Xero Brings Anthropic Claude AI Directly into SMB Accounting

Xero, the accounting software provider for small- and medium-sized businesses (SMBs), launched a live integration with Anthropic’s Claude in May. 

This global integration embeds Claude within Xero and connects real-time financial data directly into Claude.ai, allowing small businesses, accountants and book keepers to access and act on critical financial insights without switching platforms.

By pulling live intelligence on cash flow, profit performance and overdue invoices, users can now surface crucial business signals directly within their existing AI workflows. 

Built on Xero’s reusable agentic foundations, the integration combines advanced financial reasoning with strict data protection, so that proprietary data is never used for model training.

Below, Soon-Ee Cheah, SVP of AI Products at Xero, discusses how this conversation-first integration changes the traditional accounting software model, the technology driving its rapid rollout and what the future holds for AI-powered financial management.

1. To start, why was Anthropic the right partner to bridge the gap between a user’s general workspace and their financial data, and how does this partnership change the traditional model of accounting software? 

Xero has always been about helping small businesses stay on top of their finances. As tools like Claude have become part of our customers’ ways of working, making the right financial information available in those workflows was a natural extension of our approach. 

Our partnership with Anthropic adds an advanced reasoning layer for financial tasks and provides the ability to bring Claude’s AI directly into Xero, and Xero’s financial data and tools into Claude.ai.

The partnership also sits within our broader commitment to a multi-frontier AI model strategy that uses the best tool for a specific job or task, ensuring small businesses, accountants and bookkeepers have access to the most powerful, flexible AI tools available today and can leverage Xero wherever they work.

In integrating with Claude, we took what we’d learned from JAX – our AI financial super agent – and designed a conversation-first experience. It provides precisely targeted excerpts of financial data that progress a conversation, while also leaving links and breadcrumbs back to Xero if the customer wants to engage in non-conversational deep-work. 

So if a small business owner asks Claude ‘How much am I owed?’, Claude will help provide clarity on what the numbers say by identifying unpaid invoices and receivables so they can see who owes them money and what needs follow-up. 

From there, they can move into Xero, going from uncovering strategic financial insights to taking better-informed action.

This changes the dynamic of working with an advisor. Now, small businesses can stop asking for basic numbers and come to them seeing what’s changed and have deeper, more strategic conversations about what it means and what to do next.

2. The speed of this integration was impressive. By building these AI foundations to be reusable from the start, how much easier does that make it for Xero to scale and launch similar tools on other platforms moving forward? 

This integration builds on the same foundational capabilities that our agentic platform JAX uses to run financial analysis, but is tailored toward the feel and experience of conversing with Claude. 

Making these agentic reasoning foundations reusable from day one has given us the flexibility to move quickly in meeting our customers where they want to work.

Providing additional information to LLMs isn’t just a case of squeezing in as much information as possible. Information needs to be curated and structured in a way that makes it easier for the model to understand and deliver meaningful answers to the question being asked. 

Having worked with these technologies extensively with JAX, we were able to use the same strategies to deliver simple but powerful experiences through Claude.

We’ve also been refining our approach to providing information to LLMs to improve their reliability and insightfulness. Those improvements have translated into both know-how, but also reusable infrastructure that efficiently delivers just the right information to LLMs when they need it.

3. Unlike older integrations that used static snapshots of data, this allows Claude to pull live intelligence. From a product perspective, how does having that real-time connection change the types of questions a business owner can now ask their AI?

Customers can engage in conversations with Claude about their profit, cash position, receivables or overdue invoices. 

Claude will instantly pull up answers based on the period they care about such as the month, quarter, financial year or year-on-year reporting. Small businesses will then see the data without opening reports, filtering reviews or exporting data. 

Claude helps pull together the right signals so users can see what needs their attention in the business. 

They can see revenue, profit and cash together to pinpoint whether the issue is sales, collections or liquidity. Or, identify unpaid invoices and receivables to understand who owes you money and what needs follow-up or which customer drives the most revenue. 

By uncovering trends or spotting anomalies within their financial data during a Claude conversation, customers are better equipped to take decisive action when they click back into Xero. 

Whether the data prompts them to send an invoice reminder to a key customer, complete bank reconciliation, or adjust a budget, users can move forward with confidence and take action.

4. With that level of live access, privacy is obviously a huge concern. How do you balance making the AI feel smart and personalised while ensuring a business’s data stays protected and never leaves Xero’s circle of trust? 

Our users’ Xero data is session-based and is not used to train models. In line with Xero’s responsible data use commitments, data responsibility is foundational to the partnership. 

Financial data shared between the platforms is used solely for the user's specific session – proprietary business data is never used to train Claude’s AI models.

5. Finally, looking ahead, how do we get from Claude simply surfacing an overdue invoice to an AI that can actually help a business owner solve the underlying cash flow problem behind it?

Delivering smarter in-context insights is just the beginning of our integration with Claude. 

Over time, customers will be able to orchestrate financial tasks from start to finish across accounting, payroll and payments. 

Powered by Claude’s advanced reasoning, JAX will proactively do the heavy lifting, analysing revenue and profit performance, tracking real-time cash flow, and identifying unpaid invoices to suggest actions for small business owners and their advisors to take.