Train Junior Lawyers With AI Instead of Losing Them to It
By Caitlin Price, Social Media & Content Marketer•Published
When a younger lawyer on her ten-person team tells her “Well, I don't know, GC AI said to do it,” a general counsel at a private-equity-backed software company and GC AI customer has one reply: “If it told you to jump off a bridge, would you do it?” She said it on a call with our team earlier this year, and it is the whole question of how to train junior lawyers with AI.
She also has an answer, and it is the method behind this playbook. Three other in-house teams raised the same worry on calls with our team over the past year, unprompted.
Cecilia Ziniti, GC AI's CEO, answered that question in Bloomberg Law in August 2026. Within three years, she wrote, “a lot of the routine work will get handled by AI agents,” and lawyers “supervise those agents the way they supervise junior associates now.”
She added a second sentence: “In-house lawyers have been doing exactly that for their entire careers. They're already trained for this.”
The first draft moved to the AI. The supervising stayed with you, and the training came with it.
Training junior lawyers with AI means running the same supervised loop a good senior lawyer always ran, the apprenticeship of first draft, review, and correction, with the AI producing the draft and the junior lawyer producing the questions about it. On a lean in-house team, the senior lawyer supervising the junior is also the senior lawyer supervising the AI, and the same five habits cover both jobs.
On CZ and Friends, GC AI's podcast, Tricia Kinney, Chief Legal Officer at Consilio and then General Counsel at BlueLinx, said the line that stuck with her came from Wharton's Ethan Mollick:
“You've got to think of it as a new hire on your team. If you have a brand new person who starts and you just set them free, they're not gonna be effective. You've got to coach, you've got to give context, you've got to give background, and you've got to give feedback.”
GC AI is the enterprise legal AI platform a three-time general counsel (Anki, Bloomtech, and Replit) built for in-house teams, used by 2,100+ in-house legal teams as of September 2026, including the legal teams at Hitachi, Jasper, and Viant Technology. How it compares with the rest of the field is in the best legal AI tools for in-house counsel guide.
Five GC AI features carry the supervising work in this playbook:
- Easy Prompt turns a junior lawyer's plain-language ask into a full prompt, so the context you taught them to add goes in every time.
- Custom Company Profile holds the team's risk posture and standard positions, and Custom Instructions in each user's settings hold the individual lawyer's, level included.
- Playbooks run a contract against positions you wrote down once and sort each clause into pass, fallback, or flag.
- Team shared chats and a team Skill Library, the saved prompts the whole team can run, both on the Team plan, let a junior share their work with you as it happens and run the prompts you would have run yourself.
- Exact Quote returns a character-level citation for every quote pulled from a document, which turns “verify it” into a click.
Will AI Hurt Junior Lawyer Training? What In-House Lawyers Told Us
AI hurts junior lawyer training when the supervised first draft disappears with nothing in its place. The teams below kept the review and moved the junior lawyer's job from writing the draft to interrogating it.
An associate general counsel at a global apparel company, a GC AI customer on an all-senior team, asked it as a question about her own profession:
“How will our younger attorneys that are still up and coming and need to be trained, how are they going to use these tools?”
She started from the metaphor Ziniti uses, the eager intern, and pushed it a rung higher:
“Cecilia, you talk a lot about how GC AI is like having an eager first or second year intern. I mean, I go a step further and just say it's a first or second year associate, which means, yes, I have to monitor its work. Yes, I have to kind of question it because it doesn't always get it right. But because I'm a senior attorney with years of practice, I'm very comfortable with that.”
That comfort takes years to build, and a second-year lawyer is still building it. The general counsel of a 350-person consumer platform, a GC AI customer running a two-lawyer team, watches for exactly that:
“She's a second-year lawyer. So she does not have a lot of experience. And I've been trying to be a little cautious with her about using this because I don't think she knows enough to push back on when she's not asking the right question or this doesn't sound right.”
His own habit is the one she is learning. He sets up the first question with Easy Prompt, then spends the session on “wait, can you tell me a little bit more about this?”
The worry is older than the platform. On the same podcast, Mary Ambacher, Deputy General Counsel and Corporate Secretary at EverQuote, put the ten-year version of it this way:
“How do we get five to 10 years from now when you still need the [Laurens] and I who've had this business judgment and have gotten it from practicing for so long if you don't have the junior attorneys starting and getting that from the beginning?”
