No row of your firm's data is ever sent to the AI provider.
You type a question in ordinary words, such as "How many files were opened last month?". The AI writes a database query for that question. Your firm's connector, a small service that holds the database login and checks every query, then runs the query on your firm's own database, read-only. The result comes back through the CaseQuery service to your browser and is shown to you. The AI writes the query, but it never sees the figures that come back. The sentence that describes your answer is put together by CaseQuery from those figures, not by the AI.
AI CaseQuery can only read. It cannot change, add or delete anything in the database.
Some things do go to the AI: your question, and the names of the tables and columns it needs. Section 12 lists exactly what is sent, and what never is.
AI CaseQuery runs in your web browser. There is nothing to install: open your firm's CaseQuery web address in any modern browser.
Each person has their own sign-in. Your administrator adds you, gives you your first password privately, and decides your role: User (asks questions), Lead (also settles definitions and reviews answers) or Administrator (also manages Settings and people). Nobody shares a key or a password.
If you forget your password, ask your administrator to reset it. You cannot change it yourself.
If your firm uses Microsoft 365 and has switched this on, the sign-in page also offers Sign in with your work account. Your administrator sends you a one-time invitation link (it lasts seven days) to link your work account once; after that, click the button and sign in with Microsoft as you do elsewhere. Your firm may require it, in which case use it every time.
A few actions that change who can reach your firm's data, or where it goes, open a box titled Confirm it is you if you last confirmed more than fifteen minutes ago. Enter your password again, or sign in with Microsoft again if that is how you sign in. For most people this happens only when exporting a result table (section 5).
Click Sign out at the foot of the left-hand menu, and confirm. The badge at the top of the screen, which reads Signed in · Sign out (or Lead · Sign out, or Administrator · Sign out, depending on your role), signs you out as well. Sign out when you leave a shared computer.
| What you see | What it is for |
|---|---|
| Chat | Where you ask questions and read answers. Most of this guide is about this screen. |
| Fine-tuning and Business Rules | Anyone can read these. They show what your firm has told the AI, which can help you judge an answer. Changing anything there is work for leads and administrators. |
| Settings | For administrators only. Anyone else sees a padlock beside it, and it does not open for them. |
| New chat | Starts a fresh conversation. The one you were in stays in your chat history. |
| Ask Casey | Opens Casey, the assistant for questions about the product (see section 11). It is only there if your administrator offers Casey. |
| Chat history | Opens your past conversations. The number beside it shows how many are saved in this browser for you. |
| User guides | Opens this guide and the others, each in a new tab. |
| Sign out | Ends your session (see section 2). |
Near the bottom of the sidebar, a few short lines tell you the state of things before you ask:
| Badge | What it means |
|---|---|
| Database type | The kind of database you are asking, or Not connected. |
| Model | The AI model your questions go to. |
| Safe mode on | Your administrator's safe response mode is on. It favours totals and limits long lists. If it reads Safe mode off, longer lists are allowed. |
| "… provisional" | Definitions that have been settled and are already in use, but that the firm has not reviewed yet. This is for leads and administrators, and you can ignore it. |
| Fine-tuning with a tick or a count | How much of the recommended fine-tuning is in place. Click it to see a short summary. |
| Signed in · Sign out | Shows that you are signed in, and your role if you are a lead or an administrator. Hover over it to see your sign-in name and firm. Click it to sign out. |
Ask the way you would ask a colleague who knows the database but not what you have in mind. The clearer you are about what to count, which ones and over what period, the less the AI has to guess.
| Vague | Clearer | Why it helps |
|---|---|---|
| How many cases? | How many cases were opened in August 2026? | Says which cases and when. |
| Show me the active ones. | List the files that are still open, with their client and open date. | "Active" can mean several things. "Still open" is plainer, and asking for columns tells it what to show. |
| Who's behind? | Which files have a task that is past its due date and not completed? | Says what "behind" means. |
| Totals by client | Total fees billed per client this year, largest first. | Names the figure, the grouping, the period and the order. |
Before you ask anything, the chat screen shows a few example questions. Clicking one puts it in the question box. It is not sent. You can change the wording first or send it as it is. Starter questions are chosen by your administrator for each database, and everyone at your firm sees the same ones for that database, whichever computer or browser they use.
