Goldman Sachs' Silicon Valley field research: AI accelerates industry reshuffling; data barriers, workflows, and pricing power build corporate moats.

AI is reshaping the competitive landscape of the information and business services industries. The uniqueness of data assets, control over workflows, and monetization capabilities will determine which companies will emerge victorious in this transformation.

According to TrendFocus, Goldman Sachs released a research report after completing its third annual Silicon Valley AI field research. Meetings with AI companies, venture capital firms, and academic researchers from Stanford University, UC Berkeley, and other institutions reinforced four core judgments: the value of proprietary information will continue to increase; AI applications are extending from information transmission to workflow execution; physical service and education platforms are more resilient than business models reliant on white-collar workforces; and monetization pathways will expand from subscription to usage volume, transaction volume, product premiums, and outcome pricing.

Research Background: First-hand information from Silicon Valley forms the basis for analysis.

Goldman Sachs held its third annual Silicon Valley AI field trip from August 18 to 19. Participants included AI companies such as Daloopa, Clio, Harvey, Corgi, Moody's, Vercel, and ClickHouse, as well as venture capital firms such as Lightspeed Venture Partners, Kleiner Perkins, Menlo Ventures, and Index Ventures. There were also special sessions for researchers from Stanford University and UC Berkeley/UC San Francisco.

This survey focused on key issues including: the accelerated evolution of model capabilities, workload distribution between open and cutting-edge models, the transformation of AI agents from auxiliary tools to workflow executors, and the sources of enduring competitive differentiation in areas such as data, domain expertise, customer scenarios, and workflow ownership. In addition, the survey also covered topics such as AI monetization, validation and accountability, cybersecurity, and the infrastructure required for entity AI and agent-native workloads.

Theme 1: Building the Clearest Information Barrier with Proprietary Data

Goldman Sachs believes that while model capabilities continue to improve, they still rely on external information that they cannot generate themselves. This benefits companies with unique, continuously updated datasets that are difficult for competitors to replicate. Simultaneously, the ability to embed this information into customer AI agents and applications can expand usage beyond traditional user interfaces. The risk is that new interfaces may diminish the value of existing search and display products, shifting the value center to the application layer that controls the customer experience.

The positive implications of MCO are most evident. Moody's demonstrated in its research how it transforms insurance materials submitted via PDFs, spreadsheets, and free-form documents into standardized underwriting data within seconds, compared to hours or even days typically required for manual processing. The platform can also parse non-standard policy language and perform catastrophe loss analysis and reserve calculations through a natural language interface. These workflows heavily rely on MCO's proprietary data, entity graphs, catastrophe models, and domain expertise, making them difficult for generic applications to replicate. Smart APIs, MCP servers, and intelligent agent products provide customers with a path to invoke this intelligence within their own workflows, also contributing to incremental subscription and usage revenue.

In contrast, FDS faces significant pressure. Daloopa demonstrated how to automatically collect financial data and maintain models using publicly disclosed documents from listed companies, information that can also be accessed by FDS and other data providers. The Daloopa platform maps disclosed information to client-defined Excel models and provides source links for each data point. Because a large portion of FDS's core financial data is aggregated and standardized from public or third-party channels, its data is highly reproducible, potentially putting pressure on the value of its workstation functionality.

SPGI possesses differentiated assets including Market Intelligence, ratings, commodity benchmarks, indices, private markets, and professional pricing. It has multiple monetization paths in machine-based consumer pricing and new workflow products, which Goldman Sachs interprets as "positive/mixed." The key variable lies in how AI changes the user interaction model of Capital IQ Pro. IT (Gartner), on the other hand, faces greater content risk—Harvey and Simile both indicate that AI can transform standardized research content into reusable intelligent agent workflows and enable companies to generate new analyses without commissioning separate research projects. This could reduce the scarcity of standardized research and training materials, which Goldman Sachs interprets as "negative/mixed."

Theme 2: Workflow control determines value attribution

Goldman Sachs observed a key evolution in AI during its research: Harvey and Clio demonstrated how AI can evolve from retrieving information to drafting documents, coordinating tasks, and completing more complex professional workflows. Whoever controls the platform that allows clients to get their work done is likely to capture economic value beyond simply being a data inputter.

TRI possesses a strong workflow position, enabling it to combine authoritative Westlaw legal content with CoCounsel, extending to workflows such as legal research, drafting, and document review. Clio's $1 billion acquisition of vLex in 2025, gaining access to a major legal content library encompassing over 1 billion legal documents from more than 110 countries, underscores the strategic value and complexity of building global legal content assets. However, if Harvey or Clio controls the client interface and workflows, TRI could become merely a bottom-up content provider—even if its content demand remains strong. Goldman Sachs assigns a "positive/hybrid" connotation to TRI.