Law-firm leaders see the shift from the other side. In a February 2026 General Assembly survey of 258 director-level and above leaders at large consulting, accounting, and legal firms in the US and UK, 63% said junior roles will evolve with different responsibilities and skill sets. Among the legal-firm respondents, 74% named partners as the group least prepared to use AI effectively.
Every one of them kept using AI on the work. Their question was about the review around it, and the review has an owner.
Supervising Junior Lawyers Using AI: Who Owns the Mistakes?
A supervising lawyer owns the work a junior lawyer produces with AI the same way they own the work a junior produces without it. ABA Formal Opinion 512 reads Model Rules 5.1 and 5.3 to require that lawyers who manage others set clear policies on generative AI and make sure the people using it are trained to work within the rules. Our AI acceptable use policy template covers the written-policy half.
The duty belongs to you, and no platform can discharge it for you. Two courts have now said the same thing to lawyers who assumed the review was someone else's job.
In Wadsworth v. Walmart, a February 2025 decision from the District of Wyoming, the drafting attorney's in-firm AI database produced eight fake cases out of nine, and the court fined the supervising partner $1,000 for e-signing the filing without reading it. Our AI hallucination legal cases tracker carries the full record.
Singapore's High Court went further in March 2026. In Tan Hai Peng Micheal v Tan Cheong Joo, a paralegal drafted closing submissions that cited two cases that did not exist, and the court ordered both the solicitor who reviewed the draft and the counsel on record to pay S$5,000 each in personal costs.
The judgment's language is the standard to hold yourself to. The counsel on record “had a non-delegable duty to ensure that the submissions were complete, accurate, and properly verified,” and the reviewing solicitor “ought to have recognised that a paralegal is not qualified to prepare written submissions independently, and that proper supervision and the verification of all authorities was his ultimate responsibility.”
Both courts put the penalty on the lawyer who skipped the read. The junior lawyer's AI output becomes your work product the moment you sign it, send it, or let the business rely on it, so read it the way you would read a first-year's draft.
How to Train Junior Lawyers With AI: A Supervising Lawyer's Playbook
Habits one, two, and five come from the software GC's ten-lawyer team, habit four from the consumer-platform GC, and habit three from the way customers like KT Farley at Helix write their positions down. Five habits:
- Have the junior tell the AI they are new to the role.
- Make them argue with it, point and counterpoint, on every draft.
- Write your positions down once, so the junior sees your standard on every clause.
- Read the chat before you read the draft.
- Run a weekly what-did-you-learn.
Have the Junior Tell the AI They Are New to the Role
The software GC's one-on-ones now start with a redirect. When a younger team member asked her how to find something, she answered, “I don't know. Ask GC AI. I'm like, seriously, just go talk to it. And I said, give it context.”
The context she means is about the person:
“Where we can supercharge our younger or more inexperienced team members is, frankly, context... I said, tell them that you're new to this type of role.”
Her reason: “I think they get a little more of a catered response to a legal concept that maybe they have not considered before.” In our 101 class we teach that context is king, and the rule holds for a second-year lawyer as much as for the senior lawyer prompting on her behalf.
Custom Instructions, in each user's settings on GC AI, hold that context so it goes in every time. Here is an illustrative version to write with your junior in the same sitting:
I am a second-year in-house lawyer on a three-person legal team at [Company]. For every recommendation, explain the reasoning, name the legal concept it rests on, and tell me what a more senior lawyer would push back on before I send this.
From there, Easy Prompt turns the junior's plain-language ask into the full prompt, with the context already attached:
Make Them Argue With It on Every Draft
The software GC's team runs point and counterpoint with each other, and she wants the same reflex with the platform:
“Being able to defend yourself with the tool and, you know, point counterpoint. We do it all the time with each other. Use this as a way to interact.”
Her measure of progress is whether the junior can say it out loud: “The more they can speak about it or talk about it too, the more they can help articulate their position.”
Lauren Anderson, Senior Counsel at Wayfair and a GC AI customer, explained on the same CZ and Friends episode why the exercise works:
“You don't always think about every terrible outcome that comes out of a commercial contract until something bad happens. So with that experience, it's like I can build that into my review.”
