Pigment connects AI agents to financial and operational planning

Pigment is a business planning platform that connects financial, workforce and operational models with specialist AI agents. Its useful role is to keep a question about the business connected to the assumptions and formulas behind the answer. For a finance team, that means evaluating the explanation of a forecast variance together with the model that produced it, rather than judging a conversational response in isolation.

01 / ProductModels give the agents something concrete to work with

Pigment describes itself as a private enterprise software company focused on business planning and performance management in its official company overview. The scope includes financial, workforce and operational planning. Pigment AI is part of that company offer, rather than a separate provider whose output must be manually pasted back into another planning system.

The current AI documentation identifies Modeler, Analyst, Custom and Documentation agents. Modeler helps create Applications, Blocks and Boards. Analyst explores data and reports; Custom agents add tailored company knowledge; Documentation handles platform questions. Predictions provides machine learning forecasts. Those are different jobs: building a calculation, explaining its result and retrieving instructions should not be treated as interchangeable capabilities.

Pigment's AI product page presents a broader vision of coordinated planning agents. For an implementation, the current application documentation and actual enabled features are the better starting point. A demonstration of a future planning role does not establish that a particular workspace can use it today. Keep the first evaluation anchored to Modeler or Analyst and a model that a finance owner understands.

02 / AudienceA fit for teams whose assumptions cross departmental boundaries

A strong candidate is a finance team whose revenue forecast depends on sales hiring, quota capacity and the time new staff take to become productive. Sales operations owns some of those assumptions, HR maintains another set, and finance needs their combined effect. The value of a shared planning environment is making those relationships visible when one team changes its plan.

A weaker candidate is a small operation that needs occasional spreadsheet summaries but has no recurring planning process or owner for the underlying model. Adopting enterprise planning software solely for a chat interface can create more work than it removes. First identify the decision that currently stalls because definitions, inputs or versions are fragmented.

The Workday blueprint is relevant when HR and financial records already organize the employee process. The SAP blueprint provides a view from an ERP-centered environment. Those systems can remain important sources even when another platform coordinates planning. Compare where actuals originate, where assumptions are approved and where the resulting plan will be used.

03 / WorkflowA proposed hiring-to-revenue forecast review

Consider a proposed review of a sales team's next-quarter forecast after planned hiring is delayed. The finance owner wants to identify which assumptions changed, quantify their effect and prepare an explanation for the operating review. This is an evaluation design, not a report of a Pigment deployment. Use a copy or controlled scenario so that exploration cannot silently replace the approved plan.

Begin with a small, reconciled model: opening sales capacity, planned starts, ramp duration, quota and expected attainment. Define whether dates describe accepted offers or actual starts. Align the time periods and department names before asking an agent to interpret them. An apparently sophisticated explanation is not useful if the model mixes a monthly headcount input with a quarterly productivity assumption.

Ask Analyst to identify the largest differences between the baseline and revised scenario. Require the result to distinguish a changed input from a downstream effect. Hiring delay is an input; reduced productive capacity is an intermediate consequence; forecast revenue is an output. The reviewer should be able to follow that chain without accepting a causal claim merely because the language sounds plausible.

For a modeling change, use Modeler only with an authorized model builder. The AI setup guide requires both Access AI Agents and Configure Blocks for Modeler. This makes role design part of the pilot: a manager who needs an explanation does not automatically need permission to restructure calculations. Review any proposed formula against the planning convention before accepting it.

Compare the generated explanation with a manually reconciled bridge. Include a case where hiring changed but revenue did not because capacity was already sufficient. Include another where a missing input prevents a defensible answer. Those cases test whether the workflow preserves the distinction between observed model changes and a business explanation that still requires judgment.

The agent guide describes chats, attachments and reusable Missions. Start with an interactive review before considering repeated execution. A useful recurring analysis needs an agreed input cutoff and a clear owner for exceptions; otherwise it can repeatedly explain a forecast that is still being edited. An analyst should approve the narrative before it reaches the operating meeting.

Finally, record the time required to verify the result, the changes accepted into the scenario and the AI credits consumed. A faster draft explanation can still be a poor outcome if it sends reviewers through many irrelevant tables. The desired result is a shorter route from a changed assumption to a decision that the model owner can defend.

04 / PricingPlatform licensing and AI consumption are separate dimensions

The license guide says the commercial structure includes a platform fee, license counts and associated features. Permissions determine whether a member is an Explorer, Contributor or Editor. Someone granted model-building permissions can therefore change the licensing picture as well as the security picture. No universal cash tariff was available in the reviewed documentation.

The AI credit FAQ states that Modeler's free trial ended on 9 June 2026; subsequent use consumes shared credits. Do not build a budget from an older free-trial announcement. Credit management describes administrator visibility into quota and consumption by feature and member, with a contact route when the limit is reached.

For the forecast pilot, estimate access roles separately from usage. A small number of model builders may perform complex operations while many managers only inspect results. Ask the account team to map that usage to the proposed plan and clarify how unused allowance, additional credits and contract renewal work. Observed consumption should inform the estimate after the initial evaluation.

05 / DistinctionsThe distinction is the relationship between explanation and model

A planning assistant becomes more useful when its answer can be connected to a maintained model. That creates a practical way to challenge it: alter one assumption, inspect the resulting calculation and ask whether the explanation reflects the actual change. This is more specific than asking a general assistant to comment on a disconnected spreadsheet export.

There is also a separation between the person who understands an output and the person allowed to change the model. Pigment's permission model makes that distinction explicit. An organization can use the pilot to clarify who owns dimensions, formulas and commentary, rather than treating every participant as an equivalent chat user. The benefit depends on keeping those roles intelligible as adoption spreads.

06 / QuestionsCheck the access boundary and changing documentation

Pigment's AI overview updated on 24 September says workspace AI is activated by default unless the organization previously opted out. The older agent guide still describes requesting activation through support. Confirm the actual workspace state instead of assuming either passage describes every account. This is a narrow documentation inconsistency, not evidence that the agents are unavailable.

The AI data-access control also has a specific boundary: it prevents designated agents and MCP from querying a Block's data, but does not govern every structural action or Modeler operation. Review that distinction with the administrator before placing sensitive planning data in scope. A hidden data query and permission to modify a model are separate questions.

Forecast explanations need a business owner. Even a correct arithmetic bridge may miss an operational event absent from the model, such as a territory change or an unusual contract. Require the analyst to mark those additions as business context rather than information discovered by the agent. That preserves a useful record of what the system calculated and what a person concluded.

07 / DecisionStart with a forecast whose assumptions are already understood

Pigment merits a focused evaluation when cross-functional planning already matters and spreadsheet handoffs make change difficult to trace. The strongest first case has named model owners, reconciled inputs and an observable decision. Use AI to make that process easier to navigate while preserving the calculations and approvals that make the plan credible.