Thomson Reuters
Randy Bean
How is AI transforming the business of information at Thomson Reuters?
This is the question I posed to Caitlin Halferty, who serves as global head of data and analytics for this multinational information and technology company whose core businesses comprise legal services, tax and accounting, financial data and risk, corporate compliance and risk, and news and media.
Thomson Reuters came into existence in its current structure in 2008 when Thomson Corporation, a Toronto-based professional information company founded in the 1930s, acquired Reuters Group, a news and media company founded in London in 1851. Thomson, which had grown from a media company into a specialist in professional information—legal, tax, and financial data—had evolved by the early 2000s into a leading provider of subscription-based information tools for professionals.
Reuters had been a pioneer in supplying news to newspapers, governments, and financial markets worldwide. The Thomson Reuters combination created one of the world’s largest providers of financial data, legal research, and news. Today, Thomson Reuters is an $8 billion business that operates as a leading provider of legal research platforms such as Westlaw, a provider of tax and accounting software, and is the parent company of Reuters news.
In her role as global head of data and analytics at Thomson Reuters, Halferty mandate is to make sure that the company’s data is a source of business advantage, and that AI across the enterprise meets the standard that Thomson Reuters customers hold the company to. Halferty brings deep data and AI leadership expertise to her current role. She previously served as global chief data officer at Ericsson, the Swedish telecommunications company with $23 billion in annual revenue, and as a founding member of IBM’s first global chief data office.
Thomson Reuters occupies a rare position, in combining over 175 years of trusted, curated, and professionally validated content with modern AI to help professionals in law, tax, compliance, and finance make better decisions.
MORE FOR YOU
As head of global data and AI, Halferty’s mandate is to transform how data and AI are used across the organization — from commercial performance and customer experience to product intelligence and financial clarity – and to ensure that every data and AI system meets the standards that Thomson Reuters customers expect. She explains, “This means outputs that are traceable to authoritative sources, transparent in how answers are generated, and accountable in ways professionals can verify — consistent with what we describe as a fiduciary-grade approach to AI in high-stakes professions.”
A central component of Halferty’s mandate is what she calls, “democratizing access to insight” — ensuring that trusted, governed intelligence is available to everyone who needs it, not just specialists. She explains, “Through natural language interfaces connected to our semantic layer, employees across functions can ask complex questions and receive governed, auditable answers in real time — without needing specialized technical support.” Halferty adds, “Our semantic layer and conversational AI systems are key enablers of this shift.”
“Increasingly, AI is no longer a feature layer on top of our products — it is becoming inseparable from the expertise itself and embedded directly into the workflows professionals rely on,” notes Halferty. She continues, “Thomson Reuters data and AI applications share a single purpose, which is to enable better outcomes for the professionals who depend on the company’s wide range of services.” Halferty adds, “Success is measured by whether the business is faster, smarter, and more responsive as a result.”
Data, analytics, and AI operate across two reinforcing dimensions at Thomson Reuters:
The applications that matter most reflect this shift. For example, agentic AI systems can now execute multi-step analysis, a semantic layer that gives every business decision a common, trusted foundation, while natural language tools make enterprise insight accessible to anyone who needs it.
Halferty comments, “These systems are designed to anticipate needs and surface insights before they are explicitly asked.” She continues, “That’s why every AI system we build is grounded in governed data, transparent logic, and the ability to stand behind every answer it produces — what we think of as a fiduciary-grade standard for AI in high-stakes environments.” Halferty concludes, “The unifying principle across all of this is trust. Our customers are professionals who cannot afford to be wrong.”
Trusted data infrastructure is one of Thomson Reuters most durable competitive advantages in an AI-first world. “We think about data the way we think about our content: it must be curated, governed, and trusted before it can deliver value,” states Halferty. “That same standard applies to every AI system we build — ensuring outputs are reliable, explainable, and fit for professional use.”
Halferty notes that the clearest expression of this strategic conviction is the company’s enterprise semantic layer, which is a shared, authoritative foundation that governs how core business concepts like revenue, retention, and customer health are defined and used across the organization. She explains, “It enables consistency at scale, so decisions are faster, clearer, and grounded in a common understanding of the business.” Halferty adds, “While many organizations can access AI models, far fewer have AI-ready, governed data at scale.”
This data foundation is what makes more advanced capabilities possible across Thomson Reuters — from agentic systems that can reason across data, to AI that can reliably support decision-making in high-stakes environments. Halferty notes, “We are already seeing the impact: data refresh cycles reduced from 24 hours to under 2 hours, complex analyses completed in minutes instead of weeks, and manual data preparation eliminated across key workflows.” She adds, “These are not just efficiency gains — they are what enable AI to operate as a true decision-support system across the enterprise.”
