OneStream’s SensibleAI Agents and MCP Agentic Layer are Now Generally Available

Today, finance teams are surrounded by data, but that doesn’t mean they have answers. Insight often lives across spreadsheets, reports, documents, and systems never designed to work together. The result is friction.

We built SensibleAI™ Agents to remove that friction.

OneStream’s SensibleAI Agents are now generally available, bringing intelligent, explainable artificial intelligence (AI) directly into the OneStream platform — where finance already works and trust already exists. Designed to support real decisions under real constraints, this AI operates inside the system of record.

AI That Stays Inside the Lines

Many AI tools sit outside finance platforms. They work on copies of data. They summarize without context. They produce answers that cannot be audited or explained. While those conditions might be acceptable for experimentation, the approach breaks down quickly in enterprise finance.

SensibleAI Agents take a different approach. Rather than being outside the platform, they operate inside OneStream’s governed data models, workflows, and security framework. Every answer is grounded in trusted data. Every insight is traceable down to the source. Every interaction respects the controls finance depends on. Think of it like adding horsepower without losing traction. You move faster, but you stay firmly on the road.

Learn more about the architecture and approach.

From Agents to an Agentic Layer: Bringing Finance to the AI Ecosystem

Finance doesn’t work in one system, and neither does AI.

Today, finance teams are already interacting with AI across tools like Excel and PowerPoint, as well as emerging agentic platforms such as Claude, ChatGPT, and Copilot. But on their own, those tools lack the context that makes financial answers meaningful. They don’t understand how finance data is structured, what a specific account represents, or which business rules apply.

That gap is operational. Without governed access to the financial intelligence that gives every answer meaning, finance teams can’t fully trust, explain, or defend what AI returns to the business.

Introducing the Finance Agentic Layer

This is where the Finance Agentic Layer comes in.

Built on OneStream’s SensibleAI™ foundation, the Finance Agentic Layer uses open standards like Model Context Protocol (MCP) to securely extend OneStream’s governed financial intelligence beyond the platform itself. It creates a consistent, trusted layer that allows any AI agent (native to OneStream or third‑party) to interact with finance data responsibly.

Think of it as meeting finance where AI already lives, without sacrificing the controls finance depends on.

Built for the Agentic Era Without Compromising Trust

All agentic tools that finance relies on need the same foundation:

• An MCP connection that opens governed financial data to the agentic tools finance already uses
• A semantic layer that translates business language into the correct data, accounts, and hierarchies
• Finance‑specific capabilities that allow agents to support real workflows, not just surface generic answers

Accuracy, Auditability, and Governance by Design

In finance, 80% accuracy is effectively 0% useful.

Every output from the Finance Agentic Layer runs on OneStream’s deterministic computation engine. It inherits role‑based access controls, maintains a complete audit trail, and traces each result back to the source data and logic used.

Whether insights are generated inside OneStream or surfaced through a third‑party AI assistant, the outcome is the same: answers that are accurate, explainable, governed, and defensible.

What SensibleAI Agents Do in Practice

SensibleAI Agents analyze structured financial data, interpret unstructured documents, and surface insights that finance teams can explain, defend, and act on. While each agent focuses on a different kind of problem, they all work inside OneStream.

Forecast Agent

Forecast Agent is bundled with the SensibleAI Forecast solution. By providing an intelligent, conversational layer over governed forecast data, the agent helps finance teams trust, review, and operationalize AI-generated forecasts.

Forecast Agent turns forecast outputs, accuracy metrics, and the drivers behind the numbers into clear, business-ready answers grounded in actual forecast data and scoped to user access. During forecast review, teams can ask what's wrong with the latest forecast and the agent surfaces problem areas with severity rankings, accuracy issues, and growth-rate anomalies.

When forecasts miss, Forecast Agent explains why. The agent walks users through the specific business drivers (seasonality, macroeconomic indicators, channel activity, events) that pulled results up or down. As part of that process, the agent highlights what's explained and what isn't, with charts and tables ready for leadership conversations. The agent also helps teams compare forecast versions, understand what changed between cycles, and sanity-check projections before numbers flow into planning.

For forecast owners and administrators, Forecast Agent acts as a copilot for knowing where to focus next. It checks data readiness before a forecast run. It monitors jobs in progress. It flags health issues across the pipeline. It connects errors back to their root cause. Through these aspects, teams are empowered to act faster and catch issues early, not after forecasts are published.

All responses are explainable, auditable, and grounded in governed data. When walking into reviews and stakeholder conversations, users can thus come equipped with answers that can be trusted, defended, and acted on.

Sample Prompts

• "How did our latest forecast do for North America? Anything I should be worried about?"
• "I'm about to kick off the April forecast run for demand planning. Is my data clean and ready to go?"
• "My customer's FP&A lead just asked me why the forecast for Widgets – Northeast is so much higher than last month's forecast for the same period. Can you break that down, so I can walk her through it?"

Finance Analyst

Finance Analyst works directly with the financial model. It allows users to ask questions in natural language and immediately receive structured, governed answers, including reports, calculations, trend analysis, visual outputs, and detailed variance analysis. Analysts can run scheduled or automated, multi‑step analyses in the background, or invoke the same workflows on demand for ad‑hoc reporting, without rebuilding logic each time.

By drilling into detail and pulling data from external systems, the agent produces accurate, comprehensive narratives and variance explanations that stay grounded in OneStream’s governed data. Those insights are available everywhere finance works (including Excel) so analysts can move seamlessly between exploration, reporting, and collaboration without breaking context.

The experience feels less like querying a system and more like working alongside an experienced analyst who already understands your data, assumptions, and structure..

Sample Prompts

• “Explain the primary drivers of gross margin variance versus budget this quarter.”
• “Show revenue growth by product line over the last six quarters.”
• “Why did the contribution margin decline even though revenue exceeded the plan?”

Search Agent

Finance knowledge rarely lives in one place. Instead, policies evolve. Reports accumulate. Context gets buried in PDFs, decks, and documentation. Search Agent brings that knowledge back into reach.

Search Agent scans policies, reports, and OneStream platform knowledge to deliver clear answers with transparent sourcing. You see not only the answer but also where it originated. That makes it easier to trust and easier to share. No more hunting through folders.

Sample Prompts

• “What is our revenue recognition policy for multi‑year contracts?”
• “Summarize the board’s guidance on operating expense growth from last quarter.”
• “How does this account roll up in our consolidation hierarchy?”

Deep Analysis

Some questions require more than speed. They require context, comparison, and judgment. We designed Deep Analysis for those moments.

Deep Analysis connects structured financial data with unstructured documents to surface patterns, drivers, and assumptions. With that data, teams can understand not just what changed, but why it changed and how current results compare to past expectations.

It’s the difference between reading a single number and seeing the full landscape around the number.

Sample Prompts

• “Why did forecast accuracy decline in EMEA this quarter?”
• “Compare current cost trends to assumptions from last year’s plan.”
• “What risks appear repeatedly across our strategic planning documents?”

What SensibleAI Agents Make Possible for Finance Teams

With SensibleAI Agents generally available, finance teams spend less time assembling answers and more time using them. Analysis extends beyond a small group of experts. Planning cycles tighten. Fewer questions stall waiting for “one more report.” The work feels less fragmented and more intentional.

The impact isn’t just efficiency. It’s also momentum.

They are ready for real deadlines, real scrutiny, and real accountability.

Explore what’s possible with OneStream’s SensibleAI Agents today.