Summary

• Pigment’s Series E was led by the Scaleup Europe Fund managed by EQT.
• The company says annual recurring revenue exceeds $100 million and agents create 85% of new metrics.
• Investment will extend its Graphite engine and agents from model building towards complete planning cycles.

French enterprise software company Pigment has raised $250 million at a $2.3 billion valuation to develop AI systems that build financial models, prepare forecasts and eventually manage more of the planning process from beginning to end.

The Series E was led by the Scaleup Europe Fund managed by EQT, with participation from World Innovation Lab and existing investors including Greenoaks, ICONIQ, IVP and Meritech. Pigment also described an investment from NVIDIA as planned, rather than completed, in its 21 October announcement.

The company says annual recurring revenue passed $100 million earlier this year, doubling year on year. It reports that AI agents now create 85% of new metrics on its platform, suggesting that customers are using automated functions within ordinary planning work rather than solely running isolated trials.

That percentage measures how new metrics are produced, not whether the resulting forecasts are accurate or whether an entire financial process can operate without human intervention. The company has not disclosed audited productivity gains attributable to the agents, and annual recurring revenue is not equivalent to recognised annual sales.

Pigment names Danone, Siemens, Snowflake, ServiceNow and Unilever among its enterprise customers. It says Anthropic, ElevenLabs and Uber also use the platform, while L’Oréal has selected it to replace salary planning tools for more than 95,000 employees worldwide. A customer selection, however, does not establish that a migration has been completed or quantify its results.

Calculations remain separate from language models

The new financing will support Graphite, Pigment’s proprietary and patent-pending calculation engine. The company says the system can handle hundreds of billions of data points while allowing thousands of people and AI agents to operate on a shared planning model.

Its architecture distinguishes the language model interpreting a request from the engine performing the underlying arithmetic. For an enterprise finance system, this division is important: a model may suggest a method for estimating costs, but calculations must still follow consistent rules and preserve the relationships between assumptions, budgets and reported figures.

Graphite is designed to produce deterministic and traceable results. With the same inputs and model definitions, a calculation should produce the same outcome, allowing finance staff to compare versions and investigate why a forecast changed. This is a stronger foundation for audit than asking a generative model to construct numbers from a free-form prompt.

Pigment’s Modeler Agent already allows users to describe a planning requirement in natural language and generate a model or interface. The company wants a planned Planner Agent to execute planning cycles more comprehensively, progressing beyond assistance with model construction and analysis.

Consider a request to evaluate the effect of slower hiring. An agent could identify the relevant workforce assumptions, update a scenario and request recalculation across payroll, departmental budgets and cash forecasts. The financial engine can maintain mathematical consistency, but the decision to adopt that scenario as an approved plan belongs to authorised employees.

Automation becomes riskier when it moves from preparing alternatives to changing operational commitments. Forecast assumptions may come from incomplete data, departments may disagree about targets, and expenditure decisions can exceed the limits delegated to a particular employee or software service.

A useful autonomous planning system therefore needs permission controls, recorded approvals, change histories and a clear distinction between provisional scenarios and committed financial decisions. Calculations may be technically correct while the assumptions used to generate them are commercially implausible.

Replacing established systems takes more than AI

Enterprise planning commonly spans spreadsheets, finance software and operational applications that were introduced at different times. Teams may use inconsistent naming, hold competing versions of a forecast or rely on manual reconciliation before reporting to senior management.

Pigment is attempting to consolidate those activities around a shared model while agents take on more of the work required to maintain it. The proposition places it in competition with established planning providers and enterprise software vendors whose products are already connected to customers’ payroll, sales and accounting systems.

Replacing an established platform can be expensive even when the new interface is easier to use. Finance teams must validate historical calculations, connect reliable data sources, train employees and satisfy internal audit requirements. An AI agent cannot make those dependencies disappear simply by creating a planning model quickly.

The L’Oréal selection demonstrates the scale of prospective deployment, but the announcement offers no completed performance figures for that programme. Large payroll planning projects can involve country-specific employment rules, currencies, compensation structures and access restrictions, all of which require careful validation before use.

Another question concerns the economics of automation. Agents that repeatedly investigate data, recalculate scenarios or run model evaluations consume computing resources, while administrators must oversee model behaviour and maintain integrations. Savings in manual analysis need to exceed those additional costs to improve the overall business case.

The funding will expand Pigment’s engineering team and support entry into additional markets. Investor participation and the company’s reported revenue growth indicate strong commercial momentum, but the future Planner Agent remains a product objective rather than an independently verified autonomous deployment.

As finance departments consider replacing planning systems, their assessment will depend on more than the speed at which an agent can generate a forecast. They will need evidence that figures can be reproduced, changes can be explained and authority over financial decisions remains with the organisation.