Fiduciary-Grade AI: Managed AI in specialist publishing – How Thomson Reuters solves the trust issue of enterprise AI with Snowflake

Reuters opens archives: How journalistic content becomes valuable AI raw material

Fiduciary-Grade AI: Why trust is everything when using artificial intelligence

Artificial intelligence has established itself as a driver of innovation across industries – but in highly sensitive fields such as law, taxation, and regulation, its own set of rules applies. Where hallucinatory AI models or flawed legal provisions can lead not only to reputational damage but also to massive liability risks, blind trust in technology is not an option. To meet this challenge, the global information services provider Thomson Reuters is relying on a far-reaching partnership with the AI ​​data cloud Snowflake. At its core is the concept of "Managed AI" – a strictly controlled, trustworthy AI that prioritizes reliable results over sheer speed. The following article examines how Thomson Reuters is consolidating tens of thousands of databases, drastically accelerating critical work processes, and, incidentally, transforming its vast, decades-old news archive into a valuable resource for the global AI economy. Learn why excellent data governance is becoming a decisive competitive advantage in modern specialist publishing and what hurdles still need to be overcome when scaling enterprise AI.

“Fiduciary-Grade AI” is a specific type of Managed AI: It specifically describes the highest level of trust and governance within a managed AI environment, which is required for fiduciary-sensitive use cases such as law, tax and regulation, while Managed AI is the more general term for professionally operated, supervised AI infrastructure overall.

Thomson Reuters is a Canadian-British media and technology company headquartered in Toronto, formed in 2008 through the acquisition of the British news agency Reuters by the Canadian Thomson Corporation.

Reuters itself was founded in London in 1851 and is one of the oldest and largest news agencies in the world. Thomson Reuters remains majority-owned by the Woodbridge Company, the private investment company of the Canadian Thomson family, which holds approximately 67 to 70 percent of the shares. The company is listed on the New York Stock Exchange and the Toronto Stock Exchange under the ticker symbol TRI.

In fiscal year 2025, the group generated revenue of approximately US$7.48 billion and employed around 27,100 people worldwide in more than 100 countries. The company is essentially divided into five segments: Legal Professionals (legal research and workflow products for law firms and government agencies), Corporates (solutions for internal legal, tax, and compliance departments), Tax & Accounting Professionals (tax and accounting products), Reuters News (the traditional news business), and Global Print.

Thomson Reuters now describes itself as an “AI and technology company” that supports professionals with trusted content and automated workflows, building on over 150 years of accumulated content expertise. Among its best-known products are the legal research platform Westlaw and the generative AI assistant CoCounsel for legal professionals, both of which increasingly rely on a modernized, AI-powered data infrastructure – precisely the area at the heart of the previously analyzed Snowflake partnership.

Reliability before speed: The new benchmark for artificial intelligence in specialist publishing houses

AI without errors: How Thomson Reuters solves the trust problem in the legal system

Artificial intelligence is generally considered a growth accelerator, but in the legal, tax, and regulatory sectors, one thing counts above all: reliability. Anyone selling lawyers, tax advisors, and compliance departments an AI system that invents incorrect legal paragraphs or cites outdated regulations risks not only reputational damage but also legal consequences for their clients. This is precisely where the partnership between Thomson Reuters and Snowflake, presented at the Snowflake Summit 26 in June 2026, comes in. It demonstrates how a traditional media and information company can transform its extensive, highly sensitive data repositories into a robust foundation for generative AI without losing control over the origin, quality, and access of the data.

Thomson Reuters is a global provider of information services for legal, tax, and regulatory matters, supplying its clients with products such as Westlaw for legal research and CoCounsel, a generative AI assistant for legal professionals. Snowflake, in turn, positions itself as an AI data cloud, a platform that provides enterprise data centrally, securely, and in compliance with governance regulations for analytics and AI applications. The collaboration between the two companies provides a vivid example of how the concept of managed AI—professionally operated, monitored, and integrated artificial intelligence within existing business processes—proves successful in one of the most demanding regulatory environments.

From 37,500 tables to a unified truth: The data architecture behind AI

Back in 2021, Thomson Reuters chose Snowflake because the platform was able to combine enterprise-wide governance and security requirements with a scalable data infrastructure. Since then, the company has built a single, secure source of truth across more than 37,500 controlled tables and 350 different data sources. This foundation powers the internal My Data Space platform, through which central teams deliver and share trusted data products across the organization.

