What happens when expertise itself starts behaving like infrastructure?
The current AI debate is dominated by questions about replacement. Will consultants disappear? Will lawyers disappear? Will accountants disappear? Will software developers disappear? Depending on who is speaking, artificial intelligence will either eliminate large parts of knowledge work or leave the fundamentals largely unchanged. The discussion is fascinating, but it may also be distracting. While the rest of us debate jobs, prompts, agents, and productivity, some of the largest companies in the world are deploying capital at a scale rarely seen outside major infrastructure projects.
Microsoft continues to expand AI data-center capacity. Alphabet has significantly increased infrastructure spending. Amazon is investing heavily in AI infrastructure. OpenAI and its partners are discussing projects measured not in millions or billions, but in hundreds of billions of dollars. At first glance this looks like a technology story. Look more closely, however, and it begins to resemble something else. Historically, organizations do not commit capital at this scale because they believe they are building a feature. They do it because they believe they are building infrastructure. Railroads required infrastructure. Electricity required infrastructure. Telecommunications required infrastructure. Cloud computing required infrastructure. The interesting question is therefore not whether AI becomes powerful. The interesting question is what the builders of these systems believe they are actually constructing. (Microsoft Annual Report 2025, Alphabet Q4 2024 Earnings Call, Amazon Annual Report 2024, OpenAI – Announcing The Stargate Project)
That question becomes even more interesting when viewed from the perspective of professional services. For decades, firms have been built around a relatively simple assumption. Expertise is scarce. Firms recruit talented graduates, train them for years, expose them to thousands of client situations, accumulate institutional knowledge, and sell access to that expertise through projects, audits, legal opinions, tax advice, and consulting engagements. The scarcity of expertise justified the economics. It justified leverage models. It justified partner compensation. In many respects, it justified the institutions themselves.
Yet something unusual is beginning to happen. Increasingly, expertise appears capable of being separated from the expert. Tax expertise can be embedded inside software. Legal analysis can be generated through AI systems. Audit procedures can be automated through platforms. Technical research can be performed through models trained on enormous quantities of information. The professional remains important. Human judgment remains important. Accountability remains important. But expertise itself is beginning to change economic form.
We Have Seen This Movie Before
This would not be the first time a capability changed economic form.
For most of modern business history, organizations owned the capabilities that powered their operations. They purchased software licenses. They bought servers. They built data centers. They accumulated expertise inside the institution. Capabilities were treated as assets. They sat on balance sheets. They were depreciated over time. Ownership and control largely remained inside the organization.
The first major shift appeared with enterprise software. Activities that had previously existed inside manuals, procedures, training programs, and organizational memory gradually became embedded inside systems. Accounting processes became software. Procurement processes became software. Payroll processes became software. What had once existed as organizational knowledge increasingly became infrastructure. Yet ownership remained largely unchanged. Companies still purchased the software and controlled the environment in which it operated. (Marc Andreessen – Why Software Is Eating the World)
The larger shift arrived later. Software itself stopped being something organizations owned and increasingly became something they consumed. Salesforce, Workday, ServiceNow, and hundreds of other SaaS providers transformed software from an asset into a service. What had previously been purchased became rented. What had previously been depreciated became an operating expense. Most organizations experienced this as a technology transition. Economically, however, something more important had happened. Ownership moved upstream. (Salesforce FY25 Annual Report, Workday FY2025 Results, ServiceNow 2025 Annual Report)
Cloud computing repeated the same pattern. Organizations that once owned servers, storage systems, and networking infrastructure increasingly rented computing capacity from Amazon, Microsoft, and Google. Again, ownership moved upstream. Again, value accumulated around the owners of the infrastructure layer. Looking back, the pattern appears surprisingly consistent. Capabilities become standardized. Capabilities become infrastructure. Eventually, capabilities become services consumed on demand. The economic center of gravity gradually shifts toward those who own the platform. (Amazon Annual Report 2024, a16z – The Cost of Cloud, a Trillion Dollar Paradox)
Following the Pattern
At first glance, expertise appears fundamentally different from software or computing infrastructure. Expertise lives inside people. It is developed through experience. It depends on judgment, context, interpretation, and professional training. Yet many of the same things were once said about business processes and computing infrastructure. Before software, accounting appeared inseparable from accountants. Before cloud, computing appeared inseparable from servers sitting inside corporate data centers. In both cases, the capability survived. What changed was the way the capability was delivered.
The same possibility now exists for expertise. Historically, if an organization wanted expertise, it needed access to a person. Increasingly, organizations can access expertise through workflows, platforms, agents, APIs, and models. The expert remains important. The institution remains important. But expertise itself begins to look less like labor and more like a capability that can be delivered through infrastructure.
