In partnership with
For two decades, buying legal software meant the same dance: the vendor pushes a three-to-five-year term, you take the discount, everyone locks in. Both sides just stopped dancing, in the same week. This issue is about what to sign instead.
THE LEAD PLAY
The Deal Behind Your AI Tools Just Flipped — On Both Sides
On July 1, Law360 Pulse reported that legal ops teams are increasingly capping AI tool contracts at about a year, betting that the ability to walk away from the wrong product is worth more than the discount that made three-to-five-year deals attractive. The example that opens the piece says it plainly: when the legal ops team at auto glass giant Belron built its AI strategy, it warned the department's attorneys not to get too comfortable with any tool.
Vendors are moving at the same time, in the opposite direction. Legora shifted its most capable product, Agent Pro, to consumption-based pricing in late June. (Same platform Kirkland locked up exclusive custom-build rights on, if you caught Issue #3.) Anthropic and OpenAI already price this way, and The Lawyer's read is that firms should expect their AI cost base to rise. Legora's framing is friendlier: pay for what you use, track spend by matter on a real-time dashboard, set thresholds. But a threshold is a thing you install when the meter can run.
Per-seat pricing worked because using software more didn't cost the vendor much of anything. AI agents broke that. Every unit of work an agent performs costs the vendor real compute, so flat pricing bleeds them, while a long lock-in strands you with a product that might get lapped in six months. I've spent years working with Litify, the Salesforce-based legal SaaS where the seat count basically was the negotiation, and the per-seat license was never just a pricing choice. It was the entire commercial logic. Both ends of that logic came apart in the same news cycle.
One caution before you celebrate the flexibility: consumption pricing charges you for the work the agent does, not for that work being right. A bad run bills like a good one. That's a procurement problem, and procurement problems get solved in the contract.
The Play this week: Pull your next AI renewal forward and change the deal, not just the tool. Ask for a 12-month term, and price the walkaway honestly against the multiyear discount. That discount is compensation for a risk the vendor wants you to carry. If the pricing is consumption-based, put the usage dashboard, hard spend caps, and matter-level cost attribution into the contract itself. Matter-level attribution is the sleeper term: it's what makes an informed client-billing conversation about AI costs possible later. And get data portability at exit in writing while you still have leverage.
Thinking about hiring globally? Start with an EOR.
The best person for your next role might not live near your office—or even in the same country.
More companies are realizing they don't need to open entities everywhere just to access global talent. Instead, they're using EOR to hire internationally faster, stay compliant, and avoid building local infrastructure before they're ready.
Oyster's EOR helps companies hire, pay, and support employees in 180+ countries while Oyster handles payroll, compliance, taxes, and local employment requirements.
SUPPORTING PLAY 1
Cooley Is Giving AI Away to Startups. That's Not Charity.
Cooley launched Cooley GO Lab this week: an AI-powered workspace for startups, built with support from Legora and the startup accelerator Y Combinator. It sits on top of Cooley GO, the free document-and-guidance hub the firm has run for founders for years. The new part is the product. Startups get working AI tools from the firm before they've paid the firm a dollar.
Anyone who has run intake will recognize the shape. A founder who incorporates through your free tools, keeps documents in your workspace, and learns your forms isn't shopping for counsel when the Series A hits. They're already home. Cooley spent a decade building that funnel; GO Lab pours concrete into it.
The instinct at a mid-market firm is to file this under things only Kirkland and Cooley can afford. Wrong file. Building a scoped, client-facing AI workflow costs a fraction of what it did eighteen months ago, and your attorneys already answer the same twenty client questions for free every year.
The Play this week: Pick one question your firm answers for free anyway (intake screening, a compliance checklist, a 'do we actually have an issue' first pass) and scope what a self-serve AI version would look like. You're not shipping a product this quarter. You're finding out what client-facing AI teaches you about your own funnel while the lesson is still cheap.
SUPPORTING PLAY 2
Your Clients Think They're Winning the AI Race. Against You.
Law360 Pulse reports that corporate legal teams now see themselves as the primary drivers of the AI innovation cycle, and that top law firms don't agree. Put that next to Bloomberg Law's Leading Law Firms survey, where all 40 responding firms with 500-plus attorneys reported using legal-specific AI tools, and you get one relationship with two parties each convinced they're the sophisticated one.
A perception gap like that stays harmless right up until it surfaces in an RFP. And a client who believes they lead on AI reads your bill differently: they're pricing your hours against what they think the technology should have saved, whether or not they're right.
The Play this week: Firm-side, ask your top three clients one question at the next check-in: where would you rank our AI use against your own department's? Then listen without defending. In-house, run it in reverse and tell your firms what efficiency you expect AI to show up as in staffing and fees. The conversation costs nothing. The surprise later doesn't.
QUICK HITS
The voluntary framework from last issue's Quick Hits still hasn't landed. What did land: the administration had OpenAI stagger GPT-5.6's rollout, with the government approving access customer by customer during the preview period. If a tool in your stack leans on a frontier model, its roadmap now has a political dependency. Worth asking your vendor where they sit in the queue.
OpenAI proposed handing the U.S. government a 5% stake, under an arrangement that envisions Anthropic, Google, and Meta ceding similar stakes through a sovereign wealth fund. Anthropic says no such talks have happened on its end. If this goes anywhere, 'who owns your AI vendor's model provider' becomes a conflicts question instead of a trivia question.
The Florida Bar became the first state bar in the country to offer members free legal AI: four months of unlimited access to Clio Work plus training and a certification path, with registration open now and access expected later this year. If you manage Florida-barred lawyers, this tool arrives in your environment whether you procured it or not. Update the inventory.
A Canadian court reportedly set aside an arbitral award over the arbitrator's overreliance on AI, apparently a first. It surfaces a question nobody has a clean answer to yet: short of obvious hallucinations, how would you even tell? If your standard contracts route disputes to arbitration, that question now lives in your clause library.
That's Issue #4. Before you sign the next AI renewal, read the term length before the feature list. See you in the next one.