GC AI's counterargument hands a second-year lawyer the bad outcomes before she has lived through one. Here is an illustrative prompt for a limitation of liability clause, with more templates in our AI prompts for lawyers guide:
Here is my position on the limitation of liability clause in the attached MSA: [position]. Argue the counterparty's side as hard as you can. Then tell me which two of my arguments are weakest and why.
One of our own team members put the same idea to the apparel AGC on her call: use it “to create friction, to give you an opposing view,” write your read of what a contract allows, then “take that opinion and kind of rip it to shreds. Like where are my gaps?”
Write Your Positions Down Once
A senior lawyer's judgment sits in her head, and the fastest way to teach it is to write it down where the junior can read it on every clause. Anderson described the compressed version her team uses on renewals: “You can just say, here are these four things that I care about, are they contained in this agreement? If not, just add them in.”
Ambacher grades the work before anyone touches it: “SEC work, you need A plus work. That's attention to detail, that's public filings. Some commercial contracts, you know, I'm like, give me your best C.”
That grade, and the positions behind it, is what a Playbook holds. You write the standard once, per contract type, and the review comes back with each clause sorted into pass, fallback, or flag, next to the standard it was checked against.
A position in the Playbook reads like this:
Limitation of liability: cap at 12 months of fees paid. Fallback: 24 months. Flag: any uncapped indemnity, or a cap below 12 months.
KT Farley, Chief Privacy Officer and Associate General Counsel at Helix, runs the saved-prompt version of the same idea:
“The ability to create and store reusable prompts and share them across the team has completely changed the work required to review standard work. Junior teammates now run the checklist prompt first and bring me the output as the predicate for my review.”
The junior runs your standard first, and your review starts from their output. Our guide on how to build a contract playbook starts from your last five redlines.
The team-level version is the Custom Company Profile, which carries the company's risk posture, jurisdictions, and preferred positions into each chat. Ziniti named the skill it encodes in CCBJ in August 2026: “The calibration of the law to the company, that's the ultimate skill of an in-house counsel.” Written down, it is the part of your judgment a second-year lawyer can borrow on day one.
Read the Chat Before You Read the Draft
The consumer-platform GC supervises by staying close, in his words: “I've tried to be involved with the things that she's using GC AI for a little bit just to be careful.”
Team shared chats, on GC AI's Team plan, let the junior share the chat with you as they go, so you see the prompt, the follow-ups, and the point where they accepted an answer without a second question. That point is the exact moment the software GC's bridge line is for.
Rebecca Fike, a partner in Reed Smith's Global Regulatory Enforcement Group and a former SEC enforcement senior counsel, runs the law-firm version of the same habit, as she said on CZ and Friends:
“I'll have maybe a second year on a phone call with the client because that is more efficient than me after the phone call calling the second year, relaying what was said... They're also learning, they're learning from the client, they're getting these facts directly.”
Reading the chat is the in-house version of putting the second year on the call. The draft that follows gets the same read a first-year's draft always got, and Exact Quote shortens it: each quote the junior pulls from a document comes back with a character-level citation, so the junior checks the source with one click.
Run a Weekly What-Did-You-Learn
The software GC added a standing item to her team meetings:
“We share that information of, hey, what did you learn this week? Is there a concept that now you are much more familiar with that you didn't know about last week?”
Three questions cover the standing item:
- What did you learn this week?
- What did the platform get wrong, and how did you catch it?
- What would you push back on now that you would have accepted last week?
Tina Patel, former VP and Associate General Counsel at Amazon Lab126, ran the same loop at Amazon before any of this existed, as she said on CZ and Friends: “Just training people, really trying to explain why we did things the way we did... It's fine to get the challenge. And sometimes there were better ways.”
Diane Honda, Chief Administrative Officer at Redis, explained on CZ and Friends why the meeting has to exist: “Law school teaches you what to think or how to think a little bit, but it doesn't teach you how to do your job.” The weekly meeting teaches the job, and the AI supplies a fresh concept to teach every week.
Michele Murray, who leads legal at ARKO Corp, offered the one item the platform will never cover, on the same podcast: “AI can't send certified mail... Everyone should learn how to send a certified letter. That's my first advice.”
The software GC's own summary of the five habits: “You're going to use the tool to train you or teach you.”
Start this month with three moves:
- Write the Custom Instructions with your junior in one sitting: their level, their team, and the reasoning they want shown.
