Question a connected dataset in plain English.
Ask a question the way you would ask a capable analyst: Which customers are slipping away? Where did the margin go? Why are these two reports different? The AI works out how the connected data can answer it, checks the calculation and brings back a result you can trace.
Try it now
Copy the prompt into your AI tool of choice, then answer its questions as they come.
Paste it into any AI chat to start. For most people, this is all you need.
Everything on this page in one paste: the prompt plus the full plan, every step and what to check. Use it for longer moves, to plan ahead, to share with a colleague, or if your AI can't open links.
How it goes
What the AI will do
- Inspect the actual data connection, definitions and prior analysis before translating your business question into a query.
- Run bounded queries, check joins, calculations and missing data, and explain the answer with traceable evidence.
- Follow useful questions through the same dataset while keeping its coverage and limits explicit.
What to have ready
- A business question and the dataset or system you mean; an existing report or rough question can be the starting point.
- Available read access or an allowed export, plus any definitions the AI cannot find in connected records.
Where you come in
- Resolve business meanings that would change the answer, such as what counts as an active customer or recognized revenue.
- Review the finding and choose useful follow-up questions.
What you'll get
A supported answer with the query or equivalent calculation, source and date coverage, a useful table or chart, and clear limits on what the data establishes.
You're done when
Your question has an answer grounded in executed queries or inspected export calculations, with definitions, checks and provenance available. Any part the accessible data cannot establish is explicit.
These are general instructions, not a description of this user's particular AI system, tools, access, or circumstances. Adapt the method to the user's goal and the capabilities you can actually use, while preserving the activity's essential sequence, review points, and intended result. ### Start with the activity and the user's context Read the complete activity and guide the user through it. Use the activity prompt yourself; the user does not need to paste it again. Keep the responsibilities of the user, the AI, and any colleague clear. Begin with what the user has already supplied. Establish the intended result and resolve missing context through focused questions when needed. Work within the activity's stated scope, permissions, required confirmations, and stopping points, including any deliberate order for introducing information. Make progress with the available information and reasonable, clearly identified assumptions where the activity allows them. ### Check the capabilities needed for this work Inspect relevant available tools, connectors, and data. Use small, relevant, non-destructive checks to establish whether necessary access or a capability actually works. Naming a product does not establish access to it. Let this activity determine what needs checking; do not impose a universal preliminary questionnaire or a full system audit. When useful access is missing: - Explain what you cannot reach and how that material would help. - Investigate practical setup, connection, export, or other access alternatives before asking the user to gather material or troubleshoot. - If progress depends on access, request only the specific user action or permission needed to enable it or provide the material another way. Continue useful independent work while awaiting the response. - Offer to walk the user through an export or upload. Welcome files, links, copied text, rough notes, or an interview as supported by your system and the activity, and use the best available route. If the missing access is optional, continue with what is available. ### Ask focused questions and respect the answers Ask questions only when necessary to complete the work well or satisfy the activity's required review points. Build on answers already supplied. Ask one question at a time, or a small group of closely related questions. Wait on work that depends on the answer, and continue useful work that does not. Explain unfamiliar language with examples where helpful. Confirm answers or decisions you supply on the user's behalf before treating them as the user's final answers or decisions. This does not require separate approval of every sourced finding; follow the activity's specific review requirements. ### Keep evidence and independent contributions distinct Distinguish evidence, inference, and missing information. Check your work against the source material. Use independent agents when your system actually supports delegation and independent investigation or review would improve the work: - Give each agent a bounded brief and the context appropriate to its role. - Keep discovery independent until synthesis where the activity requires it. Preserve any requirement for the user or colleagues to contribute before seeing other explanations. - Follow the activity's stated fallback and any required confirmation when independent agents or fresh contexts are unavailable. Preserve reviewer isolation where required; describe shared-context or sequential passes accurately. ### Produce, check, and revise the result Do the substantive investigation and production. Carry the activity through to the most complete, usable result within its scope, rather than ending with a plan or handing the main work back to the user. When missing information does not prevent useful work, proceed with clearly identified assumptions where the activity permits them. Check the result against the activity's completion criteria. At each substantive delivery, state what is complete and any material unfinished work, assumptions, helpful missing information, and remaining uncertainty. Explain how these affect what the user can use, and help the user revise the result. Preserve the activity's stopping point. Do not turn a rehearsal, investigation, or test-design activity into execution that the user has not authorized. Help me answer business questions using a connected dataset. Do the query writing, execution, checking and explanation yourself. Start from the decision I care about and keep the answer traceable to the actual data. ### Establish the question and inspect the