If you know which table your question is about, you can say so. Type @ in the question box and a short list of your tables appears; keep typing to narrow it, use the arrow keys or the mouse, and press Enter or click to choose one. Nothing appears until you type @, so the question box stays a question box if you never want this.
Beside each table, under Often needed with it, are the tables that one is joined to — either because your database declares the link, or because queries that have run here have used it. Clicking one adds it to the question as well. That saves the common case of asking about invoices and forgetting that the case number lives on the matter.
Naming a table is a preference, not a restriction. If answering your question genuinely needs another table, it will still be used — and the answer says so, naming what it added. If nothing you typed matched a table, the answer says that too, and tells you what it read instead, so you are never left believing a question was narrowed when it was not.
Only tables your firm lets you see are offered. A table your firm has withheld does not appear in the list.
The button does not always say Run:
If your administrator has switched on Confirm token estimate, a dialog titled Review estimated token use appears before the question is sent. Nothing has reached the AI provider yet. It shows:
View estimate details shows the breakdown. Click Okay, run request to send, or Cancel to stop. The figure is an approximate estimate made by CaseQuery before anything is sent. "Tokens" are the units the AI provider charges by. A question your firm has already verified costs nothing, so no estimate appears for it.
If your wording is close to a question your firm has already verified, AI CaseQuery offers it first: Your DBA has already settled a question like this one. Your question and the settled one appear side by side. Choose Yes, that is my question to get the verified answer, or No, mine is different to have your own wording worked out as usual. You are the only one who knows what you meant, so read both before choosing.
Your administrator can have AI CaseQuery check each question before it is sent to the AI provider. This is on by default. If your question seems to contain something your organisation does not allow to be sent, such as a Social Security number, a payment card number, an email address, a telephone number or a real value from a sensitive column, you will see This question cannot be sent.
Nothing was sent to the AI provider. The dialog lists what was recognised. Click Ask a new question and your question goes back in the box for you to reword. Ask without the specific value, for example by name, date or status instead of an identifier. The application can usually still answer.
An answer is built from several parts. Which parts you see depends on the question, on your Explanation level (section 10) and on what your administrator has switched on. Anything warning that an answer may be wrong is shown at every level, except the line saying which business rules were not used, which shows from Reasoning up.
A short sentence or two saying what was found. When the result has figures, a plain description of them follows. That description is put together by CaseQuery from the results, not by the AI. Where your firm has more than one database, a line under the answer says which one it came from, for example "From: Case records".
The rows that came back, under a heading that says they are never sent to the AI, followed by how many rows are displayed. When the table is the answer, such as a breakdown by month, it opens straight away. When a sentence is the answer, the table is folded and you can click to open it. Long tables show 30 rows at a time, with a button to show more. An answer returns at most 500 rows.
Your administrator may turn the result table off for everyone. The answer in words is still shown.
The rows travel from your firm's connector, through the CaseQuery service, to your browser. The service does not keep them.
From the Standard level up, a short list shows the choices the AI made in working out your answer. For example, it might say which date it used or which records it left out. Read this list: a figure can be calculated perfectly and still measure something your firm does not mean. Each item has a Not right? link (see section 7).
Beside the answer's title you may see Verified query, followed by a name and a date. It means the answer came from a query someone at your firm checked and settled. No AI call was made, and the same checked query is run every time you ask, so the figure changes only when your data does. If nobody has re-checked it for a long time, the mark says how many days it has been.
At the Reasoning level and above, a folded line may say something like "3 of 12 business rules were not used for this question". Those rules describe tables your question does not touch, so they could not have applied. They are listed so that if one should have applied, you can spot it and report it. At Plain and Standard the line is not shown.
This folded line means the answer depends on a word or measure your firm has not yet defined. It often shows how much that matters, such as "58 rows one way, 44 the other". Open it to see the question and the possible readings.
Choosing one of the readings settles the definition for everyone, so it needs the Lead or Administrator role. If your role is User, click Neither — escalate to Lead. That sends the question, with your wording, to the top of the list a lead works through. If you are a lead, see the Lead Guide: Fine-tuning.
Sometimes the AI stops and asks which of two or more readings you meant, instead of guessing. It offers a few choices. Clicking one puts it in the question box without sending it, so you can adjust the wording, or ignore the choices and ask your own way.
If your administrator has switched on Suggest follow-up questions, two or three related questions appear under each answer. Clicking one puts it in the box, ready to edit or send.
