Nvidia just announced an enterprise agent toolkit built with SAP and Salesforce. That’s not a headline most law firm partners would circle in red, but it should be. The announcement tells you everything about where AI is heading in the next year and what that means for your firm’s technology decisions right now.

Here’s the short version: if you’re thinking about building custom AI agents or hiring a dev shop to wire up GPT-4 to your practice management system, wait. The market is about to hand you better options without the six-figure bill or the maintenance headache.

The Nvidia Signal and What It Means for Legal

Nvidia built its name on graphics cards and gaming rigs. Then it became the chip behind every serious AI model. Now it’s packaging agent-building tools for enterprises and partnering with the two biggest names in business software. That’s a signal. When infrastructure companies start selling picks and shovels to the enterprise layer, a wave of vertical applications follows 6 to 12 months later.

SAP and Salesforce aren’t legal platforms, but they’re the proof of concept. If Nvidia can help those companies build domain agents for finance, supply chain, and CRM workflows, the same toolkit will power legal-specific agents soon. The difference is that law firms won’t need to build them. Software vendors who already serve legal will integrate the same stack and ship agents that understand conflicts checks, billable hour tracking, and discovery workflows out of the box.

That’s the pattern we’ve seen in every technology wave. The first movers spend a fortune on custom builds. The second wave buys a product that does 80% of the work at 20% of the cost. Right now, law firms are at the decision point. You can be the first mover and pay for it, or you can wait a year and buy a mature product that someone else debugged.

The Real Cost of Building Too Early

A mid-sized firm I spoke with last quarter spent $180,000 building a custom intake agent. They hired a consultancy, fed it six months of intake data, and trained a model to classify new leads by practice area and urgency. It worked, sort of. The agent could route emails and score leads, but it couldn’t handle phone calls, couldn’t check conflicts in real time, and couldn’t book consultations without a human stepping in.

The firm got a partial solution and a maintenance contract. Every time their practice management system updates, the integration breaks. Every time the model drifts, they pay for retraining. They’re locked into a vendor who built something bespoke, and the total cost of ownership is climbing past $300,000 over three years.

That’s the risk of building now. You’re paying for the education of the dev team, the iteration cycles, and the inevitable rewrites when the underlying models improve. Nvidia’s toolkit doesn’t eliminate that work, but it does commoditise a big chunk of it. When legal software vendors adopt the same stack, you’ll get agents that are pre-trained on legal workflows, pre-integrated with the platforms you already use, and priced like software instead of like custom development.

The question isn’t whether AI agents will transform legal operations. They will. The question is whether you want to fund the R&D or buy the finished product.

Where Law Firms Leak Money Right Now

Let’s talk about the work that’s actually breaking your economics. Most firms I audit leak $80,000 to $250,000 a year across three areas: unbilled time, intake conversion, and document review. These aren’t rounding errors. They’re the difference between a partner taking home an extra $60,000 or watching it evaporate into admin work.

Billable-hour leakage is the big one. Every attorney in your firm spends 4 to 6 hours a week on work that never makes it onto an invoice. Reviewing intake forms, drafting conflict memos, summarising case files for a partner, chasing down documents from a client. It’s real work, but it’s not billable. Multiply that by your headcount and your blended rate, and you’re looking at $100,000 to $200,000 a year walking out the door.

Intake delays are the second leak. A high-intent lead calls your firm at 6 PM on a Thursday. They leave a voicemail. Your intake coordinator listens to it Friday morning, calls back Friday afternoon, and leaves a voicemail. By Monday, the lead has hired someone else. We see 30% to 40% of after-hours intake never convert, and most of that loss is just speed. The firm that answers first wins.

Document review is the third. Junior associates bill $200 to $400 an hour to read contracts, flag issues, and summarise positions. It’s expensive, it’s slow, and it doesn’t scale. When a new matter comes in with 500 pages of discovery, you’re either eating the cost of first-pass review or passing it to the client and watching them flinch.

