The Questions We Didn't Get To

At a recent American Bar Association CLE, Jennifer Ellis and I spent an hour walking solo and small firm attorneys through practical AI use — legal research, client management, the tools that actually hold up in daily practice. We received far more questions than we had time to answer live.

That was encouraging. It tells me the profession has moved past the question of whether AI belongs in legal practice. Attorneys are now asking the harder, more useful question: how do we use it responsibly?

Below are answers to the questions we didn't get to.

The Question That Should Come Before Any Other

Is it safe to paste confidential client information into ChatGPT?

This is probably the single most important question an attorney can ask before adopting any AI tool — and the honest answer is that it depends entirely on which version of the product you're using and what your firm's policies say.

Most major providers offer several tiers: free consumer versions, paid individual plans, and business or enterprise plans. The business and enterprise tiers generally come with stronger administrative controls, better security features, and clearer contractual commitments around data handling. For OpenAI, Anthropic, and Google specifically, modern paid consumer and business offerings generally do not use API traffic or most paid conversations to train foundation models by default — though attorneys should verify each provider's current policy directly rather than take that as settled fact, since policies change.

None of this shifts the underlying obligation. Regardless of the product, the attorney remains responsible for protecting client confidentiality. My recommendation is not complicated: understand exactly how your provider handles your data, review your own confidentiality obligations, follow your jurisdiction's guidance on AI competence, and when in doubt, don't share personally identifiable information you don't need to share.

This is also where a dedicated Client Intelligence Platform earns its place over a general-purpose chat tool. Deliberately.ai is secure and SOC 2 compliant, never trains on client data, and is built specifically to protect attorney-client privilege rather than treating it as a policy layered on top afterward. The confidentiality question we just walked through — which tier, which settings, which provider policy applies today — is precisely the question a purpose-built legal platform is designed to remove from an attorney's plate.

Build the Prompt Once

Should I use GPTs or Claude Projects?

Yes — and these are some of the most underused productivity tools available to solo and small firm attorneys right now.

Think of them as persistent workspaces rather than one-off conversations. Instead of rewriting your instructions every time, you build them once and improve them over time. A Family Law Research Assistant. A Personal Injury Demand Letter Assistant. A Discovery Objection Assistant. A Deposition Preparation Assistant. Claude Projects work the same way — you attach reference documents, templates, and style guides that stay available throughout the project instead of getting re-explained in every session.

The consistency gain is real. What doesn't change is the obligation underneath it: these are still general-purpose AI assistants, and the attorney remains responsible for reviewing the output against current law, the actual facts of the matter, and professional judgment.

There's a tradeoff worth naming here too. Building and maintaining a good Family Law Research Assistant or Deposition Preparation Assistant takes real time — time most solo and small firm attorneys don't have to spare. A dedicated Client Intelligence Platform like Deliberately.ai comes with these skills already built in, preconfigured for the practice area, and continuously maintained on the vendor's side rather than the attorney's. A purpose-built platform starts working on day one, letting attorneys get back to practicing law instead of maintaining a prompt library.

Is the Subscription Worth It

Free AI vs. paid AI — is it worth paying?

In our view, yes. Paid tiers typically bring better reasoning, more reliable output, larger context windows, faster responses, better document handling, and access to newer models before they roll out broadly. For an attorney using AI regularly, the monthly cost is often recovered in a single hour of saved work. Enterprise plans become worth the jump as firms grow, mainly because of what they add on top — centralized administration, security controls, and team management — not because the underlying model is dramatically smarter.

That said, "paid general assistant" and "purpose-built legal platform" are two different purchases solving two different problems. A paid ChatGPT or Claude subscription makes a smart, general tool smarter. A platform like Deliberately.ai is built around the legal workflow itself — the case file, the document, the deadline — from the start. For the sensitive, repetitive, practice-specific work that makes up most of a solo or small firm's day, that difference in starting point tends to matter more than the difference between model tiers.

What We Owe the Next Generation of Attorneys

How should legal educators incorporate AI into the classroom?

This was my favorite question of the session, and it deserves more than a quick answer.