The positive implications of FICO and EFX are most explicit. Stanford University researchers pointed out in their research that the main obstacle preventing enterprise intelligence agents from entering the production environment is not model capabilities, but the unresolved issue of accountability mechanisms—clearly defined, verifiable, and reversible processes will be implemented first. This framework aligns highly with FICO's capabilities in decision-making and governance. EFX's The Work Number, with its employer-contributed employment and income records, is a highly difficult-to-replicate differentiated asset. As AI agents evolve from generating suggestions to initiating decisions, the need for verification will occur earlier and more frequently.

VRSK faces some pressure regarding its competitive advantage. Corgi demonstrated how an AI-native commercial insurance company can automate its underwriting process without heavily relying on VRSK data. Corgi pointed out that most insurance pricing methods, filing rates, and loss ratio data in the United States are publicly disclosed by state regulators, allowing AI-native companies to train their systems on this basis. VRSK still retains advantages such as proprietary claims history data, catastrophe models, and deep integration into underwriter workflows, but Corgi challenges the assumption that "all traditional insurance data becomes more difficult to replace in the AI era." Goldman Sachs assigns VRSK a "mixed/positive" interpretation.

ULS (UL Solutions) possesses significant long-term option value. A Stanford research session cited practices in safety-critical industries such as aviation, highlighting the potential demand for an "AI-powered SOC 2," which aligns perfectly with ULS's expertise in testing, inspection, and certification. However, uncertainties remain regarding standard setting, market demand, and revenue contribution timelines.

Theme 3: Physical service and education platforms are more resilient to shocks

The risk of AI replacing white-collar jobs puts direct pressure on human-driven business models, but Goldman Sachs believes that physical service and education platforms are in a more advantageous position.

In the staffing sector, RHI faces the biggest impact. Daloopa's presentation directly revealed RHI's exposure in accounting, finance, administration, and junior technical positions. AI-automated data acquisition, model maintenance, and structured knowledge processing will reduce the demand for these roles, and internal hiring by clients may further reduce staffing volumes. Emerging data engineering, cybersecurity, and model implementation roles may struggle to fill the gap, a view Goldman Sachs described as "negative/mixed." MAN (Manpower) is relatively protected due to its higher proportion of blue-collar jobs, but its Experis brand still faces pressure to automate routine IT support and analytics tasks.

In the education sector, MH presents a clear AI revenue opportunity. Over 100 million paid course licenses form the foundation for distributing AI-driven course content, and 7.5 million users have already adopted AI products in assessment, learning aids, clinical simulations, and teacher tools. MH has identified fixed markups, question-based pricing, and token consumption as potential monetization models, with a potential licensing premium of 5% to 15%.

In the physical services sector, ADT, CTAS, and ROL all saw positive implications. ADT can apply AI to monitoring and incident grading, supporting advanced service levels, customer retention, and video analytics products; CTAS can enhance cross-selling and service response capabilities through route data and customer history; and ROL is expected to leverage property-level pest and disease activity data and the Stanford World Model research framework to explore predictive detection and high-end monitoring products. A common characteristic of these three companies is that AI can enhance service value without replacing the core physical activities purchased by customers.

Theme 4: Monetization Path – Diversified Pricing Logic Beyond Seats

Goldman Sachs identified three main revenue streams and emphasized that AI monetization requires combining differentiated assets with quantifiable customer value, rather than simply converting existing seat revenue into new pricing units.

The logic behind usage and transaction pricing lies in the fact that AI agent applications will generate far more data requests, verification events, and automated decision-making than human-operated software. MCOs and SPGIs can monetize machine access to proprietary information; FICOs and EFXs should benefit from the increased frequency of automated decision-making and verification. The main risk is that the growth in usage will remain bundled into existing corporate contracts rather than representing incremental expenditure.

Regarding product premiums and outcome-based pricing, TRI can charge for a broader range of legal workflows through CoCounsel, MH has identified three models: fixed markup, issue-based pricing, and token consumption, and ADT can develop advanced monitoring and video analytics products. These models may be more durable than simple token-based billing because clients can directly link spending to completed tasks or improved outcomes.

In terms of infrastructure demand, IRM offers more direct revenue opportunities. Vercel and ClickHouse's research indicates that the request volume, token consumption, and telemetry data volume generated by intelligent agent applications far exceed those of traditional software, directly supporting data center leasing and the lifecycle services needs of hyperscale assets. Since customers have already paid for infrastructure and information processing, IRM does not need to build a new AI pricing model; execution capability and capital investment intensity are the more critical constraints.

Regarding revenue disruption risk, FDS, IT, RHI, and PBI face the risk of existing revenue being pressured before the monetization of new products. Goldman Sachs believes that all four companies are likely to launch competitive AI products, but the potential time mismatch between disruption and monetization underscores a more cautious investment stance.

~~~~~~~~~~~~~~~~~~~~~~~

The above content is from Zhuifeng Trading Platform .

For more detailed analysis, including real-time updates and firsthand research, please join the [ Trading Channel Annual Membership ].

Risk Warning and DisclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.