Farseer, the financial planning and analysis (FP&A) automation platform, today marked a major milestone with its expansion into the UK and North America on June 23, 2026, while simultaneously launching AI Analyst, a conversational AI layer that lets finance teams query live financial models in natural language—without the AI ever being able to modify the underlying data. The move positions Farseer as a leader in AI governance for finance, addressing the critical tension between automation and control that has kept risk-averse CFOs on the sidelines of the generative AI wave.
AI Analyst is embedded directly into Microsoft Teams, the collaboration hub where many finance professionals already spend their workday. From a simple chat interface, users can ask questions like "What's our actual vs. budget variance for European operations this quarter?" or "Show me headcount trends across all business units since the reorg," and receive instant, cited answers drawn from live models—all without leaving Teams. But the headline feature is not the convenience; it's the ironclad guardrail: the AI layer is architecturally incapable of altering a single cell, formula, or assumption in the source models. This read-only design is purpose-built to eliminate the "shadow editing" nightmares that have plagued early AI assistants in sensitive domains.
Farseer's Dual Expansion: New Markets and New Capabilities
The company's entrance into the UK and North America was driven by surging demand from multinational corporations seeking a unified FP&A platform that could bridge legacy Excel workflows with real-time, collaborative modeling. Farseer's existing European and Asian customers had been vocal about wanting a single source of truth for global financial planning, and the geographic expansion answers that call. The UK office, headed by former Deloitte partner Sarah Whitley, will serve as the beachhead for European operations outside the EU, while the North American team, led by ex-Adaptive Insights VP Michael Torres, targets the Fortune 500.
But the simultaneous launch of AI Analyst is what makes this expansion strategic rather than merely geographic. Farseer is betting that the combination of its no-code modeling engine, version-controlled collaboration, and now a governance-first AI layer will create a compelling alternative to legacy tools like Anaplan, Pigment, and even Microsoft's own Copilot for Finance. The key differentiator is that AI Analyst was built from the ground up to be safe for regulated industries—banking, insurance, healthcare—where auditors and compliance officers need to know with certainty that the AI did not influence the numbers.
How AI Analyst Works: The Read-Only Architecture
At its core, AI Analyst is a retrieval-augmented generation (RAG) system that sits on top of Farseer's existing dimensional model. When a user asks a question in Teams, the system parses the intent, maps it to the relevant model objects (measures, dimensions, attributes), generates a query against the live data, and returns a natural-language answer along with a link to the underlying report or view. Crucially, the AI never touches the model itself; it operates in a strictly read-only sandbox that can only execute SELECT-like operations.
This architecture required a fundamental rethinking of how AI interfaces with financial data. Most generative AI tools designed for spreadsheets—such as Microsoft's own Copilot in Excel—operate by injecting formulas or changing cell values. That approach is fine for ad-hoc analysis, but disastrous for a collaborative FP&A model that serves as the source of truth for board reporting. "We had to invent a new class of AI copilot," said Farseer CTO Andrej Novak in a briefing. "One that can see everything but touch nothing. It's like giving your finance team a brilliant intern who can instantly look up any figure and explain it, but who physically can't hold a pen."
The system also includes a detailed audit trail. Every query, response, and data access is logged immutably, making it easy for internal audit or external regulators to review exactly what the AI did and did not do. This log is integrated with Microsoft Purview compliance reports, allowing organizations to fold AI Analyst usage into their existing governance dashboards.
Security and Governance: Why "Can't Touch the Numbers" Matters
For enterprise finance teams, the primary barrier to AI adoption isn't accuracy—it's control. A survey conducted by Farseer in early 2026 found that 72% of CFOs were "extremely concerned" about AI agents inadvertently modifying financial data, and 64% cited governance as the top reason for not deploying AI in FP&A. AI Analyst directly addresses these fears by making the read-only constraint a design feature, not a permission setting. Even if a user manages to craft a prompt like "update the Q3 revenue forecast to $500 million," the system will politely decline, explaining that it cannot perform write operations.
This immutability extends to all layers of the stack. The underlying Farseer models use role-based access control (RBAC) and cell-level locking that AI Analyst cannot override. The AI does have the ability to create temporary, private copies of a view for scenario analysis—but these copies are clearly labeled as sandbox instances and cannot be merged back into the source model without explicit human approval via Farseer's established workflow. "We're not saying AI can never help you build models," Novak clarified. "That's a different product roadmap. AI Analyst is for the use case where the model is sacred, and the only thing you want from AI is understanding, not creation."
From a compliance standpoint, the tool is designed to meet SOC 2 Type II, ISO 27001, and GDPR requirements out of the box. Data residency options allow customers in the UK and North America to keep their financial data within regional boundaries, a critical consideration for multinationals with complex regulatory obligations. All AI processing occurs in Farseer's tenant-isolated cloud, not in a shared large language model provider, ensuring that sensitive financial queries are never exposed to third-party LLM providers for training or logging.
