AI Tool Sprawl is Real and Likely Hurting Your Firm
We’ve Got a Rubric to Help You Curb It
Executive Summary
AI tool sprawl—the uncontrolled spread of too many overlapping or unmanaged AI tools—is draining law firms’ money, time, and security. Many firms now rely on dozens of AI programs without proper oversight, training, or proof of value, leading to hidden use, compliance issues, and inefficiency. The solution is strong AI governance supported by a clear, firmwide AI policy that defines how tools are evaluated, implemented, and monitored. By simplifying their tech stacks, measuring results, and enforcing consistent policies, law firms can cut waste, lower risks, and make sure AI truly improves client service.
There is a monster lurking in the shadows of your law firm. It is hiding in plain sight. The bad news is that it is a monster of your own making. The good news is that it is completely within your power to banish it. Its name? AI tool sprawl.
AI Tool Sprawl
AI tool sprawl is the unchecked proliferation of AI applications, models, and integrations across an organization without centralized governance or standards. Tool sprawl is tools that are available but go unused. It’s tools that do very similar, or the same things. It’s tools that don’t offer a measurable ROI. And, notably, it is shadow use—where employees continue to use free or personal versions of AI tools because they are comfortable with them, the tools are accessible, and they are prevalent, regardless of whether your firm offers other AI tools.
AI tool sprawl is not just about wasted money, although that’s a significant consideration. It brings other problems. It requires resources to adopt, train, and implement the AI tool within your firm. It makes AI governance significantly more difficult, and it increases risks around security, compliance, and reputation. So how common is tool sprawl in legal? The data is clear.
AI Tool Sprawl In Law Firms
Although it exists in most industries currently, legal’s contribution to AI Tool Sprawl is not small. The Legal Tech Hub, an online directory for legal technology that helps buyers, vendors, and investors discover and evaluate tools, services, and events, maps out AI tools for the legal space. As of September 2025, LTH lists 728 GenAI solutions. Of those, there are 102 different AI legal assistant tools and 84 litigation management tools. And those are just the tools that it’s aware of. It added 90 tools from June 2025 to September 2025, a trend almost guaranteed to continue or accelerate.
There is a plethora of AI tools to choose from and more sprouting up each month. And among those tools, there is invariably overlap in the capabilities and functionality of AI tools. According to TechRadar, Gartner estimates that up to 25% of enterprise AI investment is duplicated. Yet no one wants to miss out on the newest, best, most shiny, and promising AI tool, and so we stack functionality, resulting in inevitable tool sprawl.
For many firms, the problem isn’t a lack of AI—it’s a lack of use of the AI tools they have. This is likely due to firms not providing adequate training on those tools, or lawyers trying out the tools and finding they are not useful in practice.
LawPay’s survey of 2,800 firms shows that AI adoption is most common among larger firms. The larger the firm, the more likely that it has adopted and integrated AI. This is borne out by a recent study by Skills.Law, who surveyed more than 100 of the largest law firms in the US (70 responses), the UK (9), Canada (8), and international firms (13). The average law firm had 18 live AI tools, with 6 that they were pilot testing, and 22 additional tools they were considering. It’s worth noting that this group is AI and tech-savvy, with nearly all members having an AI policy and providing AI training to their employees; very different from law firms in general.
Shadow AI Use
Although AI tool adoption is increasing in law practice, shadow AI use is rampant. Shadow AI use refers to the practice of lawyers using personal AI tools for their professional legal work. The UK’s Society for Computers and Law called Shadow use a minefield. It reported in a recent survey of 300 UK corporate legal departments that a startling 81% of respondents surveyed are using unapproved AI tools. Shadow AI use exists everywhere but can be difficult to track since it is often only disclosed voluntarily. There are indications of shadow use, however. For example, 47% of immigration lawyers responding to LawPay’s survey reported using AI in their work, but only 17% of immigration firms had formally adopted AI tools. So either a few immigration firms are heavily using AI, or immigration attorneys are using AI without their firm’s blessing. Anecdotally, in my experience speaking with lawyers and legal professionals, shadow use is far more prevalent than any formal study that I could find.
The Risks of Hasty AI Adoption in Law Firms
The rush to adopt generative AI in law firms has created a paradox because the very tools meant to enhance efficiency can, if not managed carefully, increase ethical risk, billing confusion, and governance challenges. As well, higher administrative costs can cut away the profits promised by the use of AI.
More AI Tools Mean More Need for Training
Volume and costs aren’t the most worrisome issues; the risk of using AI rises when those tools involve client data, privilege, and billing. This involves the need for appropriate training and quality AI Governance.
