Consensus and Non‑consensus Framework: AI Will Reshape the PSF Economic Model

Consensus and Non‑consensus Framework

In investing, people distinguish between consensus and non‑consensus. Consensus reflects the prevailing market narrative, whereas a non‑consensus view is a contrarian view that diverges from the mainstream. Consensus views are typically priced in, yielding only market returns. In contrast, a non‑consensus view generates alpha if it proves correct. For example, when many believed Google Search was threatened by AI‑driven chat, putting downward pressure on the stock price, contrarians argued that Google's scale would sustain its advantage. The subsequent share‑price increase (from approximately $160 to over $260) rewarded that view. Similarly, the consensus initially viewed BABA as a goddamn retailer in a competitive industry. The potential of Alibaba Cloud and its leading role in China's AI infrastructure were only recognized by non‑consensus investors. True alpha comes from seeing what the crowd misses and being right about it.

This consensus/contrarian framework can be applied to AI's impacts on PSFs as well. My analysis splits into two parts: first, the commonly held beliefs about how AI will influence PSFs, and second, my non‑consensus view on AI's long‑term disruption potential in the industry.

Consensus on AI in Professional Services

Traditional PSFs Rely on a Leverage Model but Cannot Easily Scale

Professional services firms (PSFs), such as consultancies, law firms, and accounting firms, rely on a leverage model based on the ratio of juniors to seniors on projects. The profitability depends on delegating as much work as possible to lower-cost juniors billed at margin, while maintaining a pyramid structure. David Maister's framework categorizes professional services work into three types:

• Brain projects: Novel, complex problems requiring senior expertise with minimal delegation and limited leverage.
• Gray Hair projects: Familiar problems with customized solutions leveraging firm experience.
• Procedure projects: Well-defined, repeatable work suitable for high junior leverage, yielding high profit margins.

Success in professional services requires firms to strategically calibrate their leverage model: adjusting the macro structure to align with their client market positioning and the micro structure to match individual project types. From an economic perspective, the top‑line performance of a PSF can be roughly represented by the following formula:

Fees = Rate × Utilization × Capacity

The rate is determined by the market structure and the nature of the services; thus, it varies in the short run but remains relatively stable in the long run. Utilization represents a trade‑off between today's revenue and tomorrow's revenue and is capped in healthy firms. The available capacity per headcount is fixed. Consequently, this linear economic model means firms must add people to add revenue. True economies of scale are therefore limited.

AI Tech Stack for PSFs

AI is widely regarded as a game-changer in the economics of professional services, enhancing both profitability and scalability that transcend the constraints of the traditional leverage model. By augmenting human expertise and automating routine tasks, AI enables firms to break the linear relationship between headcount and revenue growth.

For high‑complexity Brain projects, the AI tech stack should comprise intelligent knowledge bases and retrieval‑augmented generation (RAG). While a PSF's financial balance sheet may appear modest in tangible assets, its economic balance sheet comprises substantial intellectual capital: domain expertise, proprietary data, and accumulated project experience that underpin the firm's margins. In the AI era, these intellectual assets need to be systematically documented and structured into database formats as knowledge bases. RAG represents the current optimal approach for integrating LLMs with proprietary firm data while circumventing context‑window limitations. In the medium to long term, vertical agents will be constructed atop RAG systems to rapidly retrieve and synthesize information from both proprietary and third‑party repositories, greatly accelerating senior professionals' ability to solve novel problems.

The value proposition for Brain projects centers on revenue enhancement through augmented problem-solving capability. Projects will be delivered faster and with deeper insight, potentially winning more business and commanding higher fees.

Procedure projects have clearly repeatable steps, and they can be more easily programmed. Over the past decades, software vendors have codified many standardized processes, eroding portions of the PSF market. SAP, for instance, built a monopoly by transforming diverse business processes into a unified ERP architecture, shifting much value capture to software and leaving professional services as its implementer. When we look at the profitability of SAP and some key players in SAP implementation, such as Accenture and Capgemini, SAP's adjusted operating margin is clearly above those of large implementers, and revenue per employee is almost three times theirs.

Figure 1: SAP vs SAP Implementers - Operating Margin and Revenue per Employee

AI's role for procedure projects skews toward workflow automation and AI copilot assistance. Part of the routine work can be programmed and automated with the help of AI coding tools like Cursor, GitHub Copilot, or Claude. Even partial automation with simple scripts generated by AI can save significant time; case studies report notable reductions in turnaround time for data consolidation and analysis in client projects. LLM‑based copilots such as Microsoft Copilot or copilots embedded in commercial software can understand unstructured text (emails, reports, contracts, etc.) or multimodal data (images, videos, audio) to extract first insights. Some simple comprehension and generation tasks can be delegated to copilots, with outcomes heavily dependent on prompt quality. Senior professionals' expertise and experience will gradually play a key role here.

