Professional services firms have spent the past several years examining what artificial intelligence will do to their own operating models. The questions are becoming familiar. How much professional work can be automated? How many hours can be removed from an engagement? What happens to utilisation when professionals become more productive? How should firms price work requiring less human effort? What happens to the traditional talent pyramid when junior professionals perform less of the activity through which previous generations accumulated experience and judgment?
These questions focus on how professional work is produced. They examine only one side of the economic relationship. The technologies being deployed inside law firms, accounting firms, consultancies and other professional services organisations are also being deployed inside their clients. HSBC provides a useful example. In January 2026, the bank announced that it was introducing Harvey into its global legal function, describing the objective not simply as greater productivity but as reimagining how an in-house legal function operates by combining AI with the expertise and judgment of its own lawyers. Harvey said two months later that it was already working with more than 500 in-house legal teams worldwide. (HSBC Announces Harvey AI for Its Legal Platform; Announcing Harvey’s Inaugural In-House Customer Advisory Board)
The implication is not that internal legal, tax, finance, technology and strategy functions will suddenly replace the professional services firms surrounding them. Much less needs to change for the economics to move. If a capable internal team can resolve more questions itself, perform more analysis before seeking external support and absorb work that previously required outside capacity, the division of work between client and adviser begins to shift. The relevant AI question is therefore no longer only how professional services firms can deliver existing services more efficiently. It is why clients will continue buying those services externally once they possess many of the same technological capabilities themselves.
The answer is more complicated than a simple make-or-buy decision because clients never bought professional services for only one reason.
The Boundary Between Client and Adviser
The decision to perform an activity internally or obtain it externally is much older than the current AI debate. Ronald Coase’s work on the nature of the firm established transaction costs as one explanation for why some activities are coordinated within organisations while others are obtained through markets. Oliver Williamson later developed this into a broader theory of economic governance. Whether a capability sits inside the client or outside it depends partly on the relative economics of building, maintaining, coordinating and contracting for that capability. (The Institutional Structure of Production; The Economics of Organization: The Transaction Cost Approach)
Professional services have always occupied an interesting position inside that boundary. A multinational can maintain a sophisticated tax function without employing leading specialists in every jurisdiction and every unusual tax problem it might encounter. A corporate legal department can possess considerable internal capability while still having little economic reason to reproduce the specialist depth of a major law firm. A management team may understand its own industry better than any consultant while experiencing only a small number of major restructurings, strategic discontinuities or transformations during the careers of the people involved. Professional services firms solve these problems by concentrating capabilities and spreading the cost of maintaining them across many clients.
AI changes part of that calculation because it reduces the cost of accessing and applying some forms of knowledge internally. The likely result is not a clean movement from external to internal delivery. Professional services firms have access to the same underlying technology and can combine it with proprietary knowledge, specialist communities, workflows, data and accumulated experience. More likely is a continuing reallocation of work across a client-adviser system in which both sides become more capable. This is the demand-side consequence of the argument developed in When Expertise Becomes Metered Infrastructure. Expertise becomes cheaper to produce inside the firm, but some of the same expertise also becomes cheaper for the client to access without buying the historical delivery model surrounding it.
Clients Will Move at Different Speeds
Clients will not develop these capabilities at the same speed. A large multinational can invest in enterprise AI environments, data infrastructure, specialist talent, governance and the redesign of internal legal, tax, finance and strategy functions. It can spread those investments across large populations and substantial volumes of recurring work. A mid-market organisation can access many of the same underlying models, but often has less economic reason and less organisational capacity to build the infrastructure around them.
The adoption data already shows a substantial size effect. Across OECD countries for which data was available, 52.0 percent of large firms reported using AI in 2025, compared with 17.4 percent of small firms. Eurostat shows a similar pattern across the European Union: 55.0 percent of large enterprises used AI technologies in 2025, compared with 30.4 percent of medium-sized enterprises and 17.0 percent of small enterprises. Eurostat points to implementation complexity, economies of scale, affordability of investment and access to expertise as possible explanations. Access to powerful models may be becoming widely available. The organisational capacity to embed them is not. (OECD – AI Use by Individuals Surges Across the OECD as Adoption by Firms Continues to Expand; Eurostat – The Use of Artificial Intelligence Technologies in the European Union)
The same service can therefore face very different internalisation pressure depending on the clients buying it. A large corporate tax function with dedicated technology teams, substantial data infrastructure and thousands of recurring transactions may have a strong economic case for developing AI-enabled capability internally. A mid-market organisation buying the same expertise occasionally may not. Smaller organisations can sometimes move faster because they have fewer legacy systems and simpler governance, but scale still affects how much investment can be justified and how much specialist capacity can be maintained internally.