Students entering the profession will practice in a world where AI is not a novelty but infrastructure. The instinct to treat it as something to be avoided in legal education is understandable, but it's the wrong instinct. The better path is teaching students how to use it well.

That means teaching prompt engineering directly — having students compare prompts and evaluate how better instructions produce better legal analysis. It means requiring verification as a habit, not an afterthought: students should check AI-generated work against primary authority every time, which reinforces the underlying research skill while showing them exactly where AI is strong and where it isn't. It means teaching AI critique as its own skill — spotting hallucinations, missing authority, logical gaps, unsupported assumptions, omitted issues. Learning to critique AI output is becoming just as valuable as learning to produce it. And it means being explicit about the ethics underneath all of it: confidentiality, privilege, supervision, unauthorized practice, citation requirements, professional responsibility. AI literacy isn't a separate elective anymore. It's becoming part of what legal competence means.

There's a practical piece to this too. Training students on general-purpose tools alone teaches them how to use AI; training them on purpose-built systems teaches them how the profession is actually going to practice with it. That's part of why we offer universities and students access to Deliberately.ai at no cost for training purposes — a student who has worked inside a platform built specifically around legal confidentiality and case workflow is better prepared for practice than one who has only ever prompted a general chatbot.

What Confidentiality Actually Requires of a Platform

How does Deliberately.ai protect confidential client information?

This question goes to the center of why we built Deliberately.ai differently than a general-purpose AI tool.

My cofounder and I spent years earlier in our careers building similarly sensitive AI systems — in healthcare, financial services, and defense — before starting Deliberately.ai. Confidentiality at that level isn't a feature you bolt on later. It has to be there from the first line of the system design, and that's the lens we brought with us.

The platform reflects that: each matter is logically isolated from every other matter. AI reasoning happens only within the context of that specific matter. Client information is never mixed across cases. Attorneys control who can access each matter, and every AI response is generated only from the information authorized within that case.

That's a different design premise than a consumer chat tool, where confidentiality, access control, and auditability are add-ons rather than the starting point. I'd still say the same thing to a firm evaluating us that I'd say about any vendor: do the due diligence. Review the security documentation, the contractual commitments, and the privacy terms yourself before adopting anything — ours included.

Your Clients Are Already Using It

Should attorneys give clients guidance about using ChatGPT?

Yes. Clients are increasingly using AI on their own before they ever speak with their attorney. That isn't inherently a problem — but it becomes one if expectations aren't set early.

A few lines of guidance handle most of it: don't upload privileged documents into public AI tools without talking to your attorney first; don't ask AI for legal advice specific to your case; don't edit evidence or communications with AI; don't rely on AI to decide what's important enough to tell your attorney; when in doubt, ask your legal team before you ask a chatbot. None of this is about discouraging clients from using AI. It's about making sure they use it in a way that protects their own interests rather than complicating their case.

What AI Actually Replaces — and What It Doesn't

Can Deliberately.ai track service of process, jury selection, or litigation workflow?

Yes, with an important distinction: Deliberately.ai isn't built to replace a practice management system. It's a Client Intelligence layer that continuously understands what's happening across a matter — which means it can track and reason about service of process, upcoming deadlines, filings, discovery progress, missing documents, outstanding client tasks, expert reports, deposition status, witness information, settlement milestones, and trial preparation.

For firms already running a system like Clio, the intent is to complement it, not replace it — adding continuous reasoning over the full case record on top of the system you already trust for calendaring and billing. Capabilities around jury selection support and litigation strategy assistance are areas we expect to keep building into as the platform matures.

The Future of Legal AI

One theme ran through the entire CLE, question after question: attorneys are not looking for AI to replace legal judgment. They're looking for it to eliminate administrative work, surface what matters faster, improve how clients are communicated with, and hand back the time that judgment actually requires.

That's the whole premise behind Deliberately.ai. Not a system that thinks for the attorney, but one that behaves like an exceptionally organized, tireless assistant who understands every detail of every matter — so the attorney's time goes toward the work only a person can do.