Design partners We're building agentic inventory management and are looking for design partners to shape it with us. Agentic inventory design partners Contact us

Variance analysis. Plan vs. actual, live.

Every booking is checked against your plan the moment it is posted, and every variance arrives with its driver.

Variances as they happen. With the reason.

No waiting for the monthly review: the variance, its driver and the updated forecast arrive together.

Real-timeCaught at booking

Variances surface the moment a booking is posted, not in the monthly meeting.

DriversExplained

Instead of a bare “−8.4 %”, you see the driver: volume, price, mix, FX or timing.

ForecastSelf-updating

Rolling forecasts update from bookings, contracts and supplier communication.

You see it first. Not the monthly review.

The controlling agent checks every booking against your plan – in real time. The moment a variance crosses your thresholds, a notification lands in Slack, Teams or your inbox. Before anyone asks.

  • Real-time comparison against plan, not just at month-end
  • Thresholds per cost center, account and project, freely configurable
  • Notifications in Slack, Teams, email or dashboard – you choose

With a recommendation. Not just numbers.

A naked number isn't an insight. The agent ships the driver breakdown, links the underlying postings and contracts, and proposes three concrete next steps - who to talk to, what to reallocate, which assumption to revisit.

  • Driver breakdown: volume, price, mix, FX, timing
  • Three proposed next steps – not "please investigate"
  • Linked postings, contracts and source documents for the deep dive

Ready for the conversation. With your team or an AI.

Every variance arrives with context you can use immediately. One click exports the analysis for the steering meeting. Or ask ChatGPT, Claude or Copilot directly - the agent is wired up via an MCP server and returns numbers with source references.

  • One-click export for steering committees, board or stakeholders
  • Direct MCP access for ChatGPT, Claude and Microsoft Copilot
  • What-if scenarios on demand – no model marathon

Use one agent. Or the full suite.

Every Orcha module works on its own – and compounds when combined.