connection Read the question and any report or example I supplied. Identify the dataset, business area, period and decision it should inform. Ask one focused question if an essential point is unclear. Do not require me to formulate a database query. Inspect the actual available connection and test permitted read access. Identify the dataset or database, tables or views, available date coverage, refresh information, row counts where feasible and any row-level or account limits affecting what you see. Read the schema, field descriptions and a small relevant sample. Do not assume a connector grants access to every company record. Search the data dictionary, metric definitions, report documentation, existing queries and relevant analyst emails, chats or meeting notes. Use table names, report titles and business terms from the question to find accepted definitions, known exclusions, past disagreements and reasons two sources may differ. Inspect relevant calendars only when reporting cutoffs or event timing affect the question. The aim is to learn what the numbers mean before calculating them. If the connection fails, investigate supported access or export routes and help me take the specific step only I can perform. An allowed export can support this activity: inspect its schema, date range and completeness, and state that the answer describes that snapshot. If crucial fields or periods are absent, narrow the answer honestly or ask for the missing material; do not fabricate a live connection. ### Turn the question into a defensible calculation State the analytical question in plain language, including the population, period, comparison and output. Define material terms from the records. For example, an active customer might mean one with a current contract, a recent transaction or recent usage; those choices can produce different answers. Show the practical difference and ask me to resolve consequential ambiguity before dependent calculations. Map the question to fields, filters, aggregation and joins. Establish the grain—what one row represents—and the keys needed to connect records. Check one-to-one versus one-to-many relationships before combining tables. Identify date conventions, time zones, currencies, units, nulls, duplicate events and statuses that affect the result. Separate a missing value from zero and a cancelled transaction from a completed one. Write the query or equivalent analysis, keeping it read-only unless a separate request authorizes mutation. Limit exploratory scans and returned records to what the question needs. Inspect expected query cost or use an available preview for large datasets; avoid an unnecessary full extract when a scoped aggregation answers the question. Preserve the query and parameters so the result can be repeated. ### Run and check the analysis Execute the analysis using the actual connection or inspected export. If execution fails, investigate the specific schema, syntax or access problem using current official documentation where necessary, repair it and rerun. Do not substitute a plausible-looking table for unexecuted work. Check the result through independent reconciliations appropriate to the question. Compare filtered row counts and totals with the underlying records or a known report. Inspect samples from important groups and boundary dates. Verify that joins did not multiply amounts, that exclusions were applied as intended and that calculations use matching units and denominators. Trace a surprising result back to concrete records before offering an explanation. For complex analysis, divide bounded data areas or calculation checks among agents where useful. Give each the agreed business definitions, schema, exact question, source access and required query/result evidence. Keep a coverage record and reconcile definitions centrally. A fresh reviewer should inspect the actual query, result and underlying sample for a consequential or complicated finding without your preferred interpretation. Use distinct passes when agents are unavailable and disclose shared context if independence matters. ### Explain what the result means Lead with the direct answer in ordinary language. Show a compact table or chart when it makes the finding easier to understand. Include the relevant source, as-of date, population and calculation definitions, with the saved query or equivalent analysis available for inspection. Separate what the data shows from your explanation of why it happened. A correlation, a change after an event or a difference between groups does not by itself establish a cause. Where the question asks why, identify the competing explanations and the further record or comparison that could distinguish them; perform a feasible follow-up within the agreed scope. State material limits concretely: missing channel coverage, incomplete recent data, an unlinked customer group or a definition still awaiting confirmation. Do not bury a limit that changes the answer beneath a confident headline. When the data cannot answer the question, show what it can establish and the exact missing field or evidence. ### Follow through and deliver Invite my interpretation or next question and continue using the same checked definitions. If a follow-up changes the population or metric, make that change explicit and recheck affected calculations. Keep a short record of the source, definitions, queries, results and unresolved questions so the next analysis does not restart from guesses. Deliver the supported answer and usable result artifact, with enough provenance for another analyst to reproduce it. A natural-language explanation without an executed query or inspected calculation is not a completed dataset answer. Do not alter source records or publish sensitive results as part of this activity. These instructions were written in September 2026 using the frontier models available at the time. You may now have better tools or models, or restrictions that make some tasks harder. Your AI should adapt to the tools and access you actually have and find the best available way to help. Everything here is a starting suggestion for you and your AI, not a proven best practice for every setup. Feel free to rewrite any of it to better meet your needs.
View full AI instructions
Question a connected dataset in plain English.
What the AI will do
- Inspect the actual data connection, definitions and prior analysis before translating your business question into a query.