Within legal, fewer than half of professionals report being trained on how to use AI in their practice. Meanwhile, expectations are racing ahead—95% of respondents believe GenAI will be central to workflows within five years. In short, we’re buying capability faster than we’re building the systems, metrics, and skills to use it well.
Untrained users lead to both overuse and underuse of the AI tools at hand. Overuse is where legal professionals put confidential information into untrustworthy AI tools, such as tools without adequate confidentiality and privacy controls. It is also where legal professionals are overconfident in the output they receive from their AI tool, trusting it without verifying it. This leads to filing legal documents with fake cases, sanctions, loss of reputation, and perhaps malpractice.
Underuse is more a financial consideration than a legal or ethical issue. It is where tools are not being used to the fullest extent of their capabilities. It is what leads to firms to adopt overlapping tools, dissatisfaction with the tools they’ve purchased, and simple unrealized ROI.
More Tools Mean Greater Ethical Risk
The AI “arms race” within law firms stems from FOMO – the fear of missing out on the latest and greatest tools. Speed and over-adoption create redundancy and waste. The more AI tools the more rules to be created and enforced. A single misstep—say, running a client’s data through a platform that violates their privacy—can transform an efficiency gain into a compliance crisis. And this is where governance comes into sharp relief. More than half of legal professionals say their firm does not have an AI policy, or at least they are unaware of a firm AI policy, according to a recent Clio Study. This leads staff to feel more comfortable using free or low-cost AI tools. Free and low-cost AI tools often lack the necessary data protections to meet our confidentiality duties. For example, OpenAI’s own help center documentation states, “If you’re using a consumer product like ChatGPT Free or Plus, your conversations with GPTs may be used to improve OpenAI’s models.” Although it goes on to say that users can opt out of this training, with no AI policy and no policy enforcement, there is no way of knowing if your staff are using tools that leave your firm open to exposure. They might not even understand that ‘used to improve our models’ is a euphemism for ‘scooping up your input and feeding it into our AI tools where you will no longer have any control over it, where it goes, or who sees it.’
AI adoption not only leads to ethical risks but also opens a firm to increased cybersecurity risks and governance headaches.
More Tools Mean More Governance Complexity
Governance and enablement still trail adoption. Thomson Reuters’ 2025 Generative AI in Professional Services Report states that about half of professionals across legal and adjacent fields already use GenAI in some fashion —yet good management and governance lags. From that same study, only 20% of organizations are measuring GenAI ROI (with 59% not measuring at all). And 71% of corporate legal clients say they don’t know whether their outside firms are using GenAI. Each new AI platform a firm adopts introduces a different set of data policies, security risks, and billing considerations. Or at the very least, another AI tool to babysit, integrate, test, support, and manage. Multiply this across half a dozen vendors and the governance burden grows exponentially. What began as an effort to modernize has the potential to become a tangle of overlapping terms of service, inconsistent privacy safeguards, and fragmented oversight.
Without deliberate governance frameworks, the risk is not just that tools go unused—it is that the profession’s core commitments to ethics, transparency, and client trust can be undermined by the very technologies meant to assist lawyers as they try to uphold these values. The fix is not another platform, it’s effective governance, actually put into practice.
More Tools Mean More Technical Complexity
One of the biggest detriments for law firms experiencing AI tool sprawl is the challenge of integration. Tools for billing don’t talk well with tools for research. Both of them need to be integrated with client management databases, and someone needs to babysit all of them.
Lawyers must be trained on each system, compliance teams must monitor each vendor’s data practices, and billing staff must reconcile whether AI costs are properly allocated. Instead of efficiency, the result can be confusion, frustration, and administrative overhead.
The Solution--How to Break the Sprawl
The solution to AI tool sprawl isn’t to just avoid AI tools. AI tools can bring unparalleled efficiency gains. And we are soon approaching the day when it will be malpractice not to use AI in your legal work because of the promise it brings (an article on this is soon forthcoming). The way out of this messy sprawl is through discipline.
Good AI Governance
Good implementation starts with good AI governance. Not as a slogan, but as a set of accountable behaviors tied to recognized standards. If your firm does not have good AI governance, you should rethink your choice to use AI. AI without quality governance can too easily end up being a drag or a liability, rather than a boost. And at the very least, it is wasted capital.
The NIST AI Risk Management Framework gives firms a practical loop—Govern, Map, Measure, Manage. That translates neatly for firms: decide who is responsible, know where and how a tool is used, measure its impact and risk, and adjust continuously.
