Configure agents in plain English. Revenue, budget, forecast, compliance, supply chain, treasury, and operations specialists respond to real-time market data and system state, adjusting plans continuously. Your team sets the boundaries; agents run the rest.
Sven is the conversational interface to Nordite's planning engine. Describe what you need in plain English and he helps build financial models, formulas, dashboards, reports, forecasts, simulations, and optimization workflows. He reads approved live data, operates inside governed guardrails, and produces structured, drillable outputs you can review. Upload your own documents, policies, and industry knowledge and Sven draws on them automatically, giving domain-specific answers instead of generic ones.
Nordite pairs governed multi-agent planning with task-aware capabilities and policy-aware access. The result is agentic execution that stays focused, inspectable, and safe to operate inside real finance workflows.
Agent capabilities are explicit, inspectable, and governed so specialist behavior can be reviewed and controlled.
Nordite narrows each agent task to the relevant planning context, keeping execution focused on the current business domain.
Tool access remains subject to user, tenant, and project policy, giving agents what they need without exposing unnecessary capability.
Forecasting, optimization, simulation, audit trails, live connectors, and zero-ETL open-format previews operate on the same model, so agents can react to current data rather than stale exports and handoffs.
Revenue, budget, forecast, compliance, and planning specialists can collaborate in one governed workspace with approvals, branching, traceability, continuous replanning, and in-place Parquet or Arrow data previews built in.
Agents adapt to your governance requirements. Run them interactively with approval on every action, let them operate on a schedule with periodic review, or delegate fully autonomous execution within strict guardrails. You choose how much autonomy each agent gets.
The agent proposes changes and waits for explicit approval before every action. Ideal for sensitive financial models, compliance-critical workflows, and initial deployments where your team wants full visibility before trusting automation.
Agents execute on a schedule - daily forecasts, weekly reports, monthly close workflows - and surface results for review. Low-risk actions run automatically; high-impact changes still require approval before merging to the live model.
Agents monitor live data streams, adjust assumptions, rebalance allocations, and update forecasts continuously. Everything operates within the permission boundaries and policy guardrails you define. Audit trails capture every action for full traceability.
When built-in tools are not enough, agents can propose custom analytical steps for review. Execution follows deployment policy, resource controls, and audit logging so your team can inspect what ran and why.
Agents can run approved statistical analysis, time-series decomposition, regression modeling, and custom visualizations that go beyond the standard toolset.
Custom analysis follows explicit policy controls, scoped inputs, and deployment-level resource boundaries. Generated code cannot act outside the permissions granted to the task.
In interactive and supervised modes, agents present proposed analytical steps for review before they run. Approvals, rejections, and outputs are logged in the audit trail.
Go beyond ad-hoc code generation. Define reusable Python or R scripts, declare their input schemas, and register them as callable MCP tools. Agents discover registered Code Blocks automatically and invoke them with validated parameters. Scoring models, data transformations, validation rules, custom calculations - write them once, let agents call them on every run. Every execution is logged with inputs, outputs, and duration for full auditability.
Nordite is not just an analytics layer on top of other tools. The workflow engine, data pipelines, transaction simulator, and real-time data processing are built into the platform. Agents orchestrate the full stack without external dependencies.
Drag-and-drop workflow canvas for building multi-step data pipelines, model dependencies, and agent orchestration flows. Connect modules, transforms, joins, filters, aggregators, solvers, and simulators visually. Every node is configurable and the full dependency graph resolves automatically.
Nordite processes transactional data as it arrives, not in batch cycles. Live sales transactions, inventory movements, expense entries, and cash flows feed directly into the live model. Agents react to actual business activity in real time, keeping forecasts and allocations aligned with what is happening now, not what happened last month.
Built-in extract, transform, and load pipelines connect to external databases, APIs, flat files, and cloud storage. Data flows through configurable transform, join, filter, and aggregation nodes before landing in the live model. No external ETL tool required.
Generate realistic transactional datasets for testing, demos, and scenario planning. Configure regions, transaction types, volume distributions, and seasonal patterns. Agents use generated data to validate models before live data is connected, or to stress-test plans against edge cases.
Traditional tools sit downstream of data warehouses and depend on periodic extracts. Nordite collapses the data pipeline into the platform, so agents work with current data instead of stale snapshots. The difference between daily batch cycles and continuous real-time processing is not incremental - it changes what is possible.
Each agent specialises in a business function, but works inside standards-based profiles, explicit skill packages, tenant and project overrides, and runtime guardrails. They collaborate through Nordite's shared model in real time while preserving auditability and control.
Source of TruthEvery action taken by an AI agent is logged, auditable, and reversible. Agents operate within your approval policies. No unapproved changes reach your financial model.
Agents work on isolated branches of your model. Proposed changes are visible, comparable, and reversible before they touch the live plan. Nothing merges without approval.
Sensitive actions - budget adjustments, forecast overrides, model structure changes - require explicit human approval through configurable multi-stage workflows before reaching the live plan.
Every agent action is logged with the agent identity, the creating user, the timestamp, and the full context of what was read, proposed, or changed. Audit-ready from day one.
