The general counsel at a national retailer had signed the contract, cleared security review, and picked a start date. Then she named what she was worried about: her lawyers would not book the training.
That is what legal AI change management has to solve. Getting a legal team to change how it works is harder than choosing the software, and Chuck Kable, General Counsel and Corporate Secretary at Innovative Renal Care, put it this way on GC AI's CZ and Friends podcast, hosted by CEO Cecilia Ziniti:
"Change management with lawyers is really hard. You could be a lawyer for two years, three years, five years, 10 years. It's hard to get on board with something that's different than what they've historically experienced. There's this concept in change management called the Valley of Despair, and we're gonna get there. We are gonna bottom out, and you're gonna think, this stinks, Chuck, it's never gonna work."
A rollout plan has to survive that dip.
GC AI is an enterprise-grade legal AI platform built for in-house teams by a three-time general counsel. 2,000+ legal teams use it, including 200+ public companies such as Columbia Sportswear, Interface, and Eventbrite. The Enterprise tier includes managed procurement and onboarding, change management support, and ROI forecasting.
Three GC AI features come up most often in a rollout. GC AI for Word redlines clauses and whole contracts inside Word. Exact Quote attaches a citation to every claim. Our AI courses for legal professionals run 60 to 75 minutes, and Level 101 carries an hour of California MCLE credit.
Your Team Has Already Lived Through a Failed Rollout
In a 2022 Censuswide study commissioned by ContractWorks, an Onit group, researchers asked 350 in-house legal professionals across the US and UK about their legal technology implementations. The study found 77% had experienced one that failed, and 43% had been through more than one. Among those who had struggled to use the technology, 29% said it made them doubt whether their employer knew what was best for the business (Artificial Lawyer, 2022).
Those numbers set the odds on your own team. You are handing a platform to people who have probably already watched one of these fail. That memory will shape their response more than your rollout plan will.
For a lot of in-house teams, that failure was a CLM rollout. The reasons those projects failed were organizational:
No executive sponsor
Thin user training
A data migration nobody scoped
A gap between the team that picked the platform and the team that had to live with it
Every one of those was decided above the lawyer who had to use it.
AI has not fixed this. In a 2026 survey of 528 in-house legal leaders across six countries, run by InsightDynamo for Axiom, only 7% had moved past piloting to use, optimize, and measure AI across the organization, and 83% could not show whether the previous year's spending paid off. The same study found two-thirds of legal teams running general-purpose AI in its default configuration, which is how a team ends up with output it cannot trust.
The 9 Objections Your Team Can Raise
Most of these objections are reasonable. Here is what your team could say.
The objection | The answer |
"I have lived through a failed software rollout. Why is this one different?" | Name which failure mode is absent. The projects in that 2022 survey died on migration, configuration, and training runway. GC AI runs on the documents and software a team already has, with no migration project to sequence first. |
"One bad output and I am the one explaining it." | Set the expectation in week one and make every claim traceable to its source. See "The First Time It Gets Something Wrong" below. |
"I am not spending six hours learning software. My matters do not stop." | Put the lawyer's own matter through it and let that be the training. See "The Lawyer Who Will Not Book the Training" below. |
"My judgment is the job. If it drafts, what am I?" | The read-and-decide work stays with the lawyer. Kable calls the platform "a force multiplier for my support level folks, for my paralegal team." |
"If this makes us faster, does the company need fewer of us?" | Answer it directly or it festers. Say whether the plan is more capacity or fewer people, because your lawyers have already done the math and silence reads as the second one. |
"Where does this contract go, and who trains on it?" | Point at the subprocessor list. Every AI provider in the stack, OpenAI, Anthropic, and Google among them, processes content on a zero data retention basis, backed by SOC 2 Type II certification. |
"Someone above me bought this and now I validate their purchase." | Usually accurate. Axiom's study found that in most legal departments someone outside Legal picks the AI platform. Hand that lawyer the evaluation and let their assessment carry weight. |
"Every vendor says days, not months." | Make the mechanism the test. Ask what specifically makes it fast, and listen for an answer about migration and setup. |
"If it works, the platform gets credit. If it fails, I own the error." | Name the owner of a bad output in week one, before there is one. |
Two of those need more than a row in a table: the first wrong answer, and the training nobody books.