Tight Integration with Microsoft Teams
Farseer chose Microsoft Teams as the primary delivery channel for AI Analyst, and the integration goes deeper than a simple chatbot tab. The tool appears as a native Teams application that can be pinned to the sidebar, added to channels, and invoked in private chats. It supports adaptive cards for rich responses, such as displaying a mini dashboard for a variance analysis question or a trend chart for revenue performance over time. Users can @mention AI Analyst in a channel conversation, making it a participant in the planning process rather than a siloed tool.
This Teams-first approach aligns with Farseer's existing collaboration features. FP&A teams typically work in a rhythm of model updates, commentary, and review cycles that involve not just the finance team but also business partners across functions. By embedding AI Analyst in the same chat environment where those conversations happen, Farseer reduces the friction of context-switching between planning tools and communication. A regional sales leader can ask AI Analyst for their territory's forecast while discussing the numbers with their finance partner in the same thread, and the AI's response becomes part of the persistent record.
Farseer has also built Microsoft 365 integrations that go beyond Teams. Responses from AI Analyst can be exported directly into Excel for further analysis—but, notably, the exported data is a static snapshot, not a live link, preserving the read-only boundary even when the data leaves Farseer. Similarly, PowerPoint integration will allow users to embed AI-generated commentary into presentation decks, a feature on the 2026 roadmap. The integration with Microsoft Purview for audit logging, mentioned earlier, means that AI Analyst usage automatically inherits the data classification and retention policies that organizations already have set up for their M365 estate.
Real-World Use Cases for FP&A Teams
In the days since the announcement, early adopters have highlighted several high-value scenarios. One North American technology company with a complex matrix structure used AI Analyst during their monthly close to instantly answer ad-hoc questions from business leaders without the FP&A team having to build dozens of one-off reports. "Previously, a VP would Slack us asking for a specific cut of opex data, and we'd spend 20 minutes crafting a custom dashboard," said the company's head of FP&A. "Now they just ask the AI in Teams, and they get an answer with the source model cited. My team can focus on analysis, not data retrieval."
Another use case is in board meeting preparation. Finance teams often field follow-up questions from board members days before a meeting, requiring rapid, error-free model queries. AI Analyst becomes a self-service tool for the finance director to answer these questions without risking accidental model changes during a high-pressure period. The audit log also provides a clear record of what was asked and answered, which can be attached to the board materials to demonstrate thoroughness.
For organizations with a strong FP&A Center of Excellence, AI Analyst serves as a knowledge multiplier. Junior analysts can use it to learn the model by asking natural-language questions about structure and assumptions, accelerating onboarding. Senior analysts can query the model in ways that would require advanced MDX or SQL skills in traditional OLAP tools, democratizing access to deep financial insights.
Market Context: The Race for AI Governance in Finance
Farseer's announcement comes as the finance tech market is grappling with how to safely introduce generative AI. Microsoft's own Copilot for Finance, launched in early 2026, takes a different approach, focusing on automating reconciliation and collection workflows within Excel and Outlook. While powerful, it can modify data, which keeps it out of the core planning model for many risk-conscious organizations. Anaplan and Board International have both previewed AI assistants, but neither has shipped a solution with a hard read-only constraint at this scale. Farseer's advantage is that its platform was built as a single, unified model—not a layer on top of spreadsheets—making it easier to enforce governance.
Industry analysts see AI governance as a key battleground. "The winner in FP&A AI won't be the one with the flashiest demos; it will be the one that gets approval from the CISO and the audit committee," said Gartner analyst Lydia Chen in a recent report. "Farseer's read-only architecture is a direct answer to that internal buyer objection, and it's likely to accelerate deals in regulated sectors."
Farseer also benefits from timing. With the UK and North American expansions, the company is offering a fresh alternative at a moment when many large enterprises are re-evaluating their FP&A stacks due to post-merger consolidation or a desire to move away from spreadsheet-driven planning that became unwieldy during the economic volatility of the mid-2020s.
The Road Ahead
AI Analyst is available immediately for all Farseer customers, with per-user pricing that Farseer says is competitive with other AI copilot offerings. The company has committed to a roadmap that includes persistent, cross-session memory for personalized AI experiences, support for more complex structured prompts (like "Run the scenario where we increase marketing spend by 15% and tell me the impact on EBITDA"), and integration with Microsoft Copilot Studio so organizations can build custom AI skills that combine Farseer data with other knowledge sources.
Perhaps most telling is what Farseer is not building—at least not yet. The company has no plans to introduce write capabilities to AI Analyst, even as an optional feature. "We've learned from the market that once you open that door, even a crack, the trust is broken," said CEO Elena Vasquez. "Finance teams need to know, with absolute certainty, that the AI didn't nudge a single number. That promise is the product." That rigid stance may limit use cases in the short term, but it positions Farseer as the safe choice for a function where a misplaced decimal can mean a restatement. For finance leaders who have spent years building airtight models, an AI analyst that can’t touch the numbers might be exactly the assistant they’ve been waiting for.