At the Reasoning level, a folded section called How it read your question shows the AI's reasoning and assumptions. At Expert — SQL it is called Generated SQL and reasoning and also contains the query itself. Most people never need these.
AI CaseQuery checks how each answer was worked out and tells you when something deserves a second look. These notices do not say the figure is wrong. They say what could make it wrong, so you can decide.
An amber folded panel under the answer, with a heading such as "Worth checking: …". Quieter notices read A note about this answer. Open the panel to see what was noticed. Common examples:
Some answers are checked by planning the question twice, independently. You will mostly see this on questions opened from Fine-tuning. If the two attempts disagreed, a panel headed Asked twice — and read two different ways says how they differed. The answer shown is the first of the two. If the attempts agreed, a single quiet line says so under the answer.
Most Worth checking items have this button. It opens a short guided dialog of four steps. Nothing is decided until the third step.
If one of the items under How this was measured is not what your firm means, correct it where you see it.
Where it goes. Your correction goes to the top of Fine-tuning › Definitions to settle, where a lead can settle it. It holds your question, the measurement and what you wrote, and it is kept by your firm's CaseQuery service under your sign-in name. The correction is not sent to the AI, and it does not change the answer already on your screen.
Draft a rule… turns a correction into a standing instruction. That is an administrator's action. If you are not an administrator, leave it alone.
Words like "active", "open" or "outstanding" can mean different things to different people. The people doing the work usually know what the firm means, so AI CaseQuery can ask them.
A short question appears under the answer, above the table, when both of these are true:
It is not asked about answers from a Verified query. It is also not asked when the answer used filters such as "anything except", "one of" or "like", when it filtered on numbers or yes/no flags, or when it filtered on more than one status column.
The question reads, for example: Counted "open" as Open. Is that right? There is a button for each value in that column, with the one that was used marked "— yes", and a Not sure button.
After you answer, the question is replaced by a line saying what happened. For example: "Recorded. 3 more people answering the same way would settle it." Your vote does not change the answer on your screen.
Everyone signs in as themselves, so your vote carries your name: a line under the question says that your name and your answer are visible to whoever settles definitions.
If your administrator has left this to each person, the question includes a Stop asking me link. Click it and you will not be asked again while you stay signed in. The vote in front of you can still be cast. To be asked again, tick Ask me what words mean beside the question box (see section 10). If your administrator has switched on Everyone is asked, the link does not appear.
Sometimes you will see only a line such as: Counted "active" as Active. There is nothing to click. It tells you which reading your answer used. You see it, for example, when you have switched asking off, or when the column holds 7 or more values. If the reading is not what your firm means, report it with Not right? as in section 7.
Your firm's connector may offer more than one database. Each chat is with one database, with one set of databases your administrator has combined, or with two databases your administrator has linked.
When there is a choice, a Database list appears beside the question box. It is the one place to choose what you are asking: a single database, a combined entry, or a linked pair (section 9a). Choosing another starts a new chat, because nothing said about one database carries over to the next. If you reopen a past chat that was with a different database, it stays readable, and a note says that your next question starts a new chat.
Choose a combined entry, such as "Cases + Billing", in the Database list, and ask. The answer is labelled with the databases it asked.
Answers across databases are simpler than ordinary answers. There is no estimate dialog and no Worth checking panel. Choose a single database in the list to go back to asking it on its own.
A link ties records in one database to related records in another: the records in one, say, and the invoices raised against them in the other. Across a link you can ask for a count of records, or a total of an amount, hours or another quantity, narrowed by what is true on either side and broken down by a category or by year, quarter or month. For example: what is the total invoiced on open records this year, by type?
In the Database list beside the question box, the two linked databases are one entry, named A + B after the two, such as "Records + Billing", beside the single databases. Choose it. As with any change of database, a new chat starts. Then type your question in the question box and press Run, as you would of one database. Your firm's connector reads both databases where they are and matches them.
Writing a good question:
You cannot ask for a list of records, for anything about a single record, or to filter or group by the reference that ties the two databases. A question is up to 1,000 characters. If it cannot be put as a count or a total over what the link permits, the reply says so in one sentence.
Two databases can be tied by more than one link, for example by a record's number and by an invoice's number. If CaseQuery cannot tell which one your question means, it asks back in one line, naming the links, and ends Ask your question again with that in it. Ask the whole question again, saying which.