If You Are the Lawyer Being Rolled Out To
Everything above is written for whoever is leading the rollout. If you are on the other side of it, holding a login you did not ask for, you can use the same list to ask for what you need. Ask for three things:
A task you already know cold, so you can grade the output against what you already know.
A citation you can click through to the source language, on every claim.
A name, for whoever owns it when the answer is wrong.
Ask for all three in the first week, while you still have standing to shape how this goes.
The First Time It Gets Something Wrong
In a questionnaire of 252 legal professionals fielded in October 2025, Paragon Legal found that 67% had already had to override or correct an AI-generated legal output, and 47% said AI automation had sparked internal conflict inside their legal team. The ABA Journal covered the same findings. If the rollout has not planned for that moment, the skeptics get their argument.
Kable splits risk in a way that helps here:
"I have always advocated a view where you have to evaluate risk magnitude and risk likelihood separately. Because that's been the Achilles heel of a lot of businesses. They gloss over high magnitude risk because the likelihood is low."
Apply that same split to a wrong answer. The likelihood is high, and it is worth saying so plainly. The magnitude is what you control: it depends on whether someone catches the error before it goes to the business or outside the company, which is why verification has to be built into the workflow.
Traceability is what holds the magnitude down. A citation a lawyer can click and check keeps a wrong answer the size of a corrected draft. An untraceable paragraph goes out with the lawyer's signature on it.
So set the expectation in week one:
You are going to find errors.
Finding them is the work.
Nobody gets blamed for the platform being wrong.
Kable puts the leadership half of it this way:
"Your behaviors have to match your words. That is the most impactful way that you can let them know that I got your back, I support you, this is going to be okay."
Write the Review Rule Down
Write the review rule down before the first week, because the version your team invents under deadline pressure will be worse than the one you choose calmly. Three things need an answer, and they fit on one page.
Who certifies: name the lawyer who reviews AI-assisted work before it goes to the business or outside the company. For most teams this is whoever would have reviewed the associate's draft.
What goes where: decide which matters run in the enterprise instance and which never touch a model at all. Privileged investigations and anything under a litigation hold are the usual carve-outs.
What gets kept: log the prompt and the output for work product that leaves the department, so a question six months from now has an answer.
None of this needs a committee. A general counsel can write all three in an afternoon, and once they are written you can tell a nervous team the guardrails exist before you ask anyone to trust the platform.
The Lawyer Who Will Not Book the Training
The retailer's general counsel was right to worry. Her lawyers had matters that did not pause for a calendar invite, and a training block reads as an hour taken from a day that was already short.
Kelly Noguchi, Senior Legal Technology and Operations Manager at Instacart, described what the ask feels like from inside a team:
"I feel like we are expected... to be basically testers, QA people on top of their day jobs all the time. They have to continuously learn, continuously be creative."
She also named the gap between a signed contract and a working department:
"I think we can't glaze over... changing excitement into execution. That's real."
The way around that is to run the lawyer's own matter and let that be the training. Nicole Altman, Associate General Counsel at Instacart, put the method in one line, then added the part rollouts skip:
"The best way to get people into using AI is to show concrete examples of what it can do and how to do it and walk people through it. Not just once, but periodically. And leveling up a little bit each time, so it's not boring."
A single kickoff session only ever reaches whoever showed up that day. Make it short, recurring, and built on live work. It picks up the people who skipped the first one, because by then their colleagues are doing something visible with it.
How Experienced Operators Run Legal AI Change Management
The four moves below come from in-house leaders describing what worked on their own teams. Each is something the person leading the rollout has to do personally.
Let Your Most Demanding Lawyer Test It First
On the same conversation with the Instacart team, Ziniti raised the adoption curve that shows up even inside a forward-thinking department, then told the story of a team that pointed her at its hardest audience:
"Our corporate guy, he's the Eeyore of the group. If you can get him using AI, you will have won... I'm very happy to report that Eeyore is using GCAI to do his proxy statements."
