Proprietary data ain't what it used to be and picking winners in the Data & Analytics sector
This week in Data & Analytics: Vladimir Lenin, Anthropic, Elliott Management, S&P Global, Mighty Ducks, and GLG
The sell-off continues. In this week’s newsletter we take another look at the implications of the Anthropic plugin-induced selloff and zero in on what differentiates Data & Analytics companies in the age of AI.
Asymmetrix is no longer just a Substack newsletter; we are a full-fledged data and research company providing critical intelligence to stakeholders in the Data & Analytics industry. Our clients include private equity funds, investment banks / corporate finance houses, management consultants, equity analysts and more - all keen to stay up to date and informed on the Data & Analytics industry. If that sounds like you, feel free to reach out directly at asymmetrix@asymmetrixintelligence.com or via our website www.asymmetrixintelligence.com for a demo.
“There are decades where nothing happens; and there are weeks where decades happen.”
Anyone following the software and data industries in recent weeks can relate to the above quote (wrongly attributed to Vladimir Lenin, whose name is not on the bingo board for this Substack).
Digesting the full scale of the release of Anthropic’s legal plugin and the ensuring sell-off will will take a while, but in the meantime it’s been a busy couple of weeks:
January 30th: Anthropic releases 11 open-source plugins for its Claude Cowork platform, including a legal workflow plugin.
February 2nd – 3rd: Investors interpret Claude’s new legal workflow tools as a major disruption, causing a sharp sell-off (to the tune of nearly $300bn in market value wiped off) in software, legal tech and data stocks.
February 3rd: Jefferies’ Jeffrey Favuzza coins the term “SaaSpocalypse” to describe the sell-off.
February 5th: Asymmetrix’s Substack post analyzing why we believe investors in public data companies overreacted to the news of Anthropic’s new AI tool goes viral (well, relatively speaking).
February 10th: Hundreds of millions of dollars in market value is wiped off Ion Group’s $10bn debt stack amid a sell-off in the company’s bonds; concerns over AI’s impact on software and data companies persists. The term “loan-ageddon,” first used in January, takes on new meaning.
On the same day, S&P Global reports Q4 2025 earnings, the first major information services business to do so following the sell-off. Shares continue to fall as analysts and investors remain uneasy about the impact of AI on data companies.
February 11th: The Financial Times reports that activist investor Elliot Management has built a stake in London Stock Exchange Group (LSEG), a leading exchange and financial data conglomerate that was among the victims of last week’s sell-off. No details on the size of the stake or Elliot’s demands for refocusing the business have been shared.
Last week, we shared our initial, raw reaction to the havoc Anthropic’s legal plugin wreaked on data businesses’s market caps – and why we believe investors lacked nuance in viewing data businesses as vulnerable to AI disruption.
As we continue to digest the implications of Anthropic’s new legal plugins and AI workflow tools more broadly, we thought it prudent to share more nuanced perspectives on the impact these will have on the Data & Analytics industry. Here goes it.
AI products are table stakes
Perhaps this is obvious, but it is worth stating plainly:
AI product features are table stakes for Data & Analytics businesses.
If you run a data business, prioritize rollouts of client-facing AI features. Today, clients expect them. Soon, they will demand them. You may be able to survive on proprietary data or research a while – but not for long.
On S&P Global’s Q4 2025 earnings call, CEO Martina Cheung highlighted the company’s “flexible distribution policy,” enabling clients to use S&P data wherever they prefer. In addition to S&P’s “hundreds” of distribution partners, S&P has added LLMs to the mix.
As corporate LLM use accelerates, data companies can’t afford not to meet clients in these environments – just as they did with Snowflake, Salesforce and SAP. Failing to do so mean not capturing all the value at stake and leaving money on the table.
Don’t be careless, but don’t be too careful either
Coach Orion’s axiom rings true in this context. Cheung emphasized that S&P does not allow LLMs to train on its data.
While it is certainly sensible not to let the wolf into the fold, providing limited public access to data builds trust, demonstrates credibility, and generates demand.
Placer.ai proved this playbook works: sharing small data samples and timely public reports helped the company establish authority and credibility in the real estate and retail industries. This approach doesn’t cannibalize revenue; strong clients will not settle for shallow slices of a robust, proprietary dataset.
Similarly, allowing LLMs controlled access to a quarantined subset of your data ensures your information is referenced when users query AI systems, a new and important medium of discovery.
Quality vs. and Quantity
With high-quality AI tools now far cheaper than enterprise data subscriptions, clients expect premium research products to deliver value beyond what LLMs can generate.
The tension can be articulated as follows:
Clients expect more content, since AI accelerates idea generation, research, fact‑checking, and drafting;
Clients also expect insights AI alone can’t provide, especially as tools like OpenAI’s Deep Research raise the baseline.