More than 1,500 internal users, including data engineers, analysts, and executives, now access this system daily to retrieve protected data and gain insights for their workflows. Particularly noteworthy is the acceleration of critical workloads by up to 3.4 times, enabling complex data analyses that previously took weeks to be completed in seconds. This leap is not simply a matter of computing power, but rather the result of a consolidated data pipeline that eliminates manual data preparation, allowing teams to focus on decision-making instead of the underlying data logistics.

Cortex AI and CoCo: The two tools for speed and modernization

At the heart of the technical implementation is Snowflake Cortex AI, a fully managed AI service that enables organizations to run large language models directly on their own governance-secured data without having to move it to external systems. For Thomson Reuters, this means that complex regulatory datasets are transformed into real-time insights, significantly accelerating the development of intelligent applications. Cortex AI supports models from various vendors, including Anthropic, Meta, and Mistral, as well as current models like Claude Opus 5 and Gemini 3 Pro, which run directly within the secure Snowflake environment.

In parallel, Thomson Reuters is using Snowflake CoCo, an AI-powered programming assistant that helps teams modernize legacy systems and accelerate their migration to the Snowflake environment. By simplifying development within a controlled environment, teams can scale AI and data innovations without compromising security or compliance standards. This combination of analytics tool and development assistant exemplifies how managed AI works in practice: The infrastructure, model selection, security policies, and scaling are managed by the platform provider, while the business unit retains content control over its data and use cases.

Fiduciary-Grade AI: A new benchmark for fiduciary responsibility in AI

The term "fiduciary AI," used by Snowflake and Thomson Reuters, describes a level of quality where accuracy, traceability, and defensibility of results are non-negotiable. Caitlin Halferty, head of data and analytics at Thomson Reuters, emphasized that the real value lies not just in speed, but in the ability to innovate in a controlled environment where teams can transform complex regulatory data into actionable insights while maintaining the level of trust, control, and reliability required for professional, high-risk applications.

For a company whose core products are used by lawyers, tax advisors, and regulatory experts in business-critical situations, this approach is not a marketing slogan but an economic necessity. Erroneous or undocumented AI expenditures in an area where millions of dollars or legal liability are at stake would jeopardize the entire corporation's business model. Snowflake's governance layer, which includes role-based access controls, data masking, object tagging, and audit logs, provides the technical foundation, while additional filtering features like Cortex Guard help prevent inappropriate or harmful model outputs.

Managed AI Platform - Image: Xpert.Digital

Here you will learn how your company can implement customized AI solutions quickly, securely and without high entry barriers.

A managed AI platform is your all-inclusive, worry-free solution for artificial intelligence. Instead of dealing with complex technology, expensive infrastructure, and lengthy development processes, you receive a ready-made solution tailored to your needs from a specialized partner – often within just a few days.

The key advantages at a glance:

⚡ Rapid implementation: From idea to ready-to-use application in days, not months. We deliver practical solutions that create immediate added value.

🔒 Maximum data security: Your sensitive data stays with you. We guarantee secure and compliant processing without sharing data with third parties.

💸 No financial risk: You only pay for results. High upfront investments in hardware, software, or personnel are completely eliminated.

🎯 Focus on your core business: Concentrate on what you do best. We take care of the entire technical implementation, operation, and maintenance of your AI solution.

📈 Future-proof & scalable: Your AI grows with you. We ensure continuous optimization and scalability, and flexibly adapt the models to new requirements.

More information here:

• Managed AI Platform

Westlaw and CoCounsel as beneficiaries: Practical implications for flagship products

A key effect of data consolidation is immediately apparent in Thomson Reuters' two flagship products. As Thomson Reuters merges its data pipelines for Westlaw, the established legal research platform, and CoCounsel, the generative AI assistant for lawyers, core workloads now run up to 3.4 times faster, enabling the transition from static reports to near real-time insights. This shift fundamentally changes how legal professionals work on a daily basis: Instead of waiting for daily updates, users now receive continuous, verified information on regulatory developments.

