The case you lose is the call that went to voicemail.
AI that takes the repetitive work around a case off your team, in the systems you already use. Every output is a draft. Someone reviews it. Nothing is sent to a client on your behalf.
The legal work is bespoke. The work around it is not.
A liability argument is judgement. Requesting the same records from the same providers, and telling forty clients that nothing has changed this week, is not. On contingency, that difference is capacity, and capacity is fee.
-
Intake and lead response
Speed decides who signs. An enquiry arrives, someone qualifies it, runs conflicts, books the consult, and sends the retainer. The judgement is yours. The turnaround is the problem.
-
Medical records and bills
The same requests to the same providers, then the follow-up when nothing arrives. Predictable, endless, and nobody senior should be tracking it.
-
Case status updates
"Any news on my case?" Most weeks the honest answer is no, and it still takes someone twenty minutes to say it forty times.
-
Demand package preparation
Assembling records, bills, and a chronology into a first draft that an attorney then makes an argument out of.
-
Settlement and closing admin
Disbursement letters, lien follow-up, closing communications, and the review request that gets skipped when the next case is already in.
Medical records and client health information carry their own sensitivity. That is a reason to be deliberate about which tool touches what, not a reason to avoid the question. Configuration and vendor terms are exactly the work. These are patterns the practice generally runs on, not findings about your firm; the assessment establishes which actually cost you.
You almost certainly do not need a server in the closet
The most common reason firms stall is a belief that using AI at all means putting client confidences somewhere they should not go. The published guidance says something narrower.
"ensure that the [GAI tool] is configured to preserve the confidentiality and security of information, that the obligation is enforceable, and that the lawyer will be notified in the event of a breach... determine whether the [GAI tool] retains information submitted by the lawyer"
ABA Formal Opinion 512
That is a configuration and contract question, not a hardware question. The words "on-premises" do not appear in the opinion. North Carolina is more direct still, answering whether a lawyer may provide client information to a third-party AI vendor:
"Yes, provided the lawyer has satisfied herself that the third-party company's AI program is sufficiently secure and complies with the lawyer's obligations to ensure any client information will not be inadvertently disclosed or accessed by unauthorized individuals pursuant to Rule 1.6(c)."
NC 2024 Formal Ethics Opinion 1
The work is satisfying yourself. Opinion 512 contemplates that a lawyer may do that by consulting someone who has read and analysed the terms, and notes firms may need to consult IT or security professionals. That is the job: reading what a vendor actually retains, configuring tools so they behave the way your firm has decided, and training your people on it.
Stephen is not a lawyer and does not advise on your professional obligations. He is not your ethics counsel, and engaging him does not make your firm compliant or satisfy any duty you owe. Those decisions stay with your firm. Private on-premises deployment is available if you would rather not rely on a vendor at all.
Before any of that: the questions to ask any AI vendor before client information touches it. Ten of them, each traceable to Opinion 512 or NC 2024 FEO 1. Free to read, and I will send a one-pager for the partners' meeting if you want one.
Start with the assessment
Two weeks, $995.
What two weeks buys you
- A discovery call and workflow interview
- Two weeks mapping how the business actually runs
- A written roadmap: highest-ROI opportunities, named tools, dollar estimates
- A walkthrough call to put it into action
- The roadmap is yours to keep, whoever implements it
A scoped engagement of this size starts at $3,495. The assessment that tells you what to build is $995.
If it does not identify at least 10 hours a month of time savings, you get a full refund. Every dollar back. The hours are Stephen's professional estimate of what is recoverable; actually banking them depends on implementing the roadmap.
Built in 30 days, or the meter does not start. That is the second guarantee, and it applies only if you go on to have Stephen implement the roadmap: Web form capture and alerts are live within 7 days of kickoff. Missed-call text-back switches on once your carrier registration clears, usually two to three weeks. If nothing is live within 30 days, you do not start paying the monthly fee until it is.
What comes after is scoped from what the assessment finds. Some firms want the workflows built and run for them. Some want their team trained to run it, which is a workshop from $3,495. Some read the roadmap and do it themselves, which is a fine outcome and the roadmap is yours either way.
The arithmetic is worth doing with your own numbers rather than mine. If faster intake signs you even one additional case this year, put your average fee against $995. That is the bet the guarantee makes free to take.
Contingency work is the cleanest case for this. Time saved is not lost billing, it is capacity for another matter at the same fee. If your bottleneck is case supply rather than throughput, say so on the call and Stephen will tell you whether this pays for itself.
Do we have to run AI on our own hardware to stay confidential?
No. ABA Formal Opinion 512 never mentions on-premises deployment. It treats generative AI the way earlier opinions treat cloud computing and outsourcing, and asks whether the tool is configured to preserve confidentiality, whether that obligation is enforceable, whether you would be told about a breach, and what the vendor retains. North Carolina 2024 Formal Ethics Opinion 1 permits providing client information to a third-party AI vendor where the lawyer has satisfied herself the tool is sufficiently secure under Rule 1.6(c). Private on-premises deployment is available if your firm wants it, but it is a choice, not a requirement.
Are you giving us an ethics opinion?
No, and you should not accept one from a non-lawyer. Stephen is not your lawyer and does not advise on your professional obligations. What he does is the technical work underneath them: reading vendor terms and telling you plainly what a tool retains and who can see it, configuring tools so client information is handled the way you have decided it should be, and training your team on the rules you set. Your obligations stay yours, and the decisions stay with your firm.
Will this replace lawyer judgement?
No. Everything is set up to produce a draft that a person reviews. Nothing is sent, filed, or promised to a client without someone at the firm reading it first. Stephen does not build systems that contact clients on your behalf.
Does saving time actually help our firm?
Contingency work is the cleanest case for this. Time saved is not lost billing, it is capacity for another matter at the same fee. If your bottleneck is case supply rather than throughput, say so on the call and Stephen will tell you whether this pays for itself.
Do we have to change practice management systems?
No. The work happens in the tools you already use. Replacing a practice management system is a large, disruptive project with its own risks, and it is not what this is.
What does it cost?
The AI Workflow Assessment is $995 and takes two weeks. If it does not identify at least 10 hours a month of time savings, you get a full refund, every dollar back. What follows is scoped from what the assessment actually finds, so nothing is quoted before the work is understood.
Worth twenty minutes?
A short call to work out whether there is enough repetitive work in your firm to be worth changing. If there is not, you will hear that.