In-house legal teams are getting hit from every direction: Slack pings, emails, last-minute deal questions, and "can we deviate?" requests that somehow become urgent five minutes before signing.
In this video, I go under the hood of Wordsmith, a legal enablement platform, to see what it actually does in real in-house work.
| Time | Chapter |
|---|---|
| 0:00 | Intro: why I'm reviewing legal intelligence platforms |
| 1:37 | Why Wordsmith, and what I'll cover |
| 2:10 | Getting started in the web app (cockpit) |
| 2:41 | Repositories, Playbooks, and Blueprints overview |
| 3:01 | Blueprints demo (dynamic NDA template + deliverable) |
| 4:40 | Quick reality check: can ChatGPT do this too? |
| 5:34 | Why workflow support often matters more than raw accuracy |
| 5:44 | Building an agent trained on your own policy |
| 7:26 | Testing the AI agent |
| 8:50 | UX notes (what was harder to navigate as a new user) |
| 9:06 | Security notes (start with the trust center, then validate internally) |
| 9:21 | Who Wordsmith is for, and when it's overkill |
| 10:04 | Part 2 teaser: my interview with Ross, the CEO of Wordsmith |
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.
00:00Hi, I'm Anna and I'm a lawyer that's trying to make sense of AI for the legal profession. I've seen this industry from three different angles as a private practice lawyer inhouse at big tech and also a founder myself. On top of that, I published two legal AI benchmarking reports together with lawyers in the community in 2025. This mix of a background has made me very very curious about what legal AI startups are building. So, in this series, I'll be exploring some of the coolest startups
00:30out there to see what is underneath the demos. If you're curious about whether these tools are just AI rappers, who they're building for, and how they're going to change the legal profession, then you're in the right place. Follow me and let's get started. My prediction for 2026 is that this will be the year for the emergence of legal intelligence platforms, LIPs, lips. Actually, let's just stick to LIPs. We're starting to see the term legal intelligence pop up everywhere.
01:02And they're referring to solutions that can leverage legal information from within the organization and also externally to generate legal deliverables or insights that reflect sound judgment and also analytical depth. Wordsmith is the first company I'd like to explore in this emerging category. Wordsmith is a legal enablement platform that is designed to unblock in-house legal teams by enabling them to perform a range of legal tasks and also build their own AI agents that
01:32is directly integrated into business workflows to perform those tasks. Wordsmith first came on my radar when I was just scrolling LinkedIn and I saw a lawyer refer to wordsmith as a gamecher and magical. Today I'll show you three things. Number one, how an in-house lawyer can get started in WordSmith. Number two, the core functions I find the most useful. And number three, who I'd recommend this for and who I wouldn't. Wordsmith exists on many different interfaces. The web app is one. Slack, email, notion, Xavier, you
02:04name it. Word, of course. But I just want to show you how in-house lawyer can get started on the web app. Welcome to the cockpit for the in-house legal team where you are the pilot. Once you log in, the first thing you can do is set the context for your day-to-day work on the top right corner. This is where you can define the jurisdiction you support and select all that apply and the company profile. Once your context is set, you can start in a few ways. Chat with the general AI assistant at the center of the page or use the quick actions at the bottom. There are also a number of modes and feature toggles you
02:35can see here in this control panel. What I want to focus on are the core functions at the left side panel. Let's start with repositories. Repositories function like a knowledge base. You can upload your own documents and they become the context for your AI assistant to reference. You can upload the sources from a variety of different avenues. Next is playbooks. You can create your own playbooks or use Wordsmith's built-in ones. Then we have blueprints, which is where things start to get very interesting. So, blueprints let you turn
03:06a document into a reusable dynamic template with fillable fields like company names, signing entities, deal contacts, and other variables that can change each time. Here is how it would work. I just found an NDA template that I really like from my mine and I downloaded it locally and I want to adapt it for my own use case. So, let's upload the file. Once it's uploaded, you'll see two things. The variable fields in the document are highlighted. You can further customize and edit the template here manually. So, I'm going to change the entity name to legal benchmarks
03:36and adapt the other fields as needed. Once I'm happy with the document, I will publish it and I can start drafting my NDA here. At this point, I can draft it in two ways. Either fill in the form manually on the right or instruct the AI invoice mode. Please draft a mutual NDA between legal benchmarks and WordSmith for a collaboration to establish a law school focused on training future legal AI engineers with a goal of reaching the top 10 world rankings by 2027. Wordsmith will generate a first draft based on what I provided and if any key details
04:06are missing, it will flag what it still needs me to provide. But I can continue working on this offline. I want to see how well it preserved the formatting of the document. You can see here is the updated file. It has the requested changes I made. But I'm pretty happy with this first draft. You can see on the left, it's really preserved the formatting of original document on the right quite well while adding the specific business context for my deal. And this is especially helpful for in-house lawyers who don't have as much parallegal support because formatting
04:37takes up a lot of a lawyer's time. Quick reality check here. Can chat do this too? I uploaded the same NDA and did a quick test. One eternity later.
