Plain chat with your data
Recognise the six jobs Start from Scratch handles, and how they build on each other.
A blank chat box is only useful if you know what to type.
Six plain-chat examples, each a short plain-language prompt and the result it returns:
Use casePromptWhat comes back Variance analysisPerform variance analysis with EBITDA 2025An EBITDA bridge from plan to actual Root cause analysisPerform Root Cause AnalysisA colour-coded driver tree Forecast via Monte Carlo simulationPerform a forecast for Q1A tornado chart: sensitivity of the result to ±10% swings in each driver Creation of email draftsCreate an email including summary for managementA drafted email, with buttons to copy it or open it in Gmail or Outlook Business recommendations / OKRsPropose OKRs to implement the recommendationsOKR recommendations: a set of objectives, each with an owner Deployment of AI agents for the executionSuggest agents to implement the recommendationsA page of suggested agents
Read down the prompts: the last two refer to "the recommendations", so they carry on from earlier answers. The sequence runs from diagnosis (variance, root cause, forecast) to follow-through (email, OKRs, agents).

Takeaways
Plain chat covers analysis — variance, root cause, Monte Carlo forecast — and follow-through — email, OKRs, agents.
Prompts are short and in plain language; the result type is chosen by what you ask for.
The later prompts build on earlier recommendations, so one conversation can run from diagnosis to action.
Surfaces covered
- Start from Scratch
- Circle of Advisors (CFO, CEO, CRO)
- Reports
- Value Creation
- Scenarios (what-if, break-even, Monte Carlo)
- Plans & Forecasting (SOON)
Also taught in the Academy
Module 11 · Valuesight Agent and AI
Take it as lesson 11.2 in Business Analyst & FP&A