Thomson Reuters most impactful data initiatives connect data quality, AI-powered intelligence, and direct workflow integration. Halferty explains that fragmented customer data creates friction across the organization. How customer data transformation at-scale addresses this is illustrated by how the company is using AI-driven matching and enrichment to transform the quality of the customer data foundation within months. The result is that 9M+ customer records have been enriched, and 98% segmentation coverage achieved. Halferty comments, “These improvements translate directly into better outcomes — more accurate customer insights, stronger engagement, and a more reliable foundation for downstream AI systems.”
The creation of a Finance Semantic Layer represents a major step forward in how Thomson Reuters operates as a data-driven organization. “By creating a single, governed model across financial and commercial data, we eliminated ambiguity and enabled real-time, consistent analysis across teams,” notes Halferty. The result, she explains, is 1,500+ daily users operating from a shared data language, weeks of reconciliation reduced to minutes, and support for external financial reporting, meeting the highest standards of reliability. She adds, “Across both examples, the common thread is moving beyond isolated tools toward systems where governed data, AI, and workflows are tightly integrated — enabling faster, more confident decision-making.”
Thomson Reuters measures the business value of its investment in data, analytics, and AI in speed, efficiency, and commercial outcomes. Halferty comments, “The standard is straightforward: does this improve a decision, strengthen a customer outcome, or create a measurable business advantage?” Key indicators include faster delivery of actionable insight to decision-makers, elimination of manual processes across workflows, and improved customer retention through earlier risk identification and proactive engagement. “Ultimately, the value of AI is measured not just in efficiency gains, but in how it enhances the quality, speed, and impact of decisions across the business,” says Halferty.
Critical to this success is the establishment of a data, analytics, and AI business culture. “Our approach is to eliminate the distance between data and decision-making,” says Halferty. “Intelligence — from customer health signals to performance insights — is delivered directly into the tools our teams already use, so AI becomes part of how work gets done, not something separate from it.” She continues, “Nothing builds culture faster than visible impact. When AI-powered customer intelligence was deployed across multiple products, teams gained immediate access to patterns, root causes, and emerging trends.” Halferty concludes, “This ability to understand customer needs at scale — and act on them — changes how people think about data and AI and accelerates adoption.”
Looking ahead, Thomson Reuters is focusing on development of an intelligence layer where governed data, AI, and professional expertise are inseparable. Halferty explains, “We are building toward a Decision Intelligence Layer — an architecture where governed data, semantic understanding, and AI systems work together to deliver trusted, explainable insight at the moment decisions are made.” She adds, “The goal is not better reporting — it is fundamentally changing how decisions happen.”
By expanding agentic AI capabilities, systems can not only answer questions, but autonomously explore data, surface patterns, and generate analysis. This allows AI to move beyond assistance toward actively accelerating complex work. On the customer side, Thomson Reuters is building a “Customer 360”, which integrates product usage, support, and engagement data into a unified view. As Halferty explains, “This will enable AI to anticipate needs, identify risk earlier, and support more proactive, personalized customer experiences.” Halferty notes, “The broader ambition is clear: AI that anticipates, reasons, and acts — while maintaining the trust, transparency, and accountability required in professional environments.”
Underpinning all these efforts is a commitment to responsible AI as a product and cultural standard. Halferty explains, “The professionals we serve operate in environments where the cost of error is significant. That shapes how we build AI: outputs must be traceable, logic must be transparent, and data must be governed end-to-end. Responsible AI is not an overlay — it is built into the design of our systems.”
Within Thomson Reuters, every AI use case goes through a formal Data Impact Assessment before reaching users. The company has scaled this process alongside innovation, ensuring rigor without slowing progress. The organization has also made external commitments to transparency and accountability, including published disclosures and ISO 42001 certification across flagship products like CoCounsel and Westlaw.
Halferty comments, “We are building an organization where people not only use AI confidently, but understand the principles behind it — including transparency, accountability, and the ability to stand behind every output.” She adds, “When governance is embedded and visible, it becomes a foundation for innovation — not a constraint. Ultimately, responsible AI is what enables us to move faster.”
Inevitably, the most powerful AI organizations are ones where human expertise and AI work together and leadership approaches AI as a cross-enterprise capability tied directly to business outcomes — from customer understanding to commercial performance and product innovation.
Halferty concludes, “We have a strong bias toward action: applying AI to real business challenges quickly, learning from results, and scaling what works. At the core is a clear principle: AI is most powerful when it augments human expertise and enables professionals to make decisions they can stand behind. The goal is not automation for its own sake, but better decisions made by better-equipped people.”
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。