Client segment therefore becomes part of AI exposure. The question is not only what service is being sold. It is who is buying it, what capability that client already possesses, what it can economically build and how quickly that capability is likely to develop.
Expertise and Execution Move First
Technical expertise appears particularly exposed because access to specialised knowledge has historically been one of the clearest reasons for buying professional services. A client facing an unfamiliar tax question, legal issue or regulatory requirement could either invest heavily in developing the capability internally or obtain it from an institution whose professionals dealt with similar questions continuously. AI does not eliminate the difference between a technically capable internal professional and a genuine specialist, but it expands the range of questions the internal professional can address before external support becomes necessary.
Thomson Reuters’ 2026 research into corporate tax departments illustrates this direction. Sixty-nine percent of respondents identified further automation of routine compliance processes as a high-value AI opportunity, while 61 percent identified accelerated technical research and analysis. The first reduces the amount of execution required. The second increases what the internal professional can do. Corporate strategy follows similar logic. BCG estimates that more than 80 percent of activities commonly performed by corporate strategy functions have medium or high exposure to AI automation or augmentation, including information collection, analysis, option generation and decision support. AI does not need to reproduce an entire strategy consultancy to affect consulting demand. It only needs to increase how far a relatively small internal strategy team can progress before external support becomes economically justified. (2026 Corporate Tax Technology Report; The Corporate Strategy Function in an AI-First World)
Execution capacity faces a related pressure. Professional services firms have historically been valuable not only because they possessed knowledge the client lacked, but because they could mobilise large numbers of people when temporary demand exceeded permanent internal capacity. Transactions, transformations, litigation, investigations, regulatory programmes and technology implementations routinely require workforces that few organisations would sensibly maintain between such events. AI makes both sides of that relationship more productive. The consulting firm can deliver more with fewer people, but the client’s own team can also absorb work that previously required external capacity.
Recent Financial Times reporting describes companies bringing more technology work inside as AI improves their internal capability, while clients increasingly question historical consulting team sizes. For businesses built heavily around leverage, the implication is uncomfortable. Becoming twice as productive does not automatically improve the economics if the client responds by purchasing substantially less external production. (Consultants Head for an AI Showdown With Their Own Clients)
What Is Harder to Bring Inside
Specialist capability follows different economics. A client purchases execution capacity because there is too much work for the internal team. It purchases specialist capability because maintaining that capability internally would make little sense regardless of how productive its existing people become. A company may encounter a particular tax controversy once in a decade. A board may undertake a major restructuring once during the life of the institution. A transaction may require specialist valuation knowledge with little relevance to normal operations. A sophisticated cybersecurity organisation may still need forensic capability following an unusual incident. In each case, external specialists remain economically attractive because the client does not have to maintain the capability between the relatively small number of situations in which it is required. (Knowledge Transfer Across Permeable Boundaries: An Empirical Study of Clients’ Decisions to Involve Management Consultants)
AI can nevertheless change what counts as specialist. Knowledge that is expensive and scarce today may become codified, embedded in tools and accessible to capable generalists. A specialist whose value depends mainly on possessing technical information may therefore experience the same erosion as other expertise businesses. More defensible are capabilities whose scarcity reflects unusual experience, application under uncertainty or sufficiently low-frequency client demand that maintaining them internally remains uneconomic. A highly experienced specialist supported by AI may perform more analysis personally, depend on fewer junior resources and apply accumulated experience across more situations. Clients may buy less external labour while continuing to pay for capabilities that remain genuinely scarce.
Judgment sits close to this boundary. In Manufacturing Judgment, I distinguished expertise from judgment. Expertise provides possible answers. Judgment determines which answer should be trusted, how it should be applied and what decision should ultimately be made when evidence is incomplete or the available options contain real trade-offs.