Viewed through this lens, a different interpretation begins to emerge. Perhaps the most important AI story is not that machines become intelligent. Perhaps the more important story is that expertise itself is becoming infrastructure. And if that interpretation is correct, the implications for professional services may be far larger than most current discussions assume.
Because history suggests that once a capability becomes infrastructure, the economics of the industry built around that capability rarely remain unchanged.
When Expertise Becomes Metered Infrastructure
The strongest argument that expertise may be changing economic form does not come from technology. It comes from economics.
For decades, organizations consumed expertise in much the same way they consumed professional labor. If a company needed tax advice, it hired tax specialists. If it needed legal advice, it hired lawyers. If it needed strategic advice, it hired consultants. Expertise was delivered through people. The economics were therefore largely labor economics. More demand required more professionals. More professionals required more hiring, more training, and more management. Expertise could be scaled, but only within the limits imposed by human capacity.
Increasingly, however, expertise is becoming accessible through systems rather than exclusively through professionals. A large language model can summarize regulations, generate legal analysis, propose tax structures, draft contracts, review code, identify anomalies, or evaluate strategic options. The output is not always correct. The output still requires oversight. Yet something important has changed. For the first time, expertise itself can increasingly be consumed through infrastructure. Organizations are no longer accessing expertise only through people. They are increasingly accessing expertise through platforms.
This is where the historical pattern becomes difficult to ignore. Business processes became software. Software became a service. Computing became a utility. Each transition moved a capability from ownership toward consumption. Each transition shifted value toward the owners of the infrastructure layer. The same pattern may now be appearing in expertise. Not because expertise disappears, but because expertise increasingly becomes available on demand, delivered through platforms, consumed through APIs, and paid for continuously.
At that point, a different interpretation of the current AI race begins to emerge. Perhaps Microsoft, Amazon, Alphabet, OpenAI, Anthropic, and others are not merely competing to build better models. Perhaps they are competing to build the infrastructure layer through which future expertise will be delivered. If that interpretation is correct, the current investment levels suddenly become much easier to understand.
The Margin Illusion
One of the most common assumptions surrounding AI is also one of the most intuitive. If expertise becomes cheaper to produce, margins should improve.
The logic appears straightforward. A consultant supported by AI can complete work faster. An auditor can review more transactions. A lawyer can analyze more documents. A tax professional can evaluate more scenarios. If the cost of producing expertise falls, profitability should rise. Similar arguments accompanied the rise of enterprise software, SaaS, and cloud computing. In each case, technology promised greater efficiency and lower costs.
Yet history suggests that lower costs often produce a different outcome. When cloud computing emerged, organizations expected infrastructure spending to fall. In many cases, the cost per unit of computing declined dramatically. What followed, however, was not lower consumption. It was an explosion in consumption. New applications appeared. Data volumes multiplied. Analytics expanded. Entirely new categories of software emerged. The infrastructure became cheaper. Organizations simply used far more of it.
The same pattern appeared in software. As software became easier to deploy and consume, organizations did not purchase fewer systems. They purchased more. More applications. More reporting. More analytics. More integrations. More compliance systems. Lower friction changed behavior. The efficiency gains were real, but so was the growth in demand.
The same dynamic may emerge with expertise. A board that previously commissioned one strategic scenario analysis may commission twenty. An audit committee that previously reviewed samples may expect full-population testing. A regulator that previously accepted periodic reviews may demand continuous monitoring. A private-equity investor that previously received monthly reporting may expect daily insights. Expertise may become cheaper per unit while overall expertise consumption expands dramatically. What initially appears to be a productivity story may ultimately become a demand story.
The Platform Gravity Problem
History offers another clue.
When capabilities become infrastructure, value rarely remains evenly distributed across the ecosystem. More often, economic gravity begins pulling data, workflows, users, and economics toward the owners of the platform layer. Few organizations adopted Salesforce because they wanted to strengthen Salesforce. Few organizations moved workloads to AWS because they intended to increase Amazon’s economic power. Yet that was often the result. As more processes moved into the platform, the platform itself became increasingly valuable. (Stratechery – Aggregation Theory)
One reason this happens is that platforms tend to accumulate advantages over time. More users generate more data. More data improves workflows. Better workflows attract more users. Ecosystems form around the platform. Gradually, the platform becomes the place where value is created, captured, and coordinated. What initially appears to be a technology decision becomes an economic one.
The same question now emerges around expertise. If expertise increasingly arrives through infrastructure, another question follows naturally. Who owns that infrastructure? The history of software and cloud suggests that value rarely remains evenly distributed across an ecosystem. More often, economic gravity gradually accumulates around the owners of the platform layer. This may be one of the most important strategic implications of AI. The challenge for professional-services firms may not simply be adopting AI. The challenge may be determining where they sit relative to the infrastructure that increasingly delivers expertise.