Christian Kleinerman, Executive Vice President of Product at Snowflake, emphasized that Thomson Reuters demonstrates how companies can scale AI and governance together by creating a trusted foundation that enables teams to work faster and scale AI across the enterprise. This statement underscores the strategic dual function of the partnership: it is both a technological modernization project and a tool for building trust with a customer base that is traditionally very risk-averse when it comes to technological innovation.

Content as a raw material: Reuters opens its archives to the AI ​​economy

In addition to its internal data platform, Reuters announced in August 2026 that it would make its journalistic content available through the Snowflake Marketplace, enabling companies to enrich their AI applications and models with fact-based Reuters news and multimedia content. This marketplace provides companies with access to Reuters journalism dating back to 1987, available in five languages ​​and directly integrated into AI workflows. This opening marks a significant step for the entire media industry, demonstrating how protected, copyright-backed journalistic archives can become a monetizable resource for generative AI systems without relinquishing control over usage rights.

This positions Thomson Reuters simultaneously as a provider of trusted training and contextual data for third-party AI systems and as the operator of its own governance-compliant AI applications for its core customers. This dual role fundamentally distinguishes publishers and specialist information providers from pure technology companies, as they possess exclusive, often decades-old datasets that represent a significant qualitative advantage for generative AI models compared to publicly available internet data.

Snowflake's platform strategy: From data warehouse to control layer for enterprise AI

The collaboration with Thomson Reuters aligns with Snowflake's broader strategic direction, as the company increasingly positions itself as a control layer for the agent-based enterprise world. In April 2026, Snowflake announced significant enhancements to Snowflake Intelligence and Cortex Code, designed to allow organizations to connect even more data sources, enterprise systems, and AI models with their trusted Snowflake data within a unified experience. Snowflake Intelligence acts as a personal work assistant for business users, adapting over time to individual preferences and workflows, while Cortex Code serves as the development layer for enterprise AI, enabling data-driven development across the entire ecosystem.

Since its launch in November 2025, this tool has seen rapid adoption, with more than half of customers actively using it to boost productivity. Snowflake is pursuing a similar approach in the financial services sector with Cortex AI for Financial Services, which combines rigorous security and compliance controls with access to partner data from providers such as FactSet, MSCI, and the Associated Press. This parallel between the financial and legal sectors highlights an industry pattern: wherever errors are costly and traceability is essential, the governance-first AI model prevails over open, uncontrolled approaches.

Managed AI as a business model: Why management is becoming more important than model choice

The term Managed AI describes a fundamental shift in how companies use artificial intelligence. Instead of operating individual language models in isolation, a company like Thomson Reuters transfers significant parts of the infrastructure, security architecture, model selection, and operations to a specialized platform that handles these tasks at scale and with continuous updates. The real value is no longer created solely through training or adapting individual models, but through the reliable management of the entire data and AI lifecycle: from data acquisition and governance to the delivery of results in production applications.

For specialist publishers, law firms, consulting firms, and other knowledge-intensive industries, this means that competitive advantage increasingly stems not from access to AI models themselves—these are now largely standardized and accessible to many providers—but from the quality, exclusivity, and governance of the underlying data. Thomson Reuters demonstrates this impressively: The real competitive advantage lies not in using Claude, Gemini, or other language models per se, but in the ability to use these models securely, transparently, and legally compliant on highly specialized datasets built up over decades.

Competitive dynamics in the specialist information market: Acceleration as a differentiating factor

Accelerating workloads by up to 3.4 times allows Thomson Reuters to provide its end customers—lawyers, tax experts, and regulatory analysts—with critical, up-to-the-minute information significantly faster than before. In a market where competitors like LexisNexis, Bloomberg Law, and specialized legal tech startups are also introducing generative AI tools, speed combined with reliability is becoming a crucial differentiator. The first provider to deliver robust, verified answers gains the trust of a professional group that is traditionally extremely conservative with new technologies.

At the same time, this development creates a structural advantage for established information providers with extensive, historically grown, and legally vetted datasets compared to new market entrants who, while possessing modern AI technology, lack comparable data depth and reputation. The collaboration with Snowflake thus also acts as a kind of market entry barrier: While a smaller company could theoretically use the same cloud infrastructure, it lacks the decades-long, curated data trove that constitutes the actual substantive content of AI applications.