Data & Analytics companies must deliver both. They must use AI to increase research volume and efficiency but double down on what AI can’t replicate: insightful narratives, sector-specific expertise, direct engagement with industry stakeholders and proprietary intelligence.
As the bar rises, clients will expect providers to do more of the heavy lifting for them, contextualizing information and guiding decisions – ultimately the value clients pay for.
Highlight differentiated human insights
Clients are flooded with shallow AI-generated content that gives the impression of sophistication and expertise. They will pay a premium for substantive human insights drawing from proprietary intelligence and differentiated insider knowledge.
Data & Analytics companies must convey clearly what is human insights and what is AI – this will build trust and focus clients on truly value add, difficult to obtain data and research.
Further, the onus is on Data & Analytics providers to prove to the market that they are selling unique, proprietary, and quality content worth paying a premium for. Sales and marketing teams must tell a compelling story that emphasizes companies’ unique value proposition and conveys the “must-have” nature of its data. They must draw a clear contrast with the all-knowing, confident and cheaper AI tools that lack depth of insight and proprietary content.
Not all proprietary data is created equal
In the past week, much of the market has made a clear distinction between commoditized public data and proprietary, difficult to obtain data. The latter is defensible in the face of ever-improving AI tools, while the former is likely to be disrupted and easily replicated.
Furthermore, AI workflow tools require accurate, reliable and high-quality proprietary data to provide the best results. Scraping the public web will get you part way but for the full answer, proprietary datasets from Data & Analytics providers need to be plugged in. So current AI research output effectively serves to whet users appetite for better quality Data & Analytics providers.
That said, not all proprietary data is valued the same.
Data can be proprietary without being unique. Expert network transcript libraries, such as those offered by Tegus, GLG, AlphaSights and others, are proprietary in that each library contains call transcripts that are not replicated anywhere else word-for-word. However, the underlying subject matter often overlaps significantly, as providers tend to cover similar sectors, companies and investment themes driven by client demand. As a result, while the specific transcripts may differ, the content across platforms may not be unique. In this context, differentiation may depend more on size of the library and functionality rather than on quality or insightfulness of the content.
Data can be proprietary but not deeply embedded in client workflows or critical to day-to-day operational decisions. For example, Gartner gathers proprietary information through direct engagement with industry vendors and extensive client interviews, which inform its research and market analysis. However, the perception among some is that Gartner’s output is high-level and generalized, serving primarily as strategic guidance or third-party validation rather than as operational data integrated into core business systems. As a result, while Gartner may influence high-value strategic decisions, its research is not typically embedded into execution workflows or relied upon as a real-time input for operational decision-making.
Not all workflows are created equal
In some sectors and workflows, data need only be directionally accurate rather than precise. These lower-stakes workflows inform decisions but are not directly tied to transactions, regulatory filings, or immediate financial results. For these workflows, AI tools that draw from data that is “good enough” can meaningfully disrupt incumbent data providers, particularly where cost and speed are prioritized over precision. These providers must work hard NOW to differentiate themselves.
In contrast, workflows with high financial stakes, regulatory scrutiny, or health and safety implications, such as securities trading, healthcare and clinical research, and fraud detection, demand high accuracy, auditability, and reliability. In these workflows, defensibility derives not only from proprietary data, but from accuracy and reliability. AI systems relying solely on scraped or lightly curated public data are unlikely to displace incumbents in these settings. However, AI paired with mid-tier data providers may chip away at smaller clients without fully displacing premium providers embedded in critical workflows where accuracy is imperative.
Picking winners
Proprietary data is a defensible moat for data businesses in the age of AI – but only when it serves high-stakes workflows where accuracy and reliability materially affect outcomes. Providers that sell proprietary data but serve low-stakes workflows may be more vulnerable to disruption. Lower-cost, AI-enabled alternatives may be “good enough.”
The most attractive Data & Analytics businesses will combine:
Difficult-to-obtain proprietary data – not easily scraped or replicated by AI at scale;
High-quality data critical to high-stakes workflows – where accuracy and reliability matter because clients cannot afford to make mistakes;
Insights and intelligence layer – human intelligence and anaysis beyod what AI alone can provide;
Integration with AI tools – to ensure data is embedded into client workflows.
📜 Interesting Content
Clearwater B2B Information Services Q4 2025 Report - Clearwater
The market meltdown in data companies is profoundly wrong - Evercore’s Nathan Graf in The FT
Moats Matter Again - Travis May on Medium
UK’s Argus lobbied Moscow over rules threatening its Russia business - The FT
Frontier Alpha: The Great Repricing and the Industrialization of Trust - The Judgment Observatory
Are You a Software Company or a Data Company - Dan Entrup on It’s Pronounced Data
Governance, Risk & Compliance (GRC) Technology Key Themes - D.A. Davidson on LinkedIn
Senator Accuses Equifax of ‘Price Gouging’ Medicaid Programs - New York Times
Elliott Management builds stake in London Stock Exchange Group - Reuters