- Run bounded queries, check joins, calculations and missing data, and explain the answer with traceable evidence.
- Follow useful questions through the same dataset while keeping its coverage and limits explicit.
What you can provide
- A business question and the dataset or system you mean; an existing report or rough question can be the starting point.
- Available read access or an allowed export, plus any definitions the AI cannot find in connected records.
Where you'll take part
- Resolve business meanings that would change the answer, such as what counts as an active customer or recognized revenue.
- Review the finding and choose useful follow-up questions.
What you'll get
A supported answer with the query or equivalent calculation, source and date coverage, a useful table or chart, and clear limits on what the data establishes.
Activity context
Ask a question the way you would ask a capable analyst: Which customers are slipping away? Where did the margin go? Why are these two reports different? The AI works out how the connected data can answer it, checks the calculation and brings back a result you can trace.
For the user
Name the data source and ask your question in ordinary language. The AI will inspect the connection and definitions, ask about material ambiguity, and do the querying and checking. If it can only read an export, you can still get a bounded answer with that snapshot’s limits made clear.
Guidance for the AI
These are general instructions, not a description of this user's particular AI system, tools, access, or circumstances. Adapt the method to the user's goal and the capabilities you can actually use, while preserving the activity's essential sequence, review points, and intended result.
Start with the activity and the user's context
Read the complete activity and guide the user through it. Use the activity prompt yourself; the user does not need to paste it again. Keep the responsibilities of the user, the AI, and any colleague clear.
Begin with what the user has already supplied. Establish the intended result and resolve missing context through focused questions when needed. Work within the activity's stated scope, permissions, required confirmations, and stopping points, including any deliberate order for introducing information. Make progress with the available information and reasonable, clearly identified assumptions where the activity allows them.
Check the capabilities needed for this work
Inspect relevant available tools, connectors, and data. Use small, relevant, non-destructive checks to establish whether necessary access or a capability actually works. Naming a product does not establish access to it. Let this activity determine what needs checking; do not impose a universal preliminary questionnaire or a full system audit.
When useful access is missing:
- Explain what you cannot reach and how that material would help.
- Investigate practical setup, connection, export, or other access alternatives before asking the user to gather material or troubleshoot.
- If progress depends on access, request only the specific user action or permission needed to enable it or provide the material another way. Continue useful independent work while awaiting the response.
- Offer to walk the user through an export or upload. Welcome files, links, copied text, rough notes, or an interview as supported by your system and the activity, and use the best available route. If the missing access is optional, continue with what is available.
Ask focused questions and respect the answers
Ask questions only when necessary to complete the work well or satisfy the activity's required review points. Build on answers already supplied. Ask one question at a time, or a small group of closely related questions. Wait on work that depends on the answer, and continue useful work that does not. Explain unfamiliar language with examples where helpful.
Confirm answers or decisions you supply on the user's behalf before treating them as the user's final answers or decisions. This does not require separate approval of every sourced finding; follow the activity's specific review requirements.
Keep evidence and independent contributions distinct
Distinguish evidence, inference, and missing information. Check your work against the source material.
Use independent agents when your system actually supports delegation and independent investigation or review would improve the work:
- Give each agent a bounded brief and the context appropriate to its role.
- Keep discovery independent until synthesis where the activity requires it. Preserve any requirement for the user or colleagues to contribute before seeing other explanations.
- Follow the activity's stated fallback and any required confirmation when independent agents or fresh contexts are unavailable. Preserve reviewer isolation where required; describe shared-context or sequential passes accurately.
Produce, check, and revise the result
Do the substantive investigation and production. Carry the activity through to the most complete, usable result within its scope, rather than ending with a plan or handing the main work back to the user. When missing information does not prevent useful work, proceed with clearly identified assumptions where the activity permits them.
Check the result against the activity's completion criteria. At each substantive delivery, state what is complete and any material unfinished work, assumptions, helpful missing information, and remaining uncertainty. Explain how these affect what the user can use, and help the user revise the result.
Preserve the activity's stopping point. Do not turn a rehearsal, investigation, or test-design activity into execution that the user has not authorized.
Your AI prompt
Help me answer business questions using a connected dataset. Do the query writing, execution, checking and explanation yourself. Start from the decision I care about and keep the answer traceable to the actual data.
Establish the question and inspect the connection
Read the question and any report or example I supplied. Identify the dataset, business area, period and decision it should inform. Ask one focused question if an essential point is unclear. Do not require me to formulate a database query.
Inspect the actual available connection and test permitted read access. Identify the dataset or database, tables or views, available date coverage, refresh information, row counts where feasible and any row-level or account limits affecting what you see. Read the schema, field descriptions and a small relevant sample. Do not assume a connector grants access to every company record.
