What is it like to use a legal AI tool built by a three-time general counsel and named after the role itself?
In this Legal Intelligence Platform deep dive, Anna puts GC AI through work that occupies real in-house legal time: revising a contract and researching law in an unfamiliar jurisdiction. Along the way, she shares practical techniques for getting better outputs from any AI tool used for legal work.
This video was made in partnership with the featured vendor. The vendor provided product access and reviewed the video for factual accuracy. Editorial opinions and conclusions remain those of Legal Benchmarks.
00:00What? Hold on. GCAI is telling me that I just saved 8 hours and 3 minutes on legal work this week. Does that mean I get the 4-day work week now? Hello. Welcome back to another episode of the Legal Intelligence Platform Deep Dive series. My name is Anna and I am a lawyer making sense of AI for the legal profession. Today, I will be exploring GCAI, a legal intelligence platform that is tailor-made for in-house lawyers.
Speaker change: So, over the past year through Legal Benchmarks, I've been researching how AI tools actually perform a real legal work. You can check
00:32out our latest rankings at legalbenchmarks.ai. I've also been doing deep dives into different legal AI platforms to understand what they're good at. Today, in addition to exploring GCAI, I'll be sharing my top tips on how to get more high-quality outputs from any AI tool you use for your legal work. So, you'll want to stick around for this one.
Speaker change: And no, we're not going to be using GCAI to review an NDA. I think we're all tired of NDA demos at this point. GCAI is built by a three-time general counsel with a clear focus on in-house legal teams. Today, I will be using GCAI
01:03to work through two common workflows that in-house counsel spend a lot of time on: revising a contract and legal research and analysis.
Speaker change: Let's start with where GCAI lives. Today, you can find GCAI in three places: in the web browser, in Word as an add-in, and also on your mobile. Yes, you heard me. And the best part is they all connect. So, if I'm reviewing a contract in Word and I ask GCAI a question there, that conversation history and context can carry over into the web app.
01:33Speaker change: And if I leave my desk, I can pick it up again on mobile. GCAI is designed for a lawyer to be productive all day, regardless of where they are. And who wouldn't want to be productive all day? Hopping over to Word here so I can show you how I would use GCAI to help me with contract revision. In this example today, I am an in-house counsel representing Lovern, a native AI legal consultancy based in Finland, and they are signing a lease agreement with the landlord Hagora. The landlord has asked
02:03for 3 months security deposit to be returned at the end of the lease after deductions and any tenant breaches. We negotiated two wins. The first one is the deposit drops from 3 months to 1 month and any deductions now needs sign off first from Laverne. I can see for GCAI's word added that before I do anything, I have to first select that I'm representing. So, I'm going to go ahead and select Laverne here, deselect Hagora, and click continue. This is a really smart workflow design because it
02:35forces a user to be very clear to the AI at the beginning which side it should represent and tailor its responses to. I'm going to see if GCAI can handle this request off the shelf. This commercial term update seems really simple on the surface, but it actually isn't because not only does GCAI need to figure out what 1 month's means. It also need to do some math because a signing sum, as you can see here, is calculated by the addition of the first month's rent and also the security deposit, which is
03:06going to change from 3 months to 1 month. In order to complete this task correctly, GCAI needs to do math correctly and figure out what the signing sum is and update those figures accordingly. It also needs to figure out where the deduction mechanism is for the security deposit and update it. So, let's take a look at what GCAI comes up with. There's quite a few actions I can take. I can jump to each of the proposed amendments and it's highlighted in the doc. That is super helpful. I can either
03:36edit the insertion manually before I insert it or if I don't like what I see, I can regenerate it with AI. What I like to do is accept all of them so I can see the amendments in context and then figure out what I like and don't like from there. The summary of the changes is pretty short and sweet. What I really love is at the end it gives you a risk flag, which is telling me that Leveris consent for all deposit deductions is a very strong positions and that Haggar will likely push back and what the back
04:06pocket position is. This makes me feel like it's not just like a robot that is just doing what you're asked, but it's thoughtfully providing you with negotiation advice that is actually legitimate. This extra touch is really appreciated and I find extremely useful. TCI was able to amend the security deposit and the signing sum correctly. It also added the subject to the set tenants approval for the deduction of the security deposit in the right sections. The redline insertion is pretty precise and it follows the
04:36existing format. It was also able to find buried in the conditions precedent another reference to the payment of the signing sum that needed to be updated. I definitely did not expect TCI to catch a signing sum figure that was buried under section 22 because it's like in the middle of the agreement. I I usually have a hard time finding all the relevant terms that need to be updated, especially the ones that have cascading effects through the rest of the agreement, but it's able to identify in
05:06this scenario. I also gave Claude's word again a whirl on this task and it was not able to pick up that the signing sum needs to be updated. And you can also see that the amendments proposed under section 16 is just a sea of red. It just block deleted a whole section and replaced it with wording that is technically correct, but it did so, I would say, less elegantly. I have tried to ask Claude to make more surgical edit. It didn't really make that big of a difference in how it marked up the agreement. So, this is definitely a gap
05:37that I see for Claude compared to purpose-built tools for redlining. I have two tips here for folks that are using AI tools for redlining. The first is because these systems are non-deterministic by nature, it means that each time you generate a red line, you might get very different results. I tested the same prompt a few times on GCAI and Claude, and each run handled the revisions a little differently. So, I would always suggest for clauses that you're a little bit more unsure about or want to see some variations to