The value proposition for procedure projects is direct cost reduction through labor savings, faster delivery, and improved consistency. Projects can scale without proportional headcount increases.

Across the board, beyond the excessively optimistic or pessimistic noise in media and social media, the narrative is clear: AI is seen as a productivity tool and more likely to complement human labor, not substitute it. The reason lies in AI's limitation: machines still struggle with uniquely human capabilities, like empathy, ethical judgment, and creativity. In the short to mid term, the market consensus is AI is free, but clients pay the people in the loop.

Non‑consensus: My Contrarian View on AI Disruption in PSFs

Despite the reassuring consensus, my view on AI disruption in the professional services sector is more intense: over the mid to long term, let's say 5 to 10 years, AI will gradually eat the market of traditional PSFs and fundamentally reshape industry economics and organizational structures.

AI Will Eat the Professional Services Market

In 2011 Marc Andreessen wrote that “software is eating the world,” arguing that core activities in more and more industries were moving from people delivered services to software delivered services. Software companies have steadily captured value that previously flowed to service firms due to (1) scalability, (2) embedded best practices (e.g., SAP), and (3) always‑on execution. TMT has comprised a substantial share of S&P 500 market cap since software era.

AI could be the fastest technology shift in human history, even faster than software. ChatGPT reached 800 million weekly active users as of Oct‑25 in less than 3 years; by comparison, Facebook took 8 years and TikTok took 5 years to reach 1B MAU. In terms of AI performance for complex tasks, the length of tasks AI can handle is doubling every seven months.

Figure 2: AI Task Completion Length Doubling Every 7 Months (METR Benchmark)

Looking back into software history and transitioning to AI, in a scenario where AI achieves near‑human or superhuman performance across knowledge tasks, it could follow the same trajectory as software and capture a significant fraction of the value of global professional services, which currently make up around 6 trillion in annual aggregate revenues.

Historically, software couldn't replace human professionals because it lacked adaptability: code was rigid, and edge cases abounded. But modern AI is probabilistic: it can navigate ambiguity, learn from countless similar projects through its unprecedented bandwidth, and give non‑rigid, rule‑based judgments.

In the long run, AI will eat the professional services market. The existing professional services firm will be challenged by new entrants from the tech sector or transform into a hybrid software and services company. The boundary between a software company and a professional services firm is collapsing from both sides. Players who can't make this transformation will not be able to compete with lower cost, AI‑enabled competitors.

The Professional Services Economic Model Will Be Reshaped

Today's PSF is a human pyramid: A few senior professionals at the top bring in business, manage projects, and delegate tasks to a large number of junior staff performing research, analysis, and preparation. Deliverables are sold to clients at a margin through this leverage model to guarantee profits. As we have observed, more and more routine work has been or will be assisted or even taken over by copilots or vertical agent applications. On the micro level, the billable‑hours model is being challenged:

If work takes far less time, how should pricing adapt?

If clients understand that portions of delivery are AI‑assisted at lower marginal cost, how will they expect to pay?

In the long term and on the macro level:

can a human‑pyramid model persist if AI outperforms entry‑level tasks on cost and quality?

In the short to mid term, the competitive landscape will intensify: professional services, long protected by reputation and human capital, would see technology‑driven entrants erode margins. The industry will see significant contraction and consolidation. Only firms that adapt, by adopting AI themselves and perhaps redesigning their business models, will survive the shakeout. In the long run, the current billable‑hours pricing model will diversify into a hybrid of fixed‑price, subscription, or deliverable‑based models, and tomorrow's professional services firm will be a lean core of senior experts armed with advanced AI systems, delivering solutions directly. AI‑driven productivity increases will dramatically reduce labor needs, leading to a larger pie in terms of revenue and profitability.

AI's impact on professional services may mirror the Darwinian reorganization seen in other industries faced with disruptive technology. Take the EV sector as an example: the previous big players were so dominant, and the combustion‑engine technology seemed to have unbelievable moats, that few believed new players could break through entry barriers and survive the rivalry. One day the EV technology appeared and the old empire collapsed. The same happened in the internet sector: Yahoo dominated the internet's homepage with huge traffic and branding. No one could imagine that a company could defeat Yahoo until search technology emerged; Google didn't defeat Yahoo: the technology shift defeated Yahoo.

The threat of AI to PSFs is not simply that a rival uses AI more effectively; it is that a technology shift redefines the industry's rules, dismantling the prior paradigm. Nobody can foresee the future, but it's inspiring on the other hand: When the old order falls, opportunity rises.