ZeroTwo for analysts

Prompts for analysts

The analyst's day is mostly the mechanical part of thinking: pulling, cleaning, cross-checking and formatting before the analysis can start. These prompts cover that part and stop where your judgment begins.

Every figure states its base
Data problems reported before the chart
You verify the arithmetic

Prompts for analysts

How do I check a dataset before I analyse it?

Find the gaps, duplicates and small base sizes first — before any conclusion is built on them.

The promptBefore any analysis, profile this export: row count, missing values by column, duplicates, outliers, and any segment with a base under thirty. List what you find before drawing a single conclusion.

Export_Q3.xlsx

Does my data actually support the claim I want to make?

Ask directly, and ask for the counter-case. A dataset that only confirms you is usually being read selectively.

The promptI want to argue that [claim]. Using only this dataset, tell me honestly whether it supports that, what the strongest counter-reading is, and what would have to be true for my claim to hold.

Dataset.csv

How do I check the sources in an AI-written summary?

Verify each citation resolves to something retrievable. Check all of them the first time, then sample.

The promptGo through this summary and check every citation points to a real, retrievable source. List the ones that do not, and do not silently replace them.

Summary_draft.docx

How do I explain a change in the numbers?

Decompose the movement before narrating it, so the explanation follows the arithmetic rather than the other way round.

The promptThis metric moved between the two periods. Decompose the change into its contributing factors, show the arithmetic, and rank the factors by how much of the movement each explains. Flag anything the data cannot attribute.

Period_compare.xlsx

How do I sanity-check a model someone else built?

Trace the assumptions and find the ones the output is most sensitive to.

The promptRead this model. List its assumptions, say which cells drive the output most, and identify anything hard-coded that looks like it should be an input. Do not change the file.

Model_v6.xlsx

What do I do when two sources give different numbers?

Show the disagreement and judge which is more credible. Averaging produces a figure nobody will defend.

The promptThese sources report different figures for the same measure. Show the discrepancy, explain the likely reason for it, and say which is more credible and why. Do not average them.

What belongs in a chart footnote?

Source, base, date and method — plus a label on any segment too small to stand alone.

The promptTurn this into a chart. State the base for every figure, label any segment under thirty as indicative, and write the footnote with source, base, date and method.

Segments.xlsx

Which chart type should I use for this?

Pick from what the data can support, not from what looks impressive.

The promptGiven this data and the point I need to make, recommend a chart type and say what it would mislead about. If the data cannot support the comparison I want, say so instead of drawing it.

How do I synthesise a pile of research quickly?

Theme it with sources attached, and keep the outliers rather than smoothing them away.

The promptSynthesise these documents into themes. Attach the source and date to every claim, keep contradictions visible as their own list, and flag outlier findings rather than dropping them.

Research_pack.pdf

How do I write a brief a stakeholder can challenge?

Separate what you know, what you infer, what you do not know, and what you recommend.

The promptWrite this up as a brief structured into what we know, what we think, what we do not know, and what we recommend — kept strictly separate. Every factual claim carries an inline source and date.

How do I rerun the same analysis next month?

Save the method, not just the output, so the second run costs less than the first.

The promptWrite down the method used here as a repeatable procedure: inputs required, steps, checks, and the format of the output. I will run this again next cycle against new data.

How do I explain this analysis to a non-technical audience?

Translate the finding without softening the caveats that make it defensible.

The promptRewrite this finding for an audience with no statistical background. Keep every caveat that affects how the result should be used, and do not round away uncertainty.

Findings.docx

Prompts are starting points. Filenames shown are placeholders — attach your own.

In short

ZeroTwo is an AI workspace for analysts who have to produce work someone else will challenge. It reads your exports, notes and prior analyses, keeps every claim tied to the line it came from, reports data-quality problems before building anything, and leaves the interpretation to you.

How a prompt becomes a deliverable

A prompt starts the work; it is not the work. Every card above runs the same four beats, and stops in the same place.

Inputs, as they are

Transcripts, exports, the client's own documents, last cycle's deck. No re-typing or restructuring first.

A plan you can adjust

ZeroTwo shows what it intends to do. You set the source boundaries before it starts.

Assembly, with sources kept

Claims stay tied to the line they came from. Gaps and conflicts are surfaced, not smoothed over.

A first-review artifact

Complete enough to review against a written standard. Not a signed deliverable.

ZeroTwo stops at the judgment

The approval boundary is a product feature, not a disclaimer. ZeroTwo does the assembly. You keep the part the client is actually paying for.

ZeroTwo does

Gather and organise the inputs
Extract, theme and cross-reference
Keep every claim tied to its source
Surface gaps and contradictions
Format to your conventions

You do

Set the source boundaries
Resolve the contradictions
Verify the arithmetic
Write the recommendation
Sign it and send it

Against what you are doing now

Approach	Where it works	Where it breaks on a deliverable	ZeroTwo's difference
General chat assistant	Fast answers, flexible, already paid for	Stops at a draft. analysts own the path across files, slides and sheets, and re-explain the project every session.	Context and sources persist across the whole engagement
Specialist point tool	Deep fit for one format or step	Work fragments across the rest of the process; each tool holds a different half of the context.	One accountable path from inputs to artifact
Doing it by hand	Trusted method, clear accountability	Senior time goes to production and cleanup; recurring work restarts from memory each cycle.	The second run reuses the first run's method
A contractor	Human judgment, flexible capacity	Lead time and context transfer exceed the window when the deadline is days away.	Available at the moment the brief lands

Questions analysts ask first

Will it check my data or just chart it?

The data-quality prompts run before anything is built: gaps, duplicates, base sizes and outliers get reported first, and you decide whether to proceed. An exhibit built on a dataset nobody profiled is the failure that gets noticed in the room.

Can I trust the numbers it produces?

Treat every figure as needing your verification. ZeroTwo shows its calculation and states the base so the arithmetic is checkable — that is the point. It is not a substitute for you re-running the number.

How is this different from an AI data analysis tool?

Task-level tools analyse a dataset. This is aimed at the analyst's whole job — profiling inputs, synthesising research, building the exhibit and writing it up so someone else can challenge it. For the task itself see AI agent for data analysis.

Do I have to upload client or company data?

No. Run the first pass on public, synthetic or sanitised data. Nothing about the workflow requires real data to evaluate whether the output meets your standard.

Will it keep our conventions between runs?

Your chart conventions, evidence standard and report structure are saved as project context and reapplied, so the second and third run cost less than the first.

Does this replace the analyst?

It removes the extraction and formatting, which is the part that delays the thinking. Interpretation, method choice and anything you would have to defend stay with you.

Takeaways

Data-quality problems are reported before an exhibit is built, not after it is presented.
Every figure states its base and shows its calculation, so the arithmetic stays checkable.
Research synthesis keeps sources attached and contradictions visible.
Interpretation and anything you would have to defend stay with you.