AI can make expertise much more accessible without necessarily resolving those conditions. But it would be premature to conclude that this automatically creates a judgment premium. Models will become better at comparing scenarios, exposing inconsistencies and retrieving analogous situations, while clients may value senior judgment and simultaneously expect the technology surrounding it to reduce the total cost of obtaining it. Thomson Reuters found that only 15 percent of in-house legal professionals expect outside-firm fees to maintain or increase as AI use grows. Where technically plausible answers still do not resolve an important decision, automation of the underlying analysis does not necessarily remove the reason for external involvement. That does not tell us what the client will pay for it. (Future of Professionals 2026 Legal Report)
Experience Across Clients
Professional services firms possess another potential advantage because they observe comparable situations across multiple clients. A management team experiences its own transformation. An adviser may see twenty. A CEO completes a limited number of major transactions during a career. A transaction professional may work on hundreds. A corporate tax director may redesign an international operating model once, while an adviser encounters variations of similar problems across industries, jurisdictions and ownership structures. Research into sector-specialist consultants identifies this repeated exposure as a genuine source of expertise because the adviser accumulates variations of the same underlying problem rather than only more information about it. (Configuring Expert Knowledge: The Consultant as Sector Specialist)
AI will narrow the public-information component of that advantage substantially. Clients can interrogate regulatory decisions, case studies, transaction announcements, academic research and industry reporting themselves. The harder advantage to reproduce lies in what does not enter the public record: why an apparently sensible governance model failed once implemented, which stakeholder blocked a decision, where incentives undermined execution or why two organisations produced different outcomes from superficially similar strategies. Repeated exposure can therefore remain valuable even when information becomes abundant because experience creates pattern recognition across situations rather than simply access to more facts.
But repeated client exposure does not automatically become institutional pattern recognition. Research into knowledge-intensive services shows that knowledge developed through customised client work can remain client-specific or disappear unless firms deliberately retain, diffuse and reuse it. A firm whose learning remains inside individual partners may discover that much of what it called institutional experience was never institutional at all. The advantage comes from converting experience into reusable judgment that the institution can mobilise when the next client encounters a similar problem. (Does Customization Promote Innovation? Evidence From Knowledge-Intensive Business Service Firms)
When Being External Is Part of the Product
The boundary looks different again when part of the value comes from the provider sitting outside the client organisation. Audit provides the clearest example. A company may understand its own systems better than its auditor, employ outstanding accountants and possess sophisticated analytical technology. It still cannot provide independent assurance over its own financial statements. Independence is not simply an attribute surrounding the service. It is part of the institutional reason the service exists. (AS 1000: General Responsibilities of the Auditor in Conducting an Audit)
The same logic appears elsewhere in less formal forms. Boards commission external reviews. Investors use independent diligence. Organisations bring in outside investigators when an internal review would lack sufficient credibility. Leadership teams sometimes want their own analysis challenged by someone outside the existing hierarchy and incentive structure. The client may already possess substantial expertise. What it cannot create internally is the same distance from the activity being assessed. Management consulting research has long identified these symbolic and legitimating functions alongside technical expertise, but also their weakness. External validation creates value only where the adviser is regarded as sufficiently independent to challenge the client rather than merely endorse a predetermined conclusion. (Symbolic Functions of Consultants)
Accountability sits beside this distinction. As argued in Trust Capital, professional services firms create institutions around judgment so that clients and other stakeholders become willing to rely on it. If a corporate team can produce sophisticated analysis internally at very low marginal cost, part of the remaining value of an external opinion can lie in the process surrounding the answer: experienced professionals assess the facts, quality procedures are applied, internal review takes place and a professional institution places its reputation behind the conclusion.
None of this means the economics are protected. Independence can preserve a reason for external demand while pricing pressure, investment requirements, regulation and liability make the resulting business unattractive. Something can remain structurally external without being a good business.
The Boundary Is Moving, Not Disappearing
Some of the most consequential professional services engagements expose another reason clients use outside advisers. Major restructurings, transactions, cyber incidents and difficult transformations involve multiple specialist domains, stakeholders and decisions whose interactions become more important than any individual technical problem. In these situations, part of the external value can sit in integrating contributions, connecting information and helping an organisation maintain coherence while the situation is moving quickly.
