Getting off the plane at Nashville International Airport last week — an airport set to be renamed after Dolly Parton by the time I left — I saw 5 legal tech ads before making it to baggage claim. In fact the only ads I saw on the airport’s video boards were two Harvey ads, an Everlaw ad, a Clio, a Litera, and a plug for a local Baptist Church. I assumed the latter wasn’t going to have a big presence on the exhibit floor.
Though if they were assigned booths on the wrong side of Harvey, no one would know if they were there or not. The legal AI behemoth snapped up 24 booths in a row to create an homage to the Berlin Wall, a dark-wood warren of faux law offices largely cutting off the booths in the next row over. There were billboards on highways, every Uber order flashed a legal tech ad, and Sheryl Crow played the Legora party. What word or phrase best sums up the legal tech environment right now? Bob Ambrogi mused we’re in “Peak Legal Tech.” That’s a solid option — we learned today that the final official attendee count was 5,782. To put that in perspective, last year the organization was ecstatic at the prospect of hitting 4,600. Others zeroed in on the industry’s newfound enthusiasm for partnerships and MCP connections as “everyone playing nice.” Somehow that doesn’t quite cut it though.
I’m going with: Polycule.
It’s about everyone coming together under one roof, and certainly involves a lot more jealousy and resentment than anyone is going to admit publicly.
As recently as a few months ago, vendors still hyped exclusive deals and single-panes. Providers claimed they were used by X percentage of the top 100 firms while their direct competitor claimed Y percentage and we all pretended X and Y didn’t add up to around 170. Every product claimed that it would soon become every lawyer’s preferred AI tool. Fast forward to now, and everyone seems to just assume that Biglaw will snap up every competing tool and let the lawyers sort it out. Some of that has to do with the lateral market. If a firm spends $100 million to land a group and that team arrives with different tool preferences… the firm has already committed $100 million, what are they going to say? Of course they’ll just buy the other tool.
To that end, most vendors spent this show eagerly announcing partnerships and MCP connections. It’s about every product making it as easy as possible to work with every other product. They can’t control what user interface will earn the eyeballs of each lawyer, but they can make sure their data or processes are in the mix. And with any luck they’ll emerge on top in the free-for-all environment the firm creates.
The single pane of glass got smashed
Over the last couple years, a lot of vendors explained how their AI interface would soon be the entry point for lawyers to manage everything. The single-pane would remove all the wasted time jumping between applications. I never understood how any one of them hoped to conquer the other AI products vying for user attention. It seems the answer is that none of them will.
Everlaw laid out a strategy best described as Everlaw everywhere. Connections to Gemini Enterprise for Legal, Microsoft Copilot, Harvey, and CoCounsel. The discovery platform has its own interface, of course, but the company wants to be the evidence layer and understands that the interface layer may shift to whichever product the lawyer prefers. Everlaw doesn’t need to be the user’s preferred interface, as long as the user is taking advantage of the company’s tools — why does it matter if it’s coming through CoCounsel or Harvey?
Speaking of Thomson Reuters, they spent ILTACON announcing partnerships at quite the clip. The Everlaw announcement, an iManage tie-up, Google… that’s on top of their big news of a full CoCounsel rebuild and their own home-grown language model. The company sees itself sitting on prized legal content and doesn’t care how people access the vault as long as they’re paying.
NetDocuments expanded its MCP partner roster to Google Gemini, Perplexity, Anthropic, Harvey, and Legora, all hitting the same Legal Context Graph the DMS uses itself. Box, which has 83 of the top 100 firms, described its posture to me as “headless AI” — roughly 1,600 integrations, native Box AI if you want it, Harvey or Claude if you don’t. Bundledocs, which classically helped paralegals put together documents, but has recently moved more up the workflow chain to meet more senior lawyers, has no ambition to be a generative AI hub, but lets you point Harvey or Copilot or ChatGPT at the work it does best.
Not everyone was banging the integration drum. DISCO is making the counter-bet that document review is a dedicated discipline that requires a specialized platform and tighter control over where the data goes. To that end, they’re more engaged in tying partners into their platform than plugging into others, fearing that no amount of MCP plumbing can consistently deliver the right results as a direct ask can. But by and large the chatter throughout the conference focused on the giving fellow legal-grade products the ability to draw information.