Attribute-based access policies define exactly what each agent can see and do. A forecasting agent cannot export data. A reporting agent cannot modify formulas. Boundaries are structural, not just policy.
Autonomous does not mean uncontrolled. Every agent is bound by the permissions of the user who created it, enforced at the platform level. No agent can ever access data, models, or actions beyond what its creator is authorised to do.
An agent's maximum permissions are the permissions of the person who created it. If you can only access EMEA data, your agents can only access EMEA data. This is enforced at the engine level, not by policy alone.
Agents operate within strict tenant and project boundaries. Cross-tenant data access is architecturally impossible, and cross-project access requires explicit grants from an administrator.
Agents inherit role assignments from their creator. A viewer's agent can read but not write. An analyst's agent can model but not approve. Admin actions require admin-level credentials.
Every action an agent takes is logged with the agent identity, the creating user, the timestamp, and the full context of what was read, proposed, or changed. Nothing is invisible.
Agents can recommend changes, but committing to the plan, adjusting budgets, or modifying model structure requires explicit human approval through configurable workflows.
Agents are limited to explicitly whitelisted tools per skill profile. A forecasting agent cannot trigger data exports. A reporting agent cannot modify model formulas. Surface area is minimised by design.
Nordite was built by someone who ran a planning platform at scale and understands what finance leaders need before they trust automation: hard boundaries, not soft promises. Permissions are enforced at the engine level, not by prompt engineering or policy documents.
Agents go beyond chat-driven assistance. They can run entire business functions autonomously: monitoring live market data, adjusting planning assumptions as conditions change, rebalancing allocations across departments, and coordinating through shared context. Set the governance boundaries you want; agents handle everything within them.
Configure agents with schedules to run recurring workflows with the right level of supervision. A risk monitoring agent can refresh approved market signals, simulate supply chain outcomes, and generate an updated risk report before the team arrives. A forecasting agent can refresh projections weekly and flag deviations from prior periods.
Agents understand the full dependency graph across modules. When simulating oil price volatility, the agent follows the impact from market assumptions through factory operations, logistics costs, allocation, and margin analysis, re-evaluating linked formulas at every iteration. This produces accurate end-to-end probabilistic analysis that captures cascading effects across the entire model.
Agents fetch live market data during analysis: equities, commodities, bonds and interest rates, market indices, ETFs, foreign exchange rates, and cryptocurrency prices. A treasury agent can pull current yield curves and FX rates to update hedging models. A supply chain agent can fetch Brent crude and steel prices to recalculate cost projections. No manual data entry, no stale spreadsheets.
Agents collaborate through governed shared memory stores. A demand-sensing agent writes regional demand forecasts to shared memory, which a supply planning agent reads to adjust inventory targets, which a finance agent reads to update cash flow projections. Each agent operates within its own permission boundaries while sharing structured context for coordinated planning outcomes.
For organisations that choose it, agents can run every level of a business autonomously. They observe real-time market conditions and the data flowing through the system, adjust planning assumptions, rebalance allocations across departments, update forecasts, and restructure budgets as conditions change. The business becomes a continuously self-adjusting operation, governed by the boundaries you define.
Financial leaders define the task. Agents monitor, analyse, optimise, and report back - continuously, not just at month-end. Your team focuses on decisions, not data wrangling.
Set an agent to watch EMEA margins. If anything shifts beyond tolerance, it adjusts the forecast and sends a summary to the CFO before Monday's meeting.
An agent monitors actuals as they flow in, adjusts the rolling forecast, and flags variances that need human attention. No analyst cycle required.
Schedule an agent to pull regional P&Ls together every week, reconcile intercompany, and deliver a board-ready pack to your inbox.
An agent watches receivables ageing and payment patterns. When it detects a trend towards a shortfall, it alerts treasury with recommended actions.
Agents continuously model attrition scenarios, compare actual hiring against plan, and recommend adjustments before the gap becomes a problem.
Run 50 what-if scenarios overnight across every business unit. Wake up to a ranked summary of risks and opportunities, ready for the leadership meeting.
Agents can manage entire business functions end-to-end: reacting to live market data, adjusting models in real time, and coordinating across departments. One FP&A analyst covers the ground that used to require a team. Your people set strategy and governance boundaries; agents handle continuous execution.
Planning capacity that scales with compute, not headcount.
Every agent runs on the same engine that handles 200,000+ cell writes per second. Adding analytical capacity means adding an agent, not hiring an analyst. The economics of planning change fundamentally when the operational workload is automated, continuous, and governed.
Central governance
The model owns the plan. Targets, assumptions, and constraints flow from the source of truth to every agent, ensuring alignment without manual coordination.
Agents recommend, not commit
Each agent independently analyses data, runs scenarios, and proposes adjustments, but only the central model decides what gets approved and committed to the plan.
Full audit trail
Every recommendation, every approval, and every override is logged and versioned. Governance is built into the architecture. Your compliance team will thank you.
Continuous replanning
Not a monthly cycle, but a living loop. External data flows in, agents detect shifts, propose updates, and the model reconciles. Your plan is always current, never stale.