Speaker change: After more than 27 minutes, we finally got our deliverable ready for download. So, let's take a look at what Chad GPT came up with. Failed. Failed. In the end, I was not able to download the DOCX file from Chad GPT. And this is exactly why I think
05:08accuracy performance should only make up a part of the consideration for legal teams making tech procurements. In both of our legal benchmark studies, we've seen general purpose AI and legal AI look surprisingly close on accuracy performance for the tasks we tested. The differentiator is often everything around the answer, how usable it is, and how well the platform supports real legal workflows. Formatting is such a great example of this. It's not enjoyable work, but it's where legal
05:38teams often lose a surprising amount of time. Guys, what I'm most excited about and what I find most unique is the feature that allows you to build your own agents that are trained based on your own knowledge base to assist the different stakeholders. This just sounds very futuristic. So, I am going to try it for myself and build my own anabots and see whether it can help me fend off business. Just kidding. I love business.
06:09When I was an in-house lawyer, I hated handling the routine questions that are just seemingly very simple. Which entity should sign this contract? Hey, it's in a policy. Can we deviate? No. I have said many times that we cannot deviate from our governing law policy. Oh, no. But this is the blocker now. Legal, you're the blocker now. Anyways, there are a lot of these scenarios that would be really, really helpful. So, let's see how I can do this. I start out in the repository, which I am already at. And what I did is upload a governing law
06:41policy. And this I've done by connecting to my Google drive and uploading this policy here. After uploading the policy, I went into the configuration to set my custom instructions. And you can do this in advanced here. So this part is like the brain in the engine for my agent. Uh anyways in the instruction section I did a reference assistance to create this rules for the repository which is basically telling it how it should behave. So after I've given this context
07:12um I also set up additional email instructions there on how it should respond to an email with the different steps and also a few short examples there as well. I did play around a bit so that it is giving me the appropriate answer that I want. It's an iterative process. Let's test this thing out. How I am signing a soul? Is it soul or so? So, please help me decide. I am signing so using the pack me SG entity. Which governing law should we use? Okay, I've
07:43sent this email out as business and now I'm waiting for an answer. Waiting, waiting, waiting. The moment of truth. Governing law policy has responded and it says under the governing law policy, the so should be governed by Singapore law. Correct. In my example, I use email because maybe I'm a little bit too old school now, but Wordsmith is Slack native. So, if your company lives in Slack, you can deploy your agents inside the channels of Slack to meet your teams in their comfort zones. That being said, if this was a real working scenario, I
08:14probably wouldn't toggle on autopilot unless it's for some very simple, straightforward queries because just with this level of testing, I'm probably not going to be able to anticipate all of the edge cases. All right, so here's my take after exploring WordSmith. From a functionality perspective, WordSmith is not a point solution. It aims to build a workspace that supports the full spectrum of the day-to-day of an in-house legal council from as simple as drafting a clause all the way to
08:44enabling the legal team to automate service delivery for the rest of the organization. From a user experience perspective, I definitely found the web portal a little bit difficult to navigate as a new user. It's great if you know what you're looking for, but I can definitely see a legal team needing some training or onboarding support to fully maximize the potential of the platform. Data security and privacy are a core part of assessing any enterprise solution. Wordsmith provides a trust center and states that it is SOCK 2 type
09:152 certified and GDPR compliant which is a good starting point to validate against your team's requirements. So which teams would benefit the most from Wordsmith? If you mainly need a lightweight co-pilot to help you with contract summarization or publishing an email? If your work is more bespoke and ad hoc or if you don't have time to set up and maintain workflows, then I would say WordSmith may be an overkill. If you're working in a tech scaleup or a multinational organization where there are lots of inbound requests coming in
09:45from all different communication channels for legal then the value of a platform like wordsmith is not just in supporting the legal team with any specific task in isolation. It is enabling legal to scale support across all of those channels so that business can move fast with confidence. Stay tuned for part two where I sit down with Ross, the founder of Wordsmith, for a candid conversation about why legal teams should trust Wordsmith, why platforms trying to serve both in-house and law firms can be conflicted, and
10:15what leveling up really means for lawyers when AI agents start doing the work. If you want a more systematic way to assess tools like Wordsmith, I'm also building an assessment framework with lawyers in the community and I'll link it in the comments below.