Some firms hope to build proprietary expertise platforms and keep the gravity inside the institution. Others may rely on industry-specific platforms sitting above individual firms. A third possibility is more disruptive. The ultimate gravity may settle with the owners of the foundation models themselves. In that scenario, professional-services firms increasingly operate on infrastructure they do not own. The expertise remains theirs. The client relationships remain theirs. Yet the platform economics increasingly sit elsewhere.
That possibility may help explain why the current AI race feels so intense. The largest technology companies may not simply be competing to build better models. They may be competing to own the platform layer through which future expertise is delivered. If that interpretation is correct, the strategic challenge facing professional-services firms is not merely how to adopt AI. It is how to avoid becoming economically dependent on the very infrastructure through which expertise increasingly flows. (OpenAI – Announcing The Stargate Project, Reuters – Companies Pouring Billions Into AI Infrastructure)
The history of software and cloud suggests that once platform gravity becomes sufficiently strong, value tends to migrate toward the center. Professional-services firms therefore face a question that extends far beyond technology. As expertise becomes infrastructure, where will value ultimately accumulate? Inside the firm? Inside industry platforms? Or increasingly with the owners of the models themselves?
And that raises a more fundamental question.
If expertise becomes abundant, what remains scarce?
What Remains Scarce?
For much of the past two years, discussions about AI have focused on what the technology can do. The assumption often sits just below the surface. If models become sufficiently capable, expertise itself ceases to be scarce. The conclusion appears obvious. If expertise becomes abundant, the institutions built around expertise must inevitably become less important.
Yet that conclusion may be too simplistic.
The history of professional services is not simply a history of expertise. It is also a history of judgment. A tax partner advising on a routine deduction is applying expertise. A tax partner advising on a transaction structure that has never existed before is exercising judgment. An auditor reviewing a standard control is applying expertise. An auditor deciding whether an unprecedented event fundamentally changes a firm’s risk profile is exercising judgment. A consultant applying a familiar framework is using expertise. A consultant helping a board navigate an uncertain future is exercising judgment. The distinction matters because expertise and judgment do not behave in the same way.
Much of what organizations describe as expertise consists of accumulated knowledge, precedent, learned practice, pattern recognition, and analytical capability. These are precisely the areas where AI appears strongest. Models absorb enormous quantities of information. They identify patterns. They retrieve precedent. They generate recommendations. In many situations they can perform these activities at a scale impossible for individual professionals. This is the part of expertise that increasingly appears capable of becoming infrastructure.
Judgment is different.
Judgment becomes most valuable precisely where precedent is weakest. It operates when information is incomplete, when objectives conflict, when consequences are significant, and when no obvious answer exists. Judgment is not merely about finding an answer. It is about deciding which answer to stand behind. The more expertise becomes abundant, the more visible this distinction may become.
The Challenge Nobody Is Talking About
This distinction creates a challenge that receives remarkably little attention.
For decades, professional-services firms developed judgment by exposing junior professionals to large volumes of expert work. They reviewed contracts, prepared analyses, performed audit procedures, researched regulations, and worked through hundreds of situations before eventually encountering the one situation that did not fit the pattern. Expertise and judgment developed together because the path to judgment ran through expertise.
Yet many of the activities through which firms historically developed judgment are precisely the activities now being targeted for automation.
This creates a paradox that may prove more important than many of today’s discussions about productivity, automation, or headcount. The more successful firms become at industrializing expertise, the more carefully they may need to think about how they continue developing judgment. The challenge may therefore not be whether AI replaces expertise. The challenge may be whether institutions can continue producing judgment once expertise increasingly arrives through infrastructure.
Professional-services firms have spent decades refining systems for producing expertise. They may now need to develop equally sophisticated systems for producing judgment.
Why Trust May Become More Valuable
This also helps explain why discussions about trust, governance, accountability, and oversight continue to surface despite rapid advances in AI.
At first glance, this appears contradictory. If expertise becomes more abundant, one might expect professional institutions to become less important. Yet history often moves in the opposite direction. As information becomes abundant, trust becomes more valuable. As recommendations become easier to generate, accountability becomes more important. As analysis becomes cheaper, judgment becomes more scarce.