These are the workflows where AI agents make sense. Not because they’re flashy, but because they’re repetitive, high-volume, and expensive when humans do them. The firms that solve these three problems first will pull ahead. The firms that wait for the perfect custom solution will keep leaking cash while the market moves.

What an AI Agent Actually Does in a Law Firm

Let’s walk through what an Intake Voice Agent looks like when it’s working. A potential client calls your firm Saturday morning. They were in a car accident Friday night, and they want to know if they have a case. Your phones roll to voicemail because no one’s in the office. With a voice agent running, the call gets answered.

The agent introduces itself, explains that it’s an AI assistant helping with intake, and asks the caller to describe what happened. It listens, asks follow-up questions, and captures the details: date of the accident, injuries, insurance information, whether the other driver was cited. It runs a conflicts check against your case management system in real time. If there’s no conflict, it offers to book a consultation with an attorney and sends a calendar invite while the caller is still on the line.

The whole interaction takes six minutes. The caller hangs up with a consultation booked for Monday at 10 AM. Your intake coordinator gets an email summary with the key facts, the conflict check result, and a link to the recording. The attorney walks into Monday’s meeting with a brief already written. No voicemail tag, no delay, no lost lead.

That’s one agent. Now add a Matter Triage Agent that reads every email and form submission that hits your intake inbox. It classifies the practice area, scores the lead based on case value and fit, and routes it to the right partner with a one-paragraph brief attached. The partner sees it in their inbox 30 seconds after the lead submits the form, and they can decide in 60 seconds whether to respond or pass.

Add a Document Review Agent that performs first-pass review on contracts, discovery batches, and matter files. It flags risky clauses, summarises each party’s position, and produces an associate-grade memo. Your associate still does the final review, but they’re starting from a 90% draft instead of a blank page. A task that used to take eight hours now takes two.

These aren’t hypothetical. We build all three at Enterprise DNA, and we deploy them under the Omni platform. The intake agent is Omni Voice. The triage and document agents are Omni Ops. They’re not custom builds. They’re configurable products that we adapt to your firm’s workflows, and they’re live in weeks, not quarters.

Why Waiting Makes Sense for Most Firms

The Nvidia announcement is a timing signal. It tells you that the infrastructure layer is mature, the enterprise layer is moving, and the vertical layer is 6 to 12 months behind. If you’re a 50-attorney firm doing $15 million a year, you don’t need to be on the bleeding edge. You need to be on the value edge, where the technology works and the ROI is clear.

Custom builds made sense two years ago when there were no other options. Today, the options are coming. Legal software vendors are already integrating agent capabilities into practice management, document automation, and e-discovery platforms. Clio, MyCase, and Smokeball will all have agent features in the next 12 months. They’ll be cheaper, faster, and better supported than anything you could build in-house.

That doesn’t mean you wait and do nothing. It means you get your workflows documented, your data cleaned up, and your team trained on what AI can and can’t do. When the products ship, you’ll be ready to adopt them fast. The firms that wait without preparing will be just as far behind as the firms that built too early.

The middle path is to deploy agents where the ROI is obvious and the risk is low. Intake is the perfect starting point. It’s high-volume, low-complexity, and easy to measure. You can deploy an intake voice agent in four weeks, measure conversion lift in 60 days, and decide whether to expand from there. If it works, you scale. If it doesn’t, you’re out a month of time and a few thousand dollars, not a six-figure dev contract.

For a practical starting point, download our AI Client Intake Checklist for Law Firms. It’s a worksheet that walks you through the questions you need to answer before deploying any intake automation: what data you need to capture, how to handle conflicts, where the handoff to a human should happen, and how to measure success. It’s not a sales document. It’s a planning tool, and it’ll save you from the most common mistakes we see firms make when they rush into AI.

The Build-vs-Buy Decision for 2026

Here’s the framework I use when a firm asks whether to build or buy. If the workflow is unique to your firm, mission-critical, and a source of competitive advantage, consider building. If the workflow is common across the industry, well-understood, and a cost center, buy.