This capability is not inherently external. Sophisticated clients already perform much of the orchestration themselves, and research on the active client shows how internal managers can act as gatekeepers, brokers and partners in managing outside advisers. AI and agentic systems will also become better at workflow coordination, information synthesis and dependency tracking. The question is whether a temporary, consequential and fragmented problem creates enough cross-boundary complexity that maintaining the necessary capability internally would make little economic sense. (The Active Client: Gatekeepers, Brokers and Partners of External Consultants; Inter-Team Knowledge Collaboration Across Boundaries in Multi-Team Consulting Projects)
The commercial evidence already suggests that the boundary will not move in one direction. Legal clients are pushing firms to reflect AI productivity in pricing, with Thomson Reuters reporting that 71 percent of in-house legal professionals expect outside firms to change commercial models as AI usage increases. Wall Street banks have gone further by asking law firms to quantify AI-related efficiency gains and exploring more competitive bidding and alternative-fee arrangements. At the same time, corporate legal departments continue to expect significant external demand, and AI itself is generating new consulting work. Source Global Research reported in March 2026 that 88 percent of surveyed consulting clients had paid for external AI support during the preceding year. (Wall Street Banks Push Big Law to Cut Fees Because of AI; 2026 State of the Corporate Law Department Report; Understanding the Market for AI-Related Consulting Services)
Internalisation and external spending can therefore increase at the same time. Some existing work moves inside the client. Some requires fewer external people. Some specialist capability remains outside. New categories of external demand appear. These developments only look contradictory when professional services demand is treated as one thing.
It is not.
A single engagement can contain technical expertise, additional labour, rare specialist capability, judgment, independence, validation, accountability and orchestration. Historically, those elements were bundled together inside a delivery model whose economics remained heavily connected to professional time. AI begins to separate them. Understanding demand therefore requires more than estimating how much of the underlying activity can be automated.
The Client Boundary Test
Much of the current discussion estimates AI exposure by identifying which professional tasks can be automated or augmented. That is useful for understanding production economics. A law firm needs to know how much contract review can be automated. An audit firm needs to understand which testing activities require less human effort. A consultancy needs to know how much research and analysis its professionals can perform with AI.
But task automation says surprisingly little about whether the client will continue buying the service.
A service can be highly automatable while remaining structurally external. Statutory audit is an obvious example. Technology may fundamentally change how the work is produced without removing the requirement for independent assurance. The opposite is equally possible. A service can continue to contain substantial human judgment while becoming easier for a sophisticated client to internalise because its own professionals become capable enough to perform it.
Production exposure and demand exposure are different questions.
The Client Boundary Test starts from the second. It asks why the client buys a capability externally and whether that reason remains defensible as its own technology and talent improve. The categories are not predictions about which services will survive. They are different economic explanations for why the work sits outside the client organisation.
| What the client is buying | Why it is external | Likely AI effect | What must still be true |
|---|---|---|---|
| Expertise | The client does not possess the required knowledge | External advantage weakens materially | Knowledge remains proprietary, highly contextual, difficult to verify or meaningfully superior to common AI capability |
| Execution capacity | The client needs additional labour temporarily | External volume is likely to decline as productivity increases | Demand remains volatile, urgent or large enough that permanent internal capacity remains uneconomic |
| Specialist capability | Capability is too scarce or episodic to maintain internally | More resilient than ordinary expertise, but not protected | Scarcity reflects experience or application rather than information that can simply be codified |
| Scarce judgment | Several plausible answers exist and consequences matter | Relative importance may increase, but a pricing premium is not proven | External judgment materially improves decisions and clients are willing to pay for that difference |
| Cross-client pattern recognition | The adviser has observed comparable situations elsewhere | Public-information advantage declines | The firm converts experience into genuine institutional learning rather than leaving it with individuals |
| Independence | Part of the value requires the provider to sit outside the client | Structurally durable where genuinely required | Externality remains necessary and the service can still be provided at attractive risk-adjusted economics |
| External validation | Leadership wants its analysis or decision challenged | Can remain important as internal capability grows | The adviser is regarded as genuinely independent and capable of challenging the client |
| Accountability | A credible professional institution stands behind the conclusion | May become more visible as answers become cheaper | The institution accepts meaningful professional responsibility and clients value that responsibility |
| Governance and defensibility | External involvement strengthens the credibility of the decision process | Likely to remain relevant for high-consequence decisions | External review improves the process rather than substituting for management responsibility |
| Trusted orchestration | Temporary complexity spans several specialist domains and stakeholders | Potentially resilient, but increasingly contestable | The client cannot economically reproduce the necessary boundary-spanning capability |
The table is deliberately less optimistic than much of the current discussion around judgment and trust. Judgment does not automatically become more profitable because expertise becomes cheaper. Independence preserves external demand but can coexist with unattractive economics. Pattern recognition matters only if the firm genuinely learns from its work. External validation creates value only where the adviser can challenge rather than merely legitimise. Orchestration can increasingly be performed by sophisticated clients and technology.
