The Client Is Learning to Use AI. Is the Law Firm Ready?

As businesses begin to perform legal work they once outsourced, artificial intelligence is challenging some of the assumptions on which law firms have built their business models.

Earlier this year, AT&T began an experiment that should command the attention of law firm leaders everywhere. The American telecommunications giant established LegalEdge, a 38-person internal legal unit, and began bringing work traditionally performed by external lawyers back inside the company. Within six months, the unit had handled almost 100 matters. More significantly, AT&T was seeing a reduction in the junior associate hours billed by its external law firms.

On 8 October, the company announced a collaboration with OpenAI to develop AI-powered legal workflows and make better use of its institutional knowledge. Its general counsel, David McAtee, was pursuing something considerably more ambitious than making individual lawyers more productive. He was rethinking how legal work should be organised, which activities required external expertise and what his company should continue paying law firms to do.

For the legal profession, the significance of this development is easily underestimated. Much of the discussion about AI has concentrated on adoption within law firms: which products to purchase, how lawyers should be trained and how existing processes might be accelerated. But the most consequential changes may come from the other side of the table. While law firms are deciding how to use AI, their clients are beginning to reconsider what they need law firms for.

The economics of legal expertise

For generations, the economics of legal practice have rested partly on the cost of acquiring and processing specialist information. Research, document review, drafting and due diligence require knowledge, experience and considerable human effort. Law firms organise that effort through teams of lawyers, supervise the results and charge clients, frequently by reference to the time involved.

AI is beginning to disrupt this arrangement. A corporate legal department with access to its own contracts, negotiating histories, internal policies and commercial records can now use increasingly capable systems to undertake portions of work previously sent outside the organisation. The technology remains imperfect, and professional scrutiny is essential, but the cost of producing a useful preliminary analysis is falling.

The implications for the billable hour are obvious. If a task previously required ten hours and can now be completed to an acceptable standard in two, the client may reasonably question the price. More fundamentally, if the client can perform much of that work internally, the question of how the external firm charges may become irrelevant. The instruction itself may disappear.

There is an additional complication. Clients possess something that even their most trusted external advisers cannot fully replicate: an intimate knowledge of their own business. Decades of contracts, correspondence, disputes and commercial decisions contain information that has historically been difficult to access systematically. AI offers the possibility of turning those fragmented records into a useful institutional resource.

A sophisticated client may therefore arrive at its law firm with the relevant documents assembled, a preliminary analysis completed and a defined question requiring specialist judgment. The lawyer enters the process later, at a point where the preparatory work has already been undertaken. The value of the instruction may remain substantial, but its nature has changed.

Richard Susskind has spent decades questioning the assumption that legal services must continue to be delivered through the traditional structures of professional practice. AT&T's LegalEdge initiative offers a practical illustration of his central argument: clients need solutions to legal problems, and the methods used to produce those solutions are open to reconsideration.

A different kind of legal buyer

The emerging pressure is not simply technological. It is commercial.

Thomson Reuters Institute's 2026 State of the Corporate Law Department report found a striking difference between how general counsel and their senior executives perceived the value of the legal function. While 86 per cent of general counsel regarded their departments as significant contributors to business success, only 17 per cent of C-suite executives shared that assessment.

Corporate legal departments are themselves under pressure to demonstrate value, control expenditure and explain their contribution to business performance. That pressure inevitably passes through to the external firms they instruct.

A managing partner may regard the introduction of sophisticated AI tools as evidence of competitive advantage. A general counsel using similar technology internally may regard it as a basic expectation. Clients will increasingly want to understand why particular work requires substantial professional time, what expertise the firm is contributing and whether the result could be achieved more efficiently.

This presents a particular challenge to the traditional law firm structure. Much of the work most susceptible to automation is also the work historically undertaken by junior lawyers. Research, first-pass document review and routine drafting generate revenue while providing the experience from which professional judgment develops. If clients become less willing to pay for these activities, firms will need to reconsider both their pricing and their approach to professional training.

There is little evidence that demand for external legal services is about to collapse. The same Thomson Reuters research found that more general counsel expected to increase external legal spending than reduce it. Complex disputes, regulatory obligations and sophisticated transactions continue to require specialist expertise. AI itself is generating new categories of legal work.