Or start with Lexara’s simple method: 3D’s -- Decide, Deploy, and Defend – which will take you through the entire life cycle of any AI tool, while curbing tool sprawl.
3D’s -- Decide, Deploy, and Defend
In a market obsessed with speed, this model is the competitive advantage. It converts AI from a hope, a dream, and an aspiration into an actual, useful tool. You’re not adopting a tool based on slick marketing or promises of the latest and greatest features. You’re looking critically at a tool to decide if it meets your needs and fits into your firm, and whether it can provide your firm with an acceptable ROI.
DECIDE whether the tool fits into your current tech stack. Will it talk to the other AI tools you already have, or will this be an additional burden requiring a manual transfer of information from one system to the next in order to bridge a gap? Too many times, I’ve seen or heard from law firms that have adopted an exciting new AI tool, only to realize that it doesn’t offer quite the benefit expected because, one or more times per use, someone must physically copy and paste information from one AI tool to the next. Nothing withers professional pride and office morale faster than a job that consists largely of pressing CTRL-C, CTRL-V.
DECIDE whether the AI tool offers a provable ROI. Can you map to actual profits? Start by asking the tool’s provider if they can show evidence of what return on investment you can expect. Spend some time investigating their reports or materials showing this ROI. Most, if not all, reputable AI tools offer a free trial of at least 14 to 30 days. Put the tool to the test. If the lift to actually put the tool into action is too great, run a sample test with limited data. Then track the benefit. If you are not seeing a decrease in administrative or non-billable time, improved quality, or faster results, reconsider adopting that AI tool.
DECIDE what use-cases the tool covers. You should have one research assistant, one drafting assistant, one e-discovery assistant, one billing assistant, each chosen through controlled pilots or trial periods, after reviewing pre-anticipated metrics (e.g., draft-to-file time, iteration count, error rates).
Currently, task-specific tools remain more effective than end-to-end practice software with AI features. Avoid being enamored of fancy or well-known brands. Make each tool prove its own worth before adoption.
DEPLOY the tool only once you have determined its value, its risks, and its governance requirements. Be sure to add it to the “approved tools” list in your firm’s AI policy. And train everyone on its proper and effective use. Never assume that its use is intuitive or common sense.
DEFEND the tool through ongoing monitoring and good governance. If you’re not familiar or comfortable with governing AI at your firm, or you’re unsure of other issues regarding AI governance, head here to refresh your knowledge. Governance sets the structure for your AI use, and pilot tests prove valuable. The result is that your clients benefit. You must test and prove a tool, and only then do you scale.
DEFEND seat counts. Or rather, make seat counts defend themselves. For example, you are paying for Copilot Enterprise at X number of seats, but how many of those seats are actually being used? Monitor this in order to reclaim seats and thereby find savings.
Evaluate AI tool use regularly. Talk to the end users to help determine whether the tool adds value, and fits into their workflow. Monitor this well, and the conversation changes. Instead of chasing features, latest deployments, and hollow promises, the firm runs a valuable AI tech stack.
Buy, Delay, Skip
Buy where the tool is embedded in the workstream, if possible. Often, however, point solutions are still more effective than entire end-to-end solutions. If you’re buying for a specific solution, make sure you can show measurable ROI on your matters, and check on integrations.
Delay purchasing or renewing licenses for tools unless you can prove a workflow change in quality, risk, or speed that justifies cost and billing.
Skip two tools that do one job. Limit overlap as much as possible.
Bottom Line
In legal, we’re not just overspending on AI tools; we’re over-licensing and under-using and then asking clients to trust that the sprawl makes their matters better. The fix is to prove the ROI before or at the earliest stages of adoption. Implement good AI governance. Prove value at the workflow level, collapse overlaps, and let monthly seat counts be justified by actual use. That stops waste, reduces risk, and—importantly—keeps clients at the heart of the decisions.
© 2025 Amy Swaner. All Rights Reserved. May use with attribution and link to article.
You have really put your finger on a problem I see in so many law practices. I think it often boils down to the fact that we as lawyers operate so much on instinct and habit - and we're so focused on the actual client work - that stopping to think about these nuts and bolts issues ends up in the 'too hard' basket. We really have to think about what we need these tools for rather than just adopting the next shiny thing.
The lawsuits are catching up as burned clients realize what actually happened. I use A.I. as an assistant, for a rough draft, but I check every citation, rule of court and even parties' orientation gets hallucinated into a parallel universe. So great for rough work, but the polish has to be human. In my experience.