- Build one Playbook from the contract type they review most: your last five redlines are the positions.
- Put what-did-you-learn on the standing team agenda: go first.
What the AI Drafts, What the Junior Owns, and What Stays With You
On a lean in-house team the split runs by grade, and the supervising lawyer sets the grade. Here is how the four kinds of work a junior lawyer touches divide up once the AI takes the first draft:
| The work | What the AI hands back | What the junior lawyer does | What you decide |
|---|---|---|---|
| An NDA or vendor form on your paper | Every clause sorted against the Playbook, markup drafted on the flags | Runs the review and writes a two-line note on each flag | Whether the fallback is acceptable this time (B+ work) |
| Third-party paper above the dollar threshold | The clause sort, plus the counterparty's argument on request | Argues each flag with the platform, brings you the three that matter | Which flags are worth spending leverage on |
| A research question outside anyone's practice area | A cited answer from primary law | Opens every citation and reads the full source | What the company does about it, and by when |
| A board memo or anything carrying your signature | The first draft, with Exact Quote citations | Verifies every quote to the character | Every word, before you sign (A+ work) |
Each row keeps the junior lawyer on the part that builds judgment: the question, the counterargument, the source, and the note to you. Anderson's word for how a junior gets there is “reps,” and each row is one. “You stop being the bottleneck and start being the guardrail,” as Ziniti put it in the same Bloomberg Law piece.
Our post on why judgment is the new superpower for in-house lawyers makes the longer case, and the AI skills for lawyers guide covers the four skills the junior builds along the way.
What AI Does for Senior Attorneys on Small In-House Teams
AI for senior attorneys on a small in-house team does the work a first-year to fifth-year associate would have done, and the senior lawyer supervises it the same way. The legal operations leader at a multinational travel company, a GC AI customer, described what the rest of his legal stack left uncovered:
“None of those solves for the presence of a true legal assistant or three, four or five year associate. So people that we might otherwise pay three or $500 an hour to sitting outside of the company, that's where a tool like this can come in and immediately deliver value.”
What goes to outside counsel changes shape too: “What we're giving them is much further along, as if they'd been working on it for a week or two or a month or two with a team of associates or junior folks. And they're now just the final sort of polishing.”
On a senior-heavy in-house team, the survey's least-prepared group and the supervising lawyer are the same person. The fix takes an afternoon: GC AI's legal AI classes, taught by former general counsels, have taught 8,000+ lawyers, and by GC AI's own count as of September 2026, lawyers who take a class adopt AI 3 to 5 times faster than those who skip one. The Legal AI Ethics class walks through Opinion 512's supervision duty, and Legal Ops: Rolling Out Legal AI covers the rollout across a team.
The hiring changes too. Sharon Johnson, SVP and Chief Legal Officer at Mode Global and a GC AI customer, said on CZ and Friends that it has changed who she hires:
“I may hire someone who is less experienced now, maybe direct from law school even, just because I have the ability to teach them better. We've had some interns, and the work product that we were able to get out of them is far above what previous interns have done, just because they have better tools and we like to teach them how to use them.”
Peter Katz, VP and General Counsel at Expel and the first GC at Duo Security, whose former reports have gone on to GC roles, said on the same podcast that he hires for the ceiling:
“I always assume that if the resume gets to me that the person has the ability to do the job. So it's not about how many contracts have you negotiated and all of these kinds of things. It's all of the other stuff that it's important for GCs to be looking for when interviewing.”
Ziniti replied on that episode: “We're circling back to that kind of judgment, human layer, you know, like how do you get there? And that makes me excited about AI.” With the first draft handled, the second lawyer you hire starts on the judgment work. In GC AI's December 2025 ROI study of more than 100 active customers, lawyers reported an average of 14 hours a week back to spend on it.
Where to Go Next
The department-wide view sits in AI for law departments, the rollout mechanics in legal AI change management, and the question of when a lean team adds its second or third lawyer in corporate legal department structure. If the junior on your team is a contract manager or paralegal, legal AI for non-lawyers draws the line on what they can safely run alone, and the wider replacement question has its own answer in will AI replace lawyers.
Start With the Contract Your Junior Reviews Most
Pick the agreement your junior lawyer sees most often, build the Playbook from your own last five redlines, and have them run the next one with the Custom Instructions written. Then read the chat before the draft, and ask what they learned on Friday.
