawyer runs before filing. Low risk, because it used information the firm already owned, not live client matters. High value, because it pulled judgment out of one partner’s head and made it something the whole practice works to.
The higher-risk version is the one that makes people nervous for good reason: a lawyer connecting the document management system to an AI tool without checking whether the account is enterprise grade, or experimenting on the same machine that holds client files. Play in a separate folder. Use a different machine if you have one. Treat the first month as practice, not production.
The error lawyers miss
Once the context layer exists, one habit protects it. Jessica calls it the soft hallucination, and she says it is the question legal marketers miss most on her AI assessment.
Everyone knows to check for the invented case citation. The soft version is subtler. The AI writes “across the industry, we’re seeing a shift toward X” with no source. It sounds exactly like a partner wrote it, because partners write sentences like that all the time, so nobody flags it. Her rule is that confidence is the signal. When the output sounds sure of itself and cites nothing, go and find the source.
Ten minutes a day, one model
Jessica’s closing advice on the show was the opposite of what most AI content tells you. Stop trying to keep up, because it is impossible and even the people who do this for a living feel behind. Pick one model, not three. Then spend ten minutes a day trying one thing you have not tried, the way you practice law, until you can tell when something broke and why.
She calls that competency rather than fluency, and it is the only AI skill that survives the next update. The tools will change again by spring. The context layer, and the judgment you wrote into it, is the part that stays.
If you would like to understand how a context layer fits with the rest of your firm’s visibility, start with how law firms build authority that AI search can see.
Want help building the strategy before the tools?
That is the work we do with firms at Conroy Creative Counsel. If your firm has been buying AI tools and training and is not sure any of it will matter next year, see how Marketing Co-Counsel works.