How the Modeler Agent gives stretched teams a new, scalable operating model

If you work in finance, HR, supply chain, or revenue operations in a large enterprise, you’re almost certainly feeling the squeeze: planning cycles are shorter, org structures shift faster, and the model behind your numbers has to constantly adapt.

The bottleneck is usually modeling work: structure, dimensions, calculations, views, governance, explanations, and the iteration loop after leadership feedback. It’s skilled, scarce, and expensive work.

Hiring more staff or reverting to manual spreadsheets aren’t scalable enough for modern business demands. As a result, critical projects often end up stalled in the backlog.

How the Modeler Agent helps

The Pigment Modeler Agent is the closest you can get to having a Pigment Solution Architect on staff 24/7.

It’s an extra teammate that helps you construct and refine custom business planning applications by simply asking in natural language.

The Modeler doesn’t replace your team. It requires your review and validation before it makes changes. But it does make your whole team more effective. 

It targets a critical phase of enterprise planning: the stretch where ideas slow down as design complexity increases and struggle to translate into actionable models.

Let’s look at an example of how you might put it to use.

Example: Shifting sales strategy with the Modeler

Imagine you’re David, an FP&A Manager whose company is shifting from a one-shot sales model to a subscription-based business.

The existing model was simple: units × price. But that’s changing.

David tells the agent what he needs:

• "I need to move from revenue recognition at sale to recurring revenue with monthly and annual recurring revenue."
• "I want to model acquisition, expansion, churn, and cohort behavior over time."
• "I do not want to rebuild everything from scratch, just extend the model cleanly."

With Pigment Modeler Agent, the workflow looks like a guided build with a human solution consultant:

1. The agent proposes structured changes aligned to the intent, explaining the reasoning behind its decisions
2. David reviews the proposal and, for example, adjusts the churn logic to reflect different retention assumptions for enterprise vs SMB segments before validating the plan
3. The Modeler applies changes, keeping a record of what was proposed, what was approved, and what executed, ensuring traceability

That’s just one example. There are plenty of other ways David could use the Modeler:

Modeling debt cleanup
Refactoring a legacy model that accumulated workarounds and structural shortcuts over time, without losing historical continuity.

Structural model extension
Adding driver-based rolling forecast capabilities to a model that was originally built for an annual budget and quarterly forecast.

And the feedback from customers has been extremely positive so far:

• Figma reported that work that used to take hours of designing, modeling, and framework building can now happen in minutes, improving how quickly ideas can be tested with stakeholders.
• Docker implemented a version dimension plan within an hour, a task they said would normally take one to two days of focused effort.
• Carta described a process moving from taking a week to about two days, crediting fewer cycles of trial-and-error, rework, and back-and-forth.

Guardrails and human-in-the-loop

The use of any LLM-driven functionality comes with implicit risks, owing to the non-deterministic nature of the technology. 

That’s why guardrails are so important, and why we’ve invested significantly in building them into the Pigment platform:

1. Permissions are respected: proposals must stay inside what users can do in Pigment, according to their Access Rights
2. Human validation is required for meaningful changes, especially when stakes are high, like modifying a key calculation formula
3. Hallucination risk is constrained by anchoring the agent in model context and rules, not general web and LLM knowledge

Knowledge, skills and tools

Many ‘AI’ features turn out to be a chat window glued to a product. Our design philosophy takes a fundamentally different approach, focused on building reliable, production-grade capabilities rather than superficial interactions. The Modeler Agent leverages a ‘deep agent’ design: structured expertise, scoped capabilities, and auditable behavior.

• Knowledge
  Because it has access to all the relevant contextual data, the Modeler understands your application as if it were the Solution Architect who built it.
• Tools
  Tools allow the Modeler to directly interact with your Workspace - they’re what allows it to take action.
• Skills
  Skills are structured ‘knowledge packs’ that tell the Modeler what best practice looks like in Pigment. We have architecture skills, modeling skills, and use case skills

But if you want more information, you can ask the agent directly:

It’s for this reason that the Modeler can be so much more valuable for business users than general-purpose AI like Claude or ChatGPT.

It understands your business at a deep level, it’s trained specifically on planning best practice, and it’s able to leverage Pigment’s incredibly powerful calculation engine. And it does this while maintaining watertight security and governance.

Next steps

To see the Modeler in action, join the next live tour.