06:07regenerate a few versions to see which ones are the best. Second tip here is to break a big task down into smaller tasks. So, you can see here I am asking the AI to both amend the security deposit and also add that the tenant's consent is required for any security deposit deductions. Another way you can do is by breaking down into two separate tasks, so then AI is not using as much context window as it otherwise would. It's really interesting that GCAI has a context usage tracker here, but AI can
06:40actually lose focus pretty quickly, and I would definitely recommend everyone to start a new conversation with every new query, even if you still have a lot of context usage left. All right. So, next I want to show you today how I would use GCAI to help me navigate law in unfamiliar just jurisdiction. I'm going to be using the same lease agreement between Hagora and Leveron over a office space rental in Finland that will be governed by Finnish law. What I want to find out is whether this contract is
07:11enforceable under Finnish law. So, I can see that GCAI has spun up a few sub agents to do this research in parallel. It first gave me what the requirements are under the law for a contract to be enforceable. I'm also able to see an excerpt from the source without having to leave GCAI, which is useful. It also cites from my original agreement and highlights the relevant sections of the agreement that it's referring to in this
07:41analysis, which is pretty helpful for me to verify as well. The legal sources that it cites to seem to not from the primary source, but more so from secondary sources. We can see the relevant excerpt here within GCAI, but if you want to read the entire article in context, you would have to visit the website. Here are my general tips for using AI tools for legal research and analysis tasks, especially in area that you might not be very familiar with. The first one is use AI to help you frame
08:11the question better. As you can see here in GCAI, there is easy prompt feature, and that can help you expand the queries and angles that you may have otherwise not considered. But, it's very important not to blindly accept everything that is in the easy prompt output because there could be a lot of extra things in there that you don't want, and that could crowd your context window, and sometimes prompt enhancers in general can introduce additional assumptions that deviate from your original research intent and actually give you an output
08:42that is not what you're asking for. So, I would be very careful with that, but it is a very useful feature for you to get started. If you don't have a legal specific tool, you can always ask the general purpose AI, "What are some of the key concepts to the Spiegel question that I should consider?" This helps you deal with the problem of you don't know what you don't know. And the second tip is don't accept the first answer. Always push back. So, how I would do this is, for example, in enforceability question, we got a pretty exhaustive answer here,
09:13but I can always follow up with, "You are not comprehensive enough about identifying all the risks." And just, you know, put a little bit of pressure on the AI to regenerate and to give you a more comprehensive answer. There has been some studies that suggested that threatening your AI, although it's not recommended, can perhaps increase its performance. So, it is definitely worth trying. But, at the very least, definitely push your AI to give you a second answer. The third tip I have is cross-check against different solutions. So, even if you are using a legal
09:44specific tool, I would still compare the answer against Claude or ChatGPT if you have an enterprise license because they're useful as a sense check. And if both tools reach the same conclusion and cover similar points, that gives you a little bit more confidence. But if they disagree or one tool covers the issues that others missed, then those are the areas that you need to pay closer attention to. So, for this specific task, I did use both GCAI and Claude to analyze enforceability issue. So, did GCAI actually save me a day's work on
10:15this lease agreement? According to GCAI, time saved is calculated using assumptions like 5 minutes saved per web message. But I'll be honest, it does feel a bit difficult to verify that kind of claim because legal work is not always easy to reduce to minute saved per message. There's no way I would actually know. But did it help me speed up the work and lower the cognitive burden while increasing the quality of my work product? Absolutely, yes. So, here's where GCAI lands for me. It's a tool that is built to support the substantive work of in-house lawyers. From
10:45redlining, review, drafting to research, the outputs are thoughtfully tailored to the details that lawyers care about. Oh, and lawyers care a lot about the details. US case law cited in the Bluebook format, one-click comments you can generate to both internal stakeholders and external counterparties, risk flags that counsel really appreciate. All these tiny details are things general purpose solutions usually don't tend to optimize for, but vertical solutions do because they care so much about the customers.
11:15And you can find these thoughtful touches all throughout the product experience at GCAI. The other thing that GCAI does really well is lowering the adoption barrier for users even before they try out the product. They offer a free trial, month-to-month subscription, prompting classes on a frequent and consistent basis, and a very lively Slack community for peer support. This is by far the best I've seen in the industry, and it makes lawyers feel fully supported in trying out and feeling confident
11:46using AI. But, here's the flip side. GCAI is not a contract repository or a workflow builder. It can sit alongside those platforms, but not replace them today. So, if your main goal is to empower your legal team to do their work more productively, then GCAI is a great option. However, if your main bottleneck is managing the intake,
Speaker change: storing your contracts, and letting business self-serve without legal, then GCAI is not the first option I would reach for today. If you found this product deep dive video useful, please
12:16subscribe to this channel or leave a comment below to tell me your experience using GCAI and what other tools I should check out next. And don't forget to check out my interview with Shishir Zanetti, the founder and CEO of GCAI, where we go deep into GCAI's product roadmap and what they think the future of legal work looks like. Until next time, I hope you stay healthy and curious. Ciao.