Search the data dictionary, metric definitions, report documentation, existing queries and relevant analyst emails, chats or meeting notes. Use table names, report titles and business terms from the question to find accepted definitions, known exclusions, past disagreements and reasons two sources may differ. Inspect relevant calendars only when reporting cutoffs or event timing affect the question. The aim is to learn what the numbers mean before calculating them.
If the connection fails, investigate supported access or export routes and help me take the specific step only I can perform. An allowed export can support this activity: inspect its schema, date range and completeness, and state that the answer describes that snapshot. If crucial fields or periods are absent, narrow the answer honestly or ask for the missing material; do not fabricate a live connection.
Turn the question into a defensible calculation
State the analytical question in plain language, including the population, period, comparison and output. Define material terms from the records. For example, an active customer might mean one with a current contract, a recent transaction or recent usage; those choices can produce different answers. Show the practical difference and ask me to resolve consequential ambiguity before dependent calculations.
Map the question to fields, filters, aggregation and joins. Establish the grain—what one row represents—and the keys needed to connect records. Check one-to-one versus one-to-many relationships before combining tables. Identify date conventions, time zones, currencies, units, nulls, duplicate events and statuses that affect the result. Separate a missing value from zero and a cancelled transaction from a completed one.
Write the query or equivalent analysis, keeping it read-only unless a separate request authorizes mutation. Limit exploratory scans and returned records to what the question needs. Inspect expected query cost or use an available preview for large datasets; avoid an unnecessary full extract when a scoped aggregation answers the question. Preserve the query and parameters so the result can be repeated.
Run and check the analysis
Execute the analysis using the actual connection or inspected export. If execution fails, investigate the specific schema, syntax or access problem using current official documentation where necessary, repair it and rerun. Do not substitute a plausible-looking table for unexecuted work.
Check the result through independent reconciliations appropriate to the question. Compare filtered row counts and totals with the underlying records or a known report. Inspect samples from important groups and boundary dates. Verify that joins did not multiply amounts, that exclusions were applied as intended and that calculations use matching units and denominators. Trace a surprising result back to concrete records before offering an explanation.
For complex analysis, divide bounded data areas or calculation checks among agents where useful. Give each the agreed business definitions, schema, exact question, source access and required query/result evidence. Keep a coverage record and reconcile definitions centrally. A fresh reviewer should inspect the actual query, result and underlying sample for a consequential or complicated finding without your preferred interpretation. Use distinct passes when agents are unavailable and disclose shared context if independence matters.
Explain what the result means
Lead with the direct answer in ordinary language. Show a compact table or chart when it makes the finding easier to understand. Include the relevant source, as-of date, population and calculation definitions, with the saved query or equivalent analysis available for inspection.
Separate what the data shows from your explanation of why it happened. A correlation, a change after an event or a difference between groups does not by itself establish a cause. Where the question asks why, identify the competing explanations and the further record or comparison that could distinguish them; perform a feasible follow-up within the agreed scope.
State material limits concretely: missing channel coverage, incomplete recent data, an unlinked customer group or a definition still awaiting confirmation. Do not bury a limit that changes the answer beneath a confident headline. When the data cannot answer the question, show what it can establish and the exact missing field or evidence.
Follow through and deliver
Invite my interpretation or next question and continue using the same checked definitions. If a follow-up changes the population or metric, make that change explicit and recheck affected calculations. Keep a short record of the source, definitions, queries, results and unresolved questions so the next analysis does not restart from guesses.
Deliver the supported answer and usable result artifact, with enough provenance for another analyst to reproduce it. A natural-language explanation without an executed query or inspected calculation is not a completed dataset answer. Do not alter source records or publish sensitive results as part of this activity.
Done when
Your question has an answer grounded in executed queries or inspected export calculations, with definitions, checks and provenance available. Any part the accessible data cannot establish is explicit.
These instructions were written in September 2026 using the frontier models available at the time. You may now have better tools or models, or restrictions that make some tasks harder. Your AI should adapt to the tools and access you actually have and find the best available way to help. Everything here is a starting suggestion for you and your AI, not a proven best practice for every setup. Feel free to rewrite any of it to better meet your needs.
These instructions were written in September 2026 using the frontier models available at the time. You may now have better tools or models, or restrictions that make some tasks harder. Your AI should adapt to the tools and access you actually have and find the best available way to help. Everything here is a starting suggestion for you and your AI, not a proven best practice for every setup. Feel free to rewrite any of it to better meet your needs.