Risks and open questions of governance-based AI scaling

Despite the positive portrayal in the official press releases, several questions remain unanswered that are crucial for a balanced assessment. Neither Thomson Reuters nor Snowflake publish detailed information on the actual error rates of the AI ​​systems used, the frequency of human review steps, or the specific implementation costs. The metric of 3.4x acceleration explicitly refers to certain critical workloads and not to all use cases within the company, which limits its significance for a general evaluation of success.

Furthermore, the question of dependence on a single cloud provider remains relevant: A company that bases its entire data infrastructure and AI governance on a platform like Snowflake enters into a significant strategic commitment that could have noticeable consequences in the event of price increases, technical disruptions, or strategic realignments by the provider. The question of how generative AI models handle conflicting or outdated legal sources within the massive dataset of 37,500 tables also remains unanswered in the public statements and is likely to pose a continuous challenge in practical application.

Governed Data as a new core competency of the media industry

The partnership between Thomson Reuters and Snowflake provides an insightful model of how traditional media and specialized information companies can remain relevant in the age of generative AI. Instead of directly competing with the major AI model providers, they are positioning themselves as curators, stewards, and safeguards of high-quality, audited datasets that serve as the foundation for trustworthy AI applications. This pattern is likely to spread to other industries with high regulatory requirements in the coming years, such as healthcare, financial supervision, and the public sector, where the origin, auditability, and control of data are at least as important as the sheer computing power of the models used.

In the long run, the success of managed AI strategies will not depend on which language model is the most powerful, but rather on how consistently companies manage, maintain, and prepare their data for productive use. Thomson Reuters, through its multi-year investment in a unified, controlled data architecture, has built a structural advantage that is now translating into measurable speed and efficiency gains and simultaneously serves as a blueprint for the entire professional publishing and information industry.

Snowflake and Unframe.AI both pursue the goal of providing enterprise AI securely and in compliance with governance, but differ fundamentally in their business model and target group: Snowflake is primarily a data infrastructure platform, Unframe is a fully managed solution delivery.

Core business model and approach

Snowflake originated as a cloud data warehouse platform that has evolved into an "AI Data Cloud," providing companies with the infrastructure, governance tools, and AI building blocks like Cortex AI to build their own AI applications, either independently or through partners like Thomson Reuters. Unframe on the other hand, explicitly describes itself as a "Managed AI Delivery Platform." Instead of providing DIY tools, it delivers ready-to-use, production-ready AI solutions for specific business problems within days, configured from pre-built, modular "building blocks" according to a blueprint architecture. With Unframe , the customer essentially defines the desired outcome, while the company handles the technical implementation, operation, and ongoing optimization.

Similarities

Both providers emphasize a high level of governance, security, and traceability as a key selling point for large enterprises seeking to deploy AI without losing control over sensitive data. Both are also model-agnostic, meaning they don't lock customers into a single language model but allow the use of various LLMs like Claude, Gemini, or GPT within the same platform. Both also pursue the principle of unifying data from diverse source systems and preparing it in a contextualized manner for AI applications: Snowflake via its central data platform, Unframe via its "Knowledge Fabric.".

Key differences

When might Unframe.AI become more interesting?

Unframe.AI is likely to be the more attractive choice, especially for companies that need fast, measurable results for specific, clearly defined use cases without having large in-house data engineering or AI development teams, as the platform explicitly advertises implementation in days rather than months. This is particularly relevant for organizations that already have distributed, historically grown IT systems such as SAP, Salesforce, or legacy databases and want to integrate them into AI workflows without complex migrations, since Unframe explicitly uses pre-built connectors for such systems instead of requiring the creation of a central data platform like Snowflake.

Unframe 's outcome-based pricing model also makes it attractive for companies hesitant to make large upfront investments, as they typically only pay once a solution has proven effective. Case studies such as the 45% reduction in mean time to response at a Fortune 500 insurer or the threefold faster claims settlement at a global insurance provider demonstrate that this approach is particularly well-suited for operational, process-oriented use cases with clearly measurable key performance indicators (KPIs). However, as soon as a company wants to build and maintain control over a comprehensive, enterprise-wide data infrastructure with thousands of tables and data sources, as was the case with Thomson Reuters, a platform like Snowflake remains the structurally more appropriate, albeit significantly more time-consuming, solution.

Konrad Wolfenstein

I would be happy to serve as your personal advisor.

You can contact me at wolfenstein∂xpert.digital or

Just call me on +49 7348 4088 965 .

LinkedIn