Documentation

Modeler guide — your first verification pass

Who this is for: You build or own the model. You need confidence before a decision — without waiting for a full external review cycle.

Companion docs: Install reference · First analysis · Compliance reviewer guide

What Kalc gives you (outcomes)

After a pass, expect:

Kalc identifies and suggests. You stay responsible for the analysis and the decision.

Before you run — pick the right settings

Template / sector

Choose the template closest to your model (general or a sector template your plan includes). This shapes which financial-logic checks apply.

Zero Retention

Turn on for client-confidential work if you do not want Kalc to store results after the run. Your file is still analysed in memory for the duration of the pass.

Evidence environment

New workspaces default to Contained — nothing about your model is sent to a third party. If your workspace uses Grounded advisory features, assumption labels and values may be sent to an EU AI provider — see the compliance guide.

A sensible first pass

  1. Start on a copy if you are nervous — same workflow, reversible.
  2. Run on one complete scenario case — the case you would actually present.
  3. Read Critical and High first — fix or document before Medium/Low.
  4. Use the assumptions tab — documentation gaps are often quick wins.
  5. Re-run after material fixes — see what resolved.

Interpreting findings (plain language)

ThemeWhat it usually meansWhat to do
Formula inconsistencySame row/column pattern breaksCompare neighbours; check fills
Hard-coded valuesNumbers embedded in formulasMove to input cells or document why
Undocumented assumptionsDriver with no clear labelAdd label or note in register
Statement tie issuesP&L/BS/CF may not reconcileTrace links; fix before external review

When to escalate to human review

Use Kalc passes between formal reviews. Still engage your usual external reviewer when:

Kalc is designed to prepare you for that conversation — not replace it.

Agent / API path (optional)

If your team builds models with automated agents, the same independent check is available via API/MCP — verify after generation, remediate from findings, re-verify. See Build and verify — complement.

Get help

Support · Privacy · Terms

← All docs