Xero and Anthropic have announced a multi-year partnership aimed at integrating the Claude large language model (LLM) into Xero’s global small business platform. This collaboration seeks to transition accounting from traditional manual processes to a fully automated and intelligent workflow.
In an exclusive conversation with The Fintech Times, Diya Jolly, Chief Product and Technology Officer at Xero, discussed the significance of this partnership. This initiative exemplifies a broader trend within the industry toward integrating financial tools directly into third-party AI systems, enabling users to manage payroll, payments, and tax obligations using natural language.
This strategic move follows Xero’s prior use of various models, including those developed by OpenAI and Gemini, which achieved a commendable 97% accuracy rate in bank reconciliations. Through the collaboration with notable AI developer Anthropic, Xero aims to address the remaining 3% of complex financial tasks, including handling split invoice payments and processing unstructured data that traditional ledgers find challenging.
Advancing Beyond Structured Data
Traditionally, accounting software has operated predominantly with structured data to ensure the accuracy of ledgers. Jolly emphasized that the partnership with Anthropic empowers Xero to analyze unstructured data effectively. This feature enables small businesses to upload various data sources, such as spreadsheets or legacy payroll information, directly linking them to their financial records.
“Accounting has long existed within confined web applications,” Jolly stated during the interview. “The introduction of natural language opens up the possibility of making accounting accessible wherever users find themselves.” This transition toward “headless apps” implies that financial data will no longer be restricted to a single interface, allowing interrogation through Claude.ai or other communication channels.
The Emergence of Agentic Workflows
The partnership will incorporate a reasoning layer into Xero’s financial infrastructure via its AI superagent, JAX (Just Ask Xero). Distinct from standard automation, which operates according to rigid protocols, agentic AI can execute intricate tasks from inception to completion. For instance, a small business owner may request Claude to identify a supplier, assess real-time cash flow for affordability, and generate an invoice or payment seamlessly within the AI chat interface.
This evolution raises concerns regarding the future of the bookkeeping profession. While some fear that AI may replace human jobs, Jolly clarified that the technology is intended to remove repetitive tasks rather than eliminate human oversight. “Bookkeepers will transition to roles that focus on advisory rather than performing mundane tasks,” she remarked, noting that while AI excels in automating repetitive actions, it lacks the contextual understanding of an organization’s unique needs.
Balancing Deterministic Records with Probabilistic AI
A key challenge in integrating LLMs with finance lies in reconciling the probabilistic nature of AI with the strict demands of accounting. Xero functions as a system of record where results must consistently meet deterministic criteria, whereas LLMs employ predictive algorithms, which may yield variable outcomes.
To address this, Xero positions itself as a reliable anchor. The platform operates as a financial operating system, managing bank data and compliance regulations, while Claude serves as an analytical layer. Jolly stressed that for crucial tasks such as tax filings, human expertise is indispensable for review and accountability.
Addressing Security and Data Sovereignty
A fundamental concern for financial institutions and small businesses adopting AI technologies is data privacy. The partnership is structured to ensure that financial data stays within the Xero ecosystem. When users access their financials through Claude, the LLM interacts with a Xero Model Context Protocol (MCP) server.
This configuration guarantees that proprietary business information is not utilized to train Anthropic’s models. Instead, Xero offers an applet within Claude, safeguarding the data exchange. This “trusted intelligence” model aims to provide advanced reasoning while preserving the security of financial records.
Preparing for Future Growth
As small businesses evolve from lean operations to larger enterprises, the need for integration becomes increasingly important. Xero plans to remain an infrastructural layer that connects to various point-of-sale (POS) systems, CRM software, and payroll providers. While Claude delivers analytical capabilities, Xero continues to provide the API surface and connections to a vast array of ecosystem applications.
Looking forward, Xero’s roadmap includes plans to enhance its presence across multiple interfaces, potentially integrating with popular accounting tools such as spreadsheets and office applications. The objective is to advance toward a more “default automated” state for book closing and tax preparation, leveraging the strengths of diverse AI models to tackle specific tasks.
The partnership between Xero and Anthropic represents a pivotal advancement in the fintech sector’s embrace of agentic AI. By merging the extensive datasets of a global accounting platform with the analytical capabilities of a leading LLM, this collaboration aims to empower small businesses with financial intelligence typically accessible only to firms with dedicated CFOs or financial analysts.