Pigment raises $250 million for AI business planning

Pigment announced a $250 million Series E on October 21, 2026, at a $2.3 billion valuation. Scaleup Europe Fund, managed by EQT, led the round. NVIDIA participation remains planned, rather than a completed investment.

The business planning company intends to expand engineering and research around its Graphite computing engine and AI agents. Its product roadmap reaches beyond generating reports toward running planning processes from forecast through action.

How Modeler changes a planning application

The most concrete starting point is Pigment’s Modeler Agent. A user describes a change in ordinary language, and the agent examines the existing application before proposing a plan. The user reviews and approves that plan inside Pigment before the agent applies the agreed changes. This puts an approval step between an instruction and a changed planning model.

Pigment says Modeler can create or revise dimensions, metrics, formulas, reporting views and boards. The company leaves business design, validation and governance with human modelers, and says changes are documented through its existing versioning and governance processes.

Analysis and interfaces have separate jobs

Pigment describes its Analyst Agent as the reporting side of that workflow. It retrieves data from a model, investigates trends or deviations and turns a conversation into a report. Reports can be shared or scheduled as recurring tasks. The product page also describes code execution for analysis, using the model’s metrics and business logic. Those are vendor descriptions, not independently measured performance results. Its page labels the trial as beta, so buyers should confirm access and feature availability.

A separate feature, Frames, builds custom interfaces over the live model. Pigment lists forms, executive views and interactive scenario controls among the uses. Its product page says a new or revised Frame stays visible only to its author until publication, and creation requires the Configure Blocks permission. Published interfaces retain each user’s existing data access rights.

That distinction matters operationally. An interface can collect or present planning inputs without granting its audience unrestricted access to the underlying model. A deployment review should check both the interface’s publication state and the permissions of the people expected to use it.

Shared data is part of the product

Pigment’s platform description puts Graphite beneath these tools. The company says the engine supports people and agents working concurrently, with shared definitions for metrics, dimensions and organizational hierarchies. It also describes a common data layer connecting business systems, including finance, customer management and human resources software. Graphite is described as patent pending. These architectural descriptions come from Pigment.

The practical issue is consistency. If different departments define the same measure differently, an attractive report does not resolve the disagreement. A shared model gives teams a place to define those measures and their relationships, while leaving them responsible for the assumptions and source data they put into it.

The Pigment MCP Server extends that model to outside assistants. Pigment says tools including Claude, Cursor and Codex can query data and build or edit models through the connection. In the other direction, Agent Connectors let Pigment agents read context from systems such as Notion, Google Drive and Slack.

According to Pigment, administrators approve connected assistants and source systems, while existing user permissions limit access. Pigment describes authentication through enterprise identity and access scopes tied to roles, models or individual metrics. That makes the setup more consequential than attaching a file to a chat. A connected assistant may work with a live model. Buyers should establish which operations a connection can perform and test permissions with the roles they intend to deploy. Read access and model editing warrant separate checks, even when both connections use the same identity.

Google’s proposed contextual privacy checks for agents address a related question of what data a task justifies sharing.

Planner remains on the roadmap

As of October 21, Pigment’s homepage still marks Planner Agent as coming soon. It describes scenario simulation and recommendations, without specifying a general release date. The current product descriptions should not be read as proof that an agent already runs every stage of a company’s planning cycle.

In their funding post, cofounders Eléonore Crespo and Romain Niccoli describe autonomous planning as a goal that combines building forecasts, running scenarios, updating plans and executing within team rules. Planner is the next step they identify.

The investor context also reaches beyond one product. EQT says Scaleup Europe Fund is backed by the European Commission and institutional investors, with a €5 billion fundraising target. Its remit spans Series B through the period before a public listing, including AI and other technology sectors. That figure is the fund’s target, not Pigment’s funding or a confirmed final fund size.

For buyers, the useful next step is to evaluate the documented workflow against a bounded planning task, with explicit owners for approvals and validation. ByteForward has not tested these products or verified their accuracy, time savings or financial returns. Product access and deployment requirements also need confirmation for an individual customer’s setup.

Illustrative workplace planning photograph by Helloquence, via Wikimedia Commons, under CC0 1.0.