How to Outcompete Your Client's AI

Andy Carter/Ikon Images The in-house lawyers at real estate investment firm Alturas Capital Partners used to rely on outside counsel for much of its lease work. Thanks to generative AI, it can now do that work internally , saving the firm hundreds of thousands of dollars in spending and compressing lease work that once stalled for weeks. But for the outside counsel that lost the work, the bigger problem is existential: AI has eliminated the need for their services. This is redefining the boundaries of firms, since some firms are expanding the scope of their work, while others are losing it to their former customers. This effect is also affecting the prices that the remaining providers can justify. Just recently, Wall Street banks pushed big law firms to cut fees because of AI. This scenario is widespread: Generative AI has lowered the cost of producing legal documents, market analyses, creative assets, and software in situations where a capable in-house team equipped with AI can credibly replicate what an outside provider had been supplying. Service providers must find ways to ensure that there is still a need for their work. Services firms have traditionally built their value propositions around the specialized expertise and experience of their staff, as well as their proprietary methods. Generative AI has changed that by providing access to a fair degree of knowledge that was once the sole province of human experts. This is not the first time technology has changed the economics of what companies handle in-house and what work they outsource. In the 1990s, technological advancements in communications and software enabled the expansion of services as it became feasible — and cheaper — to outsource back-office work. Calculating Customer Costs Service providers must rethink existing sales strategies that are based on expertise alone, now that generative AI is changing the economics of the decision to outsource higher-level knowledge work and making it cheaper to bring it in-house. They need to focus on how to win based on cost as well. Here are three ways they can do so. Improve unit economics. Services firms’ first move should target production cost. The instinct for providers under threat is to defend their expertise — to argue that their people produce better work than a client’s in-house team armed with AI tools. That defense is collapsing. The durable advantage lies not in the skill behind each output but in the economics of producing thousands of them. Providers that build the infrastructure to produce work at scale — automated pipelines, open models, costs spread across many clients — can price each finished output below what any single client could match by doing the work in-house. Cheap production pulled the work inside; cheaper production can pull it back out. One of the world’s largest marketing services groups, WPP, offers an example of how a traditional services firm can win this way. It built WPP Open, an agentic marketing platform that draws on decades of proprietary intelligence, including 30 years of data from the world’s longest-running brand equity study, and behavioral science frameworks from its agency Ogilvy. Coca-Cola is among the brands using it. WPP has also built a self-serve offering through which marketers build strategies, generate assets, and activate media campaigns. While a competitor could acquire the same technology, it would still lack the accumulated expertise that powers the agents — expertise that takes decades to build. This requires a fundamental shift in how professional services firms think about their business, to service productization rather than bespoke delivery. Generative AI has made that shift both more urgent and more achievable. Providers that have already begun the journey — by standardizing workflows, training models on proprietary data, and packaging expertise into repeatable systems — are best placed to win on unit economics. As one analysis of professional services firms argued, the service providers that thrive will be those that shift from delivering their expertise through people alone to delivering it through people and systems together. Make buying as easy as asking. The second move targets a different cost: the effort of buying itself. Hiring a provider has never been free of effort. The customer has to find the right expert, negotiate terms, explain its needs, review drafts, and integrate the result into its own systems. Every hour spent on that strengthens the argument for doing the work in-house. Agentic AI lets providers eliminate that friction, with agents that take the request, do the work, and deliver the answer directly into the tools the customer is already using. Moody’s shows this advantage at work. The credit intelligence firm watched its own customers pick up generative AI and saw that they would soon be able to answer their own credit questions. Its response was to build AI agents that run analyses automatically and deliver the results within the Microsoft applications customers already use, such as Excel. A portfolio manager checking a counterparty’s credit risk now gets Moody’s ratings, data, and research right in the spreadsheet via Microsoft Copilot, the AI tool they would have used to do Moody’s work themselves. The answer arrives in moments. Law firms are taking the same path. A&O Shearman, for example, has built AI agents — developed with legal AI firm Harvey — that distill the reasoning of its senior lawyers for tasks like antitrust filing analysis and reviewing loan documentation, and makes them available to clients and other firms by subscription. When buying from a provider becomes as easy as asking, building an alternative in-house stops looking like a worthwhile investment. Own the operational burden. The third move targets the cost that clients encounter last: quality assurance and maintenance. Bringing work in-house with AI looks easy at first, but someone has to check every output before it can be trusted, fix mistakes, rewrite prompts and workflows as models change, and clear outputs through compliance. The costs accumulate quietly — in staff hours, in rework, in constant maintenance — until they rival the fee the client was paying the provider in the first place. The service provider’s move is to make that burden visible to the customer and be able to carry it for them. Thomson Reuters illustrates how a traditional knowledge provider can win on this ground. The company’s CoCounsel Legal is an AI research product built on its Westlaw legal platform, which contains decades of curated case law, 35 million legal classifications, and the ongoing work of more than 650 attorney-editors. A firm building its own AI legal research tool would have neither the content infrastructure nor the verifiable citations and audit trails that courts and compliance functions require. What Service Providers Under Threat Can Do Now Service providers survive when they win the client’s full make-or-buy comparison: cheaper to run, easier to buy, and less painful to manage. Here’s how to get started on each of the three moves. List which of your deliverables generative AI can already produce just as well as a skilled person, because those are the ones clients will pull in-house first. For each, compare what one output costs you, all in, against what the client would pay in tools, tokens, and staff time to produce it internally. Wherever your numbers win, package the system and sell it. Count the steps and the days between a customer asking you for something and getting a usable answer. Every step is a reason to build instead. Cut those steps to deliver your expertise as close to the customer’s use case as possible. Agentic AI can help here. For every customer considering building in-house, write out what running the work would actually cost their business: the hours checking outputs, the rework, the model updates, the compliance reviews, and the salaries behind it all. Put that total ne