But growing demand for legal services does not guarantee that the existing methods of delivering them will remain commercially viable.

The implications for Irish firms

It would be easy to dismiss AT&T as an exceptional example from a corporate market far removed from the realities of Irish legal practice. Most Irish solicitors' firms do not serve clients with the resources to establish their own AI-enabled legal departments. Their relationships are often built on years of trust, local knowledge and personal service.

Yet the underlying technology is available to businesses of almost every size. An Irish employer can generate a preliminary employment policy, a company director can review a supplier agreement and an entrepreneur can produce a first draft of a shareholders' agreement without ever contacting a solicitor.

Whether the resulting work is legally sound is another matter entirely. AI systems can misinterpret legislation, overlook important facts and produce convincing but unreliable analysis. A document that appears complete may require substantial professional intervention before it can safely be relied upon.

This creates a new commercial problem. A client who arrives with an AI-generated agreement may expect to pay only for a quick review. The solicitor may discover that establishing the document's reliability requires more work than preparing it from scratch. The client sees a finished document; the lawyer sees unresolved legal risk.

The profession will need clearer ways of explaining and pricing the value of verification, specialist judgment and professional responsibility. There may also be opportunities to provide ongoing oversight of legal processes that clients increasingly manage themselves.

The shift is unlikely to affect every practice area equally. Litigation, conveyancing, probate and complex advisory work involve procedural requirements, professional obligations and judgments that cannot readily be automated. But no firm should assume that its clients' expectations will remain unchanged simply because the work is legally complex.

The value of responsibility

There is an important limit to the idea that clients can simply replace external legal services with AI. Producing legal information is not the same as providing reliable legal advice, and the ability to generate a document does not establish that it is accurate, appropriate or legally effective.

For organisations bringing legal work in-house, questions of governance and accountability become unavoidable. Who determines whether an AI system is suitable for a particular task? How are its outputs checked? What information can be processed? When must a qualified lawyer intervene, and who assumes responsibility if something goes wrong?

AT&T's own announcement emphasises human oversight, professional judgment and governance as essential components of its approach. These are not incidental safeguards. They are part of the infrastructure required to make AI-supported legal work dependable.

For law firms, that points towards a potentially different relationship with clients. External advisers may increasingly be engaged to supervise, validate and provide specialist intervention within legal processes operated by the client. The value of professional expertise may become more closely associated with judgment and accountability than with the volume of documents produced.

This will require changes in how firms organise their work, develop their lawyers and charge for their services. It may also create opportunities for practices capable of combining deep legal expertise with a sophisticated understanding of technology and business processes.

Who will define the future of legal services?

The legal profession has been remarkably successful at absorbing successive generations of technology without fundamentally changing its business model. Electronic research, document management and automation have transformed daily practice, while the underlying structures of private legal services have remained recognisable.

Artificial intelligence may prove different because it is becoming available to clients at the same time as their advisers. The technology is changing the economics of producing legal work, but it is also changing who can undertake that work in the first place.

AT&T is an early example, not proof of an inevitable transformation. Its experiment will need to demonstrate sustained commercial value, and many categories of legal work will continue to require substantial external expertise. Nevertheless, the question it is asking is one that law firms should be considering themselves: which parts of the legal process will clients continue to purchase, and where will they expect their advisers to contribute something they cannot obtain elsewhere?

For managing partners, that is a business strategy question of considerable importance. Investment in AI tools and training is necessary, but it should be accompanied by a serious examination of the firm's services, pricing, knowledge management and relationship with clients.

The legal profession has spent the past three years asking what artificial intelligence will allow lawyers to do.

Its clients are beginning to ask what they will no longer need lawyers to do.

Sources and further reading

• Reuters | 8 October 2026 — How AT&T is using AI to reduce its reliance on law firms
• AT&T | 8 October 2026 — AT&T Launches Flagship Collaboration with OpenAI to Transform AI-Powered Legal Services
• Thomson Reuters Institute | March 2026 — 2026 State of the Corporate Law Department Report
• Richard Susskind | 2023 — Tomorrow's Lawyers: An Introduction to Your Future. Third edition, Oxford University Press.

About the author

Caroline Malone BL is a practising barrister and co-founder of Praxis Consulting, advising law firms on artificial intelligence strategy, governance and legal workflow transformation.