Most family law firm owners fall into one of two camps on AI. The first is holding off because it feels too technical. The second has dipped a toe in but does not know what to do beyond typing questions into a chat window.
Our team has been mapping out how AI adoption actually plays out inside a firm. It has six floors. The first three are accessible to anyone with no technical skills at all.
Floor One: Prompting
Prompting is the basic act of telling AI what you want. Most people start here nervously and get mediocre responses because they give the AI far less information than it needs.
Our team’s recommendation for new users is to use voice-to-text and just talk for three to five minutes. Explain the situation, what you are trying to accomplish, what constraints matter, and what a good output would look like.
Then end every prompt with two directives. Recap what I just told you, and ask me questions about anything I did not cover before you start working. That last piece matters because AI will otherwise produce a plausible-sounding answer that agrees with you rather than the answer that is actually correct.
A useful mental model our team shared is that every prompt has three parts. A raw data payload, a classification instruction, and an output example. Give it all three and the quality of what comes back changes dramatically.
Floor Two: Skills
Once you find yourself doing the same prompt over and over, you have reached floor two. A skill is a written procedure the AI follows every time you invoke it, so you do not have to re-explain the task.
Our team walked through a concrete example. At Sterling, our Consult Legal Assistants used to write manual notes after every consultation for the attorney. That work is now handled by a skill.
The team member drops the transcript into Claude, tells it to run the consult notes skill, and gets a properly formatted dossier back every time in the same structure. Building a skill is easier than most people expect. In Claude, you ask the AI to create a skill for the task, walk it through the process, and it generates the skill file for you.
Floor Three: Skills Plus Tools
Floor three is where AI starts reaching into the other systems your firm already runs on. Your CRM, your document storage, your billing platform. This is where the real efficiency gains start compounding.
The consult notes workflow originally required a virtual assistant to manually pull the transcript from Google Meet, paste it into Claude, run the skill, and copy the output into Salesforce. Faster than doing it by hand, but still a lot of copying and pasting.
The floor three version eliminates that middle work. The AI has connections to Google Drive and Salesforce. You tell it to run consult notes for a specific client, and it pulls the transcript, runs the skill and writes the finished notes directly into the correct Salesforce field.
The intake quality assurance system we built at Sterling runs on this same principle. Every call gets transcribed the moment it ends, scored against a rubric, and the results feed a daily report that lands with the intake trainer before the morning huddle. Coaching happens in near real time with data from every call rather than a small weekly sample.
Why These Three Floors Matter
Everything above floor three gets more technical and involves loops, orchestration, and full factory-level automation. But the first three floors do not.
Any attorney or staff member can learn to prompt well within a week. Skills are a matter of writing down a process you already do. Tool connections are configuration, rather than code.
Most firms that say they cannot adopt AI have not actually tried floors one through three. The productivity gains from just those three floors are meaningful enough on their own to justify the effort.