So maybe the single pane didn’t go away as much as the vendor drive to own that pane has. Let a thousand flowers bloom when it comes to which pane a lawyer chooses to interface with their tech.
Tokengeddon
While the show wallowed in all the money gushing into legal tech, cracks appeared in the happy peak legal tech mood. This is the part of the old Behind the Music show where the narrator says, “but behind the scenes, storm clouds were gathering.” At some point, someone is going to have to actually PAY for all this AI stuff.
The whole AI industry is built on subsidizing usage. Here’s a good explainer borrowing a popular meme format:
And this trickles down through all the various wrappers to legal. Flat-rate seat licenses are all well and good when the top of the chain charges flat fees, but when the consumption billing hits, what are lawyers going to do? The closing keynote panel conversation shared that clients have spoken generally favorably about outside counsel passing tech costs onto them directly. Putting aside how long that might last when they start seeing the bill, that still leaves the lawyer providing value without capturing the return. Here’s where we talk about the death of the billable hour or $10,000/hr lawyers. But whatever response the industry adopts, it’s going to have implications for what products firms want on the menu and what best practices they want lawyers bringing to their work.
To bring it back to the early-aughts, when Westlaw and Lexis charged by time, associates lived in fear of accidentally leaving a research project open.
One vendor described a company burning $75 million in tokens with no ability to say what it bought. For all the talk about ROI, no one seems to be able to consistently quantify that across tools. Throw in “agentic” AI, a concept that investors love, gobbling up tokens in far greater numbers than humans might, and firms have reasonable fears about what happens when they finally have to pay for consumption. Whole workflows are getting built atop the premise that this stuff will be affordable and everyone’s starting to worry that it won’t be and they won’t really know what can be cut and what’s indispensable to the work.
There’s a certain irony to law firms being pushed toward flat fees because AI compresses the hours they can ethically bill, while the AI vendors are moving from flat fees to consumption pricing.
Maybe clients will absorb the direct cost of the tech. But then what? The history of lawyer-client relations suggests the client will start nit-picking about why this model was used for this task and why so many tokens were burned on this ask. One thought raised in the closing keynote was to take a “data scientist, leading process engineer, business analyst,” and examine all the cost that it takes to get the optimal solution under the circumstances and price it accordingly. “Make it simple for the client.” Bringing us back to the flat-fee universe.
Which is why this show marked a shift from always talking about the “best” and “most accurate” results toward delivering results lawyers can use at the right cost. Some still bristle at the idea that lawyers could ever sacrifice quality, but we do it all the time! Assigning a draft to a first year instead of a fourth year sacrifices quality, but you do it because it’s cheaper and you trust that more senior lawyers will intervene and fix the shortcomings cheaper than if they did it all from scratch. AI shouldn’t be any different.
Getting to the point where less powerful AI spend achieves comparable quality is a matter of data and process. NetDocuments published its benchmark holding model and harness constant to get a real look at what good, properly enriched data can do for less powerful models. In a conversation with HIKE2 resulted in the memorable frame that accuracy problems in legal AI are usually data problems wearing a model costume. That sounds right.
Which brings us back to the polycule
It brings back memories of the early-2000s legal research wars, where Lexis and Westlaw gave up trying to become the exclusive provider to a firm and fought an unceasing war for associate eyeballs. They adopted the tactics of the most successful businesses of the 80s and 90s — drug cartels — and offered free samples and special prizes to law students to give school kids hooked on the product from their first legal writing assignment.
That’s where this battle seems headed. Who wins the most fans of their interface?
It’s an interesting development because we’ve been in a period of the firm technology staff being the primary driver of a vendor’s sales. Soon it’s going to matter just as much to capture the hearts and minds of the individual lawyers that rack up time on your product as opposed to the competitor that also rests on the firm’s system.
Everyone wants in on the MCP key party. Even if they secretly don’t want in on it, and just feel kind of pressured by their partners about the situation.
Joe Patrice is a senior editor at Above the Law and co-host of Thinking Like A Lawyer. Feel free to email any tips, questions, or comments. Follow him on Twitter or Bluesky if you’re interested in law, politics, and a healthy dose of college sports news.