Public-interest audit provides perhaps the clearest example. The value of an audit opinion has never resided solely in the analysis performed. The value comes from the institution standing behind the conclusion. The same logic increasingly applies to legal opinions, tax advice, cyber-security assessments, regulatory compliance, and many other professional activities. In each case, the client is not merely purchasing expertise. The client is purchasing judgment, accountability, and trust. (IAASB – Auditor Reporting and the New Auditor’s Report, IFAC – Achieving High-Quality Audits)
Viewed through this lens, AI may not weaken the Regulated Trust Layer. It may strengthen it. The more expertise becomes infrastructure, the more important trusted institutions may become in helping clients interpret, validate, govern, and ultimately act on that expertise.
A Different Future for Professional Services
Many current discussions about AI implicitly assume that professional-services firms face a choice between resisting AI and embracing AI. The reality may be more complicated.
The firms that succeed may not be those that simply accumulate the most AI tools. Nor may they be the firms that attempt to protect traditional expertise models from change. Instead, the winners may be the firms that understand how expertise changes once it becomes infrastructure and adapt their institutions accordingly.
For decades, professional-services firms were built around the scarcity of expertise. Recruitment, training, promotion, leverage, and economics all reflected that reality. If expertise increasingly becomes abundant, those institutions may gradually evolve. Expertise remains important. But the source of value may increasingly shift toward judgment, trust, governance, accountability, and the ability to make decisions under uncertainty.
The most successful firms may therefore look less like expertise factories and more like institutions that combine industrialized expertise with exceptional judgment. AI may allow them to automate the production of expertise while concentrating human talent on interpretation, governance, decision-making, and trust. The opportunity is not to compete with the infrastructure layer. The opportunity is to build value on top of it.
Closing Thoughts
The current AI debate is often framed around replacement. Will consultants disappear? Will lawyers disappear? Will accountants disappear? Will auditors disappear?
Those questions are understandable. They are also increasingly becoming the least interesting questions.
A more interesting question is what happens when expertise changes economic form.
The history of business over the past fifty years can be viewed as a series of transitions in which scarce capabilities became infrastructure. Business processes became software. Software became a service. Computing became a utility. The next transition may already be underway. Expertise itself may be becoming infrastructure.
If that interpretation is correct, many current developments begin to look different. The largest technology companies are not simply competing to build better models. They are competing to own the infrastructure layer through which future expertise may be delivered. Professional-services firms are not simply adopting AI. They are adapting to a world in which expertise becomes increasingly abundant. The challenge is no longer how to protect expertise. The challenge is how to create value once expertise is no longer scarce.
Professional-services firms spent decades building institutions around the scarcity of expertise.
They may spend the next decade building institutions around the scarcity of judgment.
What This Means for Boards
Boards should resist viewing AI primarily as a technology initiative. The larger issue is economic and strategic.
The questions worth asking are not simply how AI can reduce costs or improve productivity. The more important questions may be where expertise is becoming infrastructure, where platform gravity is accumulating, which capabilities remain scarce, and how the institution continues developing judgment in an environment where expertise increasingly arrives pre-packaged through platforms.
Continue Exploring
The Bigger Picture
The Professional Services Transformation Theory
Why artificial intelligence, changing economics, new operating models and growing investment requirements are part of a broader structural transformation of professional-services firms.
Explore the Economics Underneath AI
The AI Cost Stack: Why Professional Services Firms Are Looking for AI Savings in the Wrong Place
Why cheaper cognitive work does not necessarily translate into lower firm costs when technology, orchestration, governance and infrastructure become larger parts of the economic system.
Follow the Argument to Judgment
Manufacturing Judgment: How Professional Services Build Their Most Valuable Asset
If expertise becomes increasingly abundant and accessible through infrastructure, professional judgment becomes more important—and raises the question of how firms will continue developing it.
Follow the Argument to Trust
Trust Capital: Why Professional Services Manufacture Judgment but Accumulate Trust
Why professional judgment only creates economic value when clients, regulators and other stakeholders are prepared to rely on it, making institutional trust another critical asset of the professional-services firm.
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
Primary Sources
• Microsoft Annual Report 2025
• Alphabet Q4 2024 Earnings Call
• OpenAI – Announcing The Stargate Project
• Anthropic – Amazon Compute Collaboration
• Salesforce FY25 Annual Report
• ServiceNow 2025 Annual Report
• NIST – The Definition of Cloud Computing
• Thomson Reuters – Future of Professionals Report 2025
• World Economic Forum – Jobs of Tomorrow: Large Language Models and Jobs
• McKinsey – The Economic Potential of Generative AI
• IAASB – Auditor Reporting and the New Auditor’s Report
• IFAC – Achieving High-Quality Audits
Secondary Sources
• Marc Andreessen – Why Software Is Eating the World
• a16z – The Cost of Cloud, a Trillion Dollar Paradox
• Stratechery – Aggregation Theory
• Reuters – Companies Pouring Billions Into AI Infrastructure