Two businesses can therefore sit inside the same service line and contain similar amounts of automatable work while having completely different demand exposure. One may depend mainly on expertise and additional labour. Another may depend on specialist scarcity, independence or institutional accountability.
Client capability then affects how quickly that exposure becomes economically visible. A large corporate and a mid-market organisation may both be capable of internalising more work eventually, but the investment case, available talent and speed of adoption can be very different. AI exposure therefore follows the underlying economic portfolio more closely than the service-line label suggests.
And even structural exposure does not translate immediately into lost revenue.
Trust and switching costs create time.
Moving work inside, changing advisers or redesigning a long-standing relationship can require the transfer of historical knowledge, new data and system connections, procurement effort, governance changes and substantial management attention. Relationships themselves matter. Clients may stay with an adviser long after part of the original economic reason for the relationship has weakened.
That resilience should not be confused with long-term defensibility. Switching costs delay change. They do not eliminate the forces causing it.
Demand Defensibility Is Not Economic Attractiveness
The Client Boundary Test should not be read as an invest-or-exit model. It answers one question: how defensible is the reason this capability remains external as client capability improves? That does not determine whether serving the demand remains economically attractive.
A portfolio can have structurally protected external demand and still be an unattractive business. Statutory audit provides the clearest example. Independence can preserve the need for an external provider while pricing pressure, regulatory cost, liability exposure and continuing investment requirements determine whether the economics make sense.
The opposite is also possible. A service whose historical external advantage is highly exposed can remain an excellent business if the firm industrialises it successfully. Recurring expertise can move from human delivery into technology, lower unit cost, expand the addressable market and create economics that are more attractive than the professional labour model being disrupted.
Demand defensibility and economic attractiveness are therefore separate dimensions. The Client Boundary Test establishes whether there remains a credible reason for the client to buy a capability externally. The economic analysis asks whether serving that demand can generate an attractive risk-adjusted return. Only when both are understood does the portfolio decision become meaningful.
This connects directly with the Professional Services Transformation Theory. Client × Service × Team remains the economic unit through which the business can be understood. The client dimension now requires a deeper question: why does this client buy the capability externally, and how likely is that reason to survive as the client becomes more capable?
A global corporate with substantial internal functions may answer that very differently from a mid-market organisation buying the same service. The resulting economics then determine whether the firm should invest behind that demand, redesign the model, industrialise it, harvest it or eventually leave it.
Closing Thoughts
Professional services firms have historically been able to sell several different sources of value through the same delivery model. A tax engagement could contain technical expertise, execution capacity, specialist judgment, external validation and institutional accountability, while most of those elements were still priced through the professional time required to produce the work. Firms did not need to distinguish precisely why the client was buying externally because the different sources of value largely travelled together.
Artificial intelligence begins to separate them. Technical knowledge becomes easier for the client to access. Internal teams can perform more analysis themselves. Work that once required additional professional capacity can be completed with fewer people. At the same time, unusual specialist experience, independent challenge and institutional accountability do not necessarily move inside at the same rate.
The change will also happen unevenly. Some clients can invest heavily in internal capability and move quickly. Others have little economic reason to reproduce externally available capability themselves. That difference may become as important as the service itself in determining where demand changes first.