Intake is a cost center. Every firm does it, and every firm does it roughly the same way. There’s no competitive advantage in having a custom intake agent. The advantage is in answering faster, converting more leads, and freeing up your coordinators to do higher-value work. That’s a buy decision.

Document review is a cost center. First-pass contract review doesn’t differentiate your firm. The quality of your final advice does, but the grunt work of reading and summarising is the same everywhere. That’s a buy decision.

Matter triage is a cost center. Routing leads to the right attorney based on practice area and case value is table stakes. It’s not strategy. That’s a buy decision.

The only time building makes sense is when you have a workflow that’s genuinely unique and you can afford to maintain it long-term. A large firm with an in-house dev team and a seven-figure technology budget might build a custom conflicts engine that integrates with proprietary risk models. A 10-attorney firm should not.

The Nvidia toolkit will make building easier, but it won’t make it cheap. You still need developers who understand both the AI stack and your legal workflows. You still need to manage data pipelines, model drift, and integration updates. The total cost of ownership for a custom agent is $150,000 to $500,000 over three years, depending on complexity. A productised agent from a legal software vendor will cost $5,000 to $20,000 a year and include support, updates, and integrations.

The math is clear. For most firms, waiting 6 to 12 months and buying a mature product is the right move. For firms that are bleeding cash on intake or document review right now, deploying a proven agent today and upgrading later is the right move. The wrong move is spending six figures on a custom build that will be obsolete in 18 months.

What to Do While You Wait

If you’re going to wait for the market to mature, use the time to get ready. Start by documenting your current workflows. Map out your intake process from the first call to the signed engagement letter. Write down every step, every decision point, and every handoff. Do the same for matter triage and document review.

Next, clean up your data. AI agents need clean, structured data to work. If your case management system is full of duplicate records, inconsistent naming conventions, and missing fields, fix that now. You don’t need perfect data, but you need usable data.

Then train your team. Most attorneys and staff have never worked with AI agents. They don’t know what the technology can do, what it can’t do, and how to supervise it. Run a workshop. Show them examples. Let them ask questions. The firms that adopt AI successfully are the ones where the team understands it and trusts it.

Finally, pick one workflow to pilot. Don’t try to automate everything at once. Pick the workflow with the clearest ROI and the lowest risk. Intake is usually the best starting point. Deploy an agent, measure the results, and learn from it. Once you’ve proven the model, you can expand to other workflows.

If you want help with any of that, the AI audit for law firms is built for this exact moment. We spend 60 minutes mapping your workflows, identifying the highest-value automation opportunities, and building a phased implementation plan. You walk out with three things: a process map, a dollar estimate of what you’re losing today, and a prioritised list of which agents to deploy first.

It’s not a sales call. It’s a working session. We’ve done this with 40+ law firms in the last year, and the pattern is consistent. Firms that start with intake see 20% to 35% conversion lift in the first 90 days. Firms that add document review next save 4 to 8 hours per attorney per week. Firms that deploy both recover $80,000 to $150,000 in the first year.

If you want the playbook other teams are using with Claude and Codex right now, grab the free Working With Claude field guide. Download it here.

The Next 12 Months in Legal AI

Nvidia’s toolkit is a forcing function. It will accelerate the pace of agent development across every industry, including legal. The firms that position themselves now will have a 12-month head start when the market floods with new products. The firms that ignore it will be playing catch-up in 2027.

You don’t need to build custom agents to win. You need to understand your workflows, know where you’re leaking money, and be ready to adopt the right tools when they ship. If you can’t wait, deploy proven agents in high-ROI areas like intake and document review. If you can wait, use the time to prepare.

The worst move is to do nothing and assume the technology will pass you by. It won’t. AI agents are coming to legal whether you’re ready or not. The only question is whether you’ll be in the first wave of adopters or the last.

If you want to be in the first wave, start with the Omni Audit for law firms. Sixty minutes, three outputs, no deck. We’ll show you exactly where your firm is leaking money and which agents will close the gap. That’s the starting point. Everything else follows from there.