That lawyer is a quality gate. Their standard is the one the platform has to clear, and their sign-off carries more weight inside a department than any mandate from above. Give them a task they know cold and invite them to break it.
Try Again on What Failed Before
Altman's team rebuilt a triage process for the hundreds of emails a week arriving at a shared legal inbox, work that had been close to somebody's full-time job:
"We had actually tried it last year and it didn't work. But tools evolved, things evolved, we tried again."
Match Behavior to Words
Kable's rule is simple: a leader's behavior has to match what they said. In a rollout that means using the platform visibly yourself, on your own matters, before asking anyone else to.
Name the Dip in Advance
Kable tells his team about the Valley of Despair before they reach it, so when adoption dips they read it as the phase they were promised rather than as evidence the skeptics had it right all along.
Where GC AI Fits for In-House Teams
Four of the objections above trace back to something specific in the product.
Start with the fear that this rollout looks like the one they already lived through. With GC AI's Contract Intelligence, a team can read a whole portfolio wherever it already sits, in Google Drive, SharePoint, OneDrive, Dropbox, or Ironclad, and turn it into a searchable view without tagging or consolidating anything first. That skips the step those CLM projects stalled on.
The accountability question has the same answer. Exact Quote lets a lawyer check any claim against its source in the time it takes to click, which makes the week-one ownership rule enforceable.
Security has a paper trail. GC AI's subprocessor list names every AI provider in the stack and its retention basis in plain language, backed by SOC 2 Type II certification. Hand the list to whoever runs security review and let them draw their own conclusion.
Every vendor promises fast. Here is the number for us: in GC AI's February 2026 survey of 200 in-house legal professionals, 62% saw value within the first hour. Hold any other vendor's timeline promise against that.
Everything else on the list, the training time, the headcount question, who signs off, comes down to how your team runs the rollout. Training is open to every team, regardless of tier: more than 8,000 lawyers have already gone through GC AI's courses, free to take. Enterprise adds change management support and managed onboarding for teams that want a dedicated partner on the rollout.
Lead the Rollout on Live Work
Pick a platform your team can check, hand it to your most demanding lawyer, run it on real matters, and tell everyone about the dip before it arrives.
Frequently Asked Questions
What Is Legal AI Change Management?
Legal AI change management is the work of getting a legal team to use an AI platform on real matters after the purchase is agreed. In practice that means setting expectations before the first wrong answer, running training on a lawyer's live work, naming who owns an error, and planning for the dip in adoption that follows.
Who Should Own a Legal AI Rollout, Legal or IT?
Legal should own the outcome, with IT as a partner on security and access. The common failure is IT handing legal a login and moving on, which leaves nobody accountable for whether lawyers use the platform. Name a lawyer or legal-ops lead who owns adoption and give them authority to redesign the workflow around it. GC AI's Enterprise tier includes change management support for teams that want a partner on that work.
How Do You Measure Legal AI Adoption?
Measure it by weekly use on live matters. Track how many lawyers open the platform each week, which workflows have moved onto it, and the time saved on those tasks. A deployment that looks active in the admin panel while weekly usage stays flat has stalled, which is a signal to coach the team before buying more licenses.
Should You Pilot Legal AI With a Small Group or Roll It Out to Everyone?
Start with a small group when your team has been burned before, because a pilot that produces checkable results gives you internal proof no vendor reference can. Pick two or three lawyers whose standards are high, give them live matters, and let their verdict drive the wider rollout. GC AI offers a 14-day free trial, which is long enough for those lawyers to grade it on work they know cold.
How Long Does a Legal AI Rollout Take?
Plan on one full cycle of real work, which is when you will know whether the rollout is working. Individual lawyers can see value in the first week, and full team adoption takes longer. Expect a dip in the middle as lawyers trade a familiar workflow for one they are still testing, and judge the deployment after the team has climbed out of it.