Some businesses will lose their historical reason for external demand. Others will retain it while facing deteriorating economics. Some may become more valuable as scarcity shifts away from information and toward experience, judgment or institutional capability. Others may become attractive precisely because the underlying work becomes highly automatable.
Task automation tells leadership how the work may change.
It does not tell leadership what clients will still buy.
What This Means for Boards
Boards and executive teams should be cautious about assessing AI exposure primarily through technology adoption, estimates of automatable activity or broad service-line labels. Those measures describe how the work can be produced. They do not explain why clients buy it externally.
A better starting point is to examine each economically significant portfolio and ask what the client is actually buying from the firm. Is it knowledge the client does not possess? Temporary labour? Rare specialist capability? Judgment? Repeated experience across other clients? Independence? External validation? Institutional accountability?
Then ask how quickly the client could develop a credible alternative. How sophisticated is its internal function? How much recurring demand does it have? What technology and data infrastructure already exists? Can it justify dedicated investment? Does it have the talent and management capacity to redesign the work?
Finally, separate demand from economics. A business can have protected external demand and still destroy value. Another can face substantial internalisation pressure while becoming more attractive because automation and scale change the production model. Pricing power, recurrence, investment intensity, liability, regulation, scalability, client concentration and competitive position remain part of the decision.
The central strategic question is therefore not whether AI can perform the service.
It is why the client still needs to buy the capability externally, how long that reason is likely to survive, and whether serving the remaining demand is economically attractive.
Continue Exploring
The Bigger Picture
The Professional Services Transformation Theory
Why changing economics, technology and operating models are increasingly forcing professional services firms to rethink how the institution itself is designed.
Follow the Argument Back
When Expertise Becomes Metered Infrastructure: How AI Is Changing the Economics of Professional Services
How AI is making expertise more accessible and changing the economics that historically kept professional capability inside professional services firms.
Explore What Remains Scarce
Manufacturing Judgment: How Professional Services Build Their Most Valuable Asset
Why professional judgment develops differently from technical expertise, and why AI may make that distinction increasingly important.
Explore Why Clients Still Rely on Firms
Trust Capital: Why Professional Services Manufacture Judgment but Accumulate Trust
Why expertise and judgment only create economic value when clients and other stakeholders are willing to rely on the institution behind them.
Apply the Thinking to Your Firm
AI Economics Review
An independent review of what AI is actually changing in your firm’s economics, including client demand, delivery economics, pricing, value capture, investment requirements and the wider operating model.
Sources
Current Client and Market Evidence
- HSBC Announces Harvey AI for Its Legal Platform
- Announcing Harvey’s Inaugural In-House Customer Advisory Board
- The Corporate Strategy Function in an AI-First World
- 2026 Corporate Tax Technology Report
- Future of Professionals 2026 Legal Report
- 2026 State of the Corporate Law Department Report
- Consultants Head for an AI Showdown With Their Own Clients
- Wall Street Banks Push Big Law to Cut Fees Because of AI
- Understanding the Market for AI-Related Consulting Services
- AS 1000: General Responsibilities of the Auditor in Conducting an Audit
- OECD – AI Use by Individuals Surges Across the OECD as Adoption by Firms Continues to Expand
- Eurostat – The Use of Artificial Intelligence Technologies in the European Union
Academic and Economic Foundations
- The Institutional Structure of Production
- The Economics of Organization: The Transaction Cost Approach
- Knowledge Transfer Across Permeable Boundaries: An Empirical Study of Clients’ Decisions to Involve Management Consultants
- Configuring Expert Knowledge: The Consultant as Sector Specialist
- Does Customization Promote Innovation? Evidence From Knowledge-Intensive Business Service Firms
- Symbolic Functions of Consultants
- The Active Client: Gatekeepers, Brokers and Partners of External Consultants
- Inter-Team Knowledge Collaboration Across Boundaries in Multi-Team Consulting Projects
Further Reading
- When Expertise Becomes Metered Infrastructure: How AI Is Changing the Economics of Professional Services
- Manufacturing Judgment: How Professional Services Build Their Most Valuable Asset
- Trust Capital: Why Professional Services Manufacture Judgment but Accumulate Trust
- The Professional Services AI Paradox: How the AI Platform Economy Is Colliding with the Partnership Model
- The Professional Services Transformation Theory