FP&A · self-service forecast explanation

FP&A Self-Service Forecast Journey

Shows how a forecast variance turns into a governed, evidence-backed explanation as agents connect actuals, drivers, assumptions, and scenarios, draft the narrative and bridge, and route it for partner approval before the management pack updates — with replay and driver-accuracy learning.

Open FP&A command center Open forecast intelligence

Demo spine

Signal → context → agent → policy → evidence → decision → action → replay

Signal

Forecast variance

A material variance between actuals and the latest forecast enters the FP&A decision queue.

Context

Actuals + drivers

Agents read actuals, prior forecast, driver trees, assumptions, and comparable scenarios.

Policy

Guardrails + materiality

Capital guardrails, scenario policy, and materiality determine whether the explanation needs partner approval.

Agent

Forecast fleet

Variance, driver-attribution, and narrative agents build the bridge and recommend the explanation with confidence.

Decision

Approve explanation

Finance partner reviews the bridge and narrative and approves; the management pack is updated.

Replay

Driver trace

The explanation can be replayed to the actuals, drivers, and assumptions that produced it.

Evidence contract

What stakeholders can inspect

Fact	Source	Proof
Actuals + forecast	GL / planning context	Period actuals, prior forecast, budget, and variance by driver.
Driver tree	Forecast intelligence runtime	Volume, price, mix, cost, and FX contributions to the variance.
Assumptions	Scenario library	Assumptions, scenario set, and comparable prior explanations.
Narrative basis	Policy + history	Materiality, guardrails, and accepted prior narratives.

Outcome

What changes

Partner effortManual bridge rebuild avoided

Cycle timeExplanation ready in-session

Control statePartner approval before pack update

LearningDriver accuracy updated

Evidence Replay Value

Assurance proof

Policy, human gate, learning, value, and SOR access

Control	Visible proof
Policy evaluated	Capital guardrails, scenario policy, materiality threshold, and disclosure conditions.
Human gate	Finance partner approval required before the management pack or forecast narrative is updated.
Learning captured	Driver-attribution accuracy and accepted/rejected narratives feed forecast-driver learning.
Value attribution	Analyst effort avoided, cycle-time reduction, and forecast-accuracy improvement are attributed at case, driver, tower, and CFO levels.
SOR access	Actuals and driver facts read from the context/planning layer; no SOR write — the output is an approved explanation, not a posting.