The answer starts with a short title and one sentence that says what was counted or totalled, across how many linked records, in which two databases, and under which conditions: for example, Across 1,240 linked records in Records and Billing where records.status is "Open": the total of invoices.amount is 312,450.00, over 3,105 records. Fields are named as table.field, as your databases name them. A breakdown follows as the answer's table, one row for each group: the group's name (the value as your database holds it, such as each status or type, or a period such as 2025, Q3 2025 or July 2025; (empty) where there is none), how many linked records it covers, and its count or total. Beneath the answer, From: names the pair it was asked across.
Notes under the answer say what it covers:
A total can be read from many records and still describe one. To stop figures being used to find out about a single record, some are held back, and the answer says why:
| What you see | What it means |
|---|---|
| not shown: fewer than 5 | Fewer than 5 linked records are in it. When the whole answer is that small, it says so instead of giving the figure. |
| not shown: one record carries most of it | A single linked record makes up most of the figure. |
| not shown: it would reveal a group that is not shown | Seen only on an answer from an older connector, which held back a second group beside a small one. Today only the small group itself is held back, and every other group is shown. |
| not shown: with an answer given in the last 30 days, it would reveal a figure for fewer than 5 | Together with a figure already shown to someone at your firm in the last 30 days, it would let a figure for fewer than 5 linked records be worked out. It never says which earlier answer. Asking the same question again is not affected. |
Ask a broader question, for example a whole year instead of a month, or with fewer conditions, to see a figure.
Your firm's AI provider turns your question into a plan, using names only: your question, today's date, the links' names, your firm's word for the kind of record, and for each database the table, the field that ties them, and the fields you can use. It is not sent your answer. Your firm's connector reads both databases, matches them and works out the figures; the answer goes to you and is kept only in your own browser, in your chat history, like any answer.
| What you see | What to do |
|---|---|
| The breakdown has more than 50 groups | Ask for a narrower breakdown, or one grouping instead of two. |
| The question would read more than the connector allows | Narrow it, for example to a period or a status. |
| Questions across linked databases have been asked many times recently | Your firm's or your own allowance is used up for now. Try again later. |
| A database, or its Schema rules, changed since the link was accepted; or the connector does not hold the link, or holds a different version | Tell your administrator. The link needs looking at under Settings. |
| The connector needs updating | Tell your administrator. |
| No linked pair in the Database list | Your firm has no link ready to ask across, or its databases are not offered in Chat. Ask your administrator. |
Next to the question box, marked Yours only, are choices that affect only your screen. Each one appears only if your administrator has left it open to you. They last while the page stays open: when you sign out, or reload the page, they go back to your firm's settings.
| Choice | Options | What it does | When you see it |
|---|---|---|---|
| Explanation | Plain · Standard · Reasoning · Expert — SQL | How much of each answer you see. Plain: the answer only. Standard: adds How this was measured. Reasoning: adds the AI's reasoning. Expert: adds the SQL it wrote. Warnings show at every level; which business rules were not used shows from Reasoning up. | Unless your administrator has switched on Everyone uses this level. |
| Columns | Supporting · Requested only | Supporting adds supporting columns to the ones you asked for, which make answers easier to read on screen. Requested only returns only the columns you asked for, with no supporting figures, joined-up names, sorting or rounding. Use it when you will export or reconcile figures line by line. | Only when your administrator has set Result columns to Each person chooses. Hidden while Ask Casey is ticked. |
| Ask me what words mean | Ticked or not | Ticked: you may be asked what a word means under an answer (section 8). Unticked: you are not asked, but you still see which reading was used. | Only when On Floor Voting is on and Everyone is asked is off. Hidden while Ask Casey is ticked. |
Your last 30 conversations are saved automatically in this browser only, under your firm and your sign-in name, so someone else signing in on the same browser sees their own and not yours. They are not kept on the CaseQuery service, and they will not appear in a different browser or on another computer.
Large result tables are saved with only a shortened preview of their rows. Clearing your browser's data also clears your history.
These are set in Settings, which only administrators can open, so you cannot change them yourself. They are your firm's settings, the same for everyone at it, whichever computer or browser they use.
Casey is "Your context-aware CaseQuery assistant". Casey explains how AI CaseQuery works: what a warning means, what a rule is, how to word a question. Casey does not answer questions about your data, never runs a query and changes nothing. To ask Casey, click Ask Casey in the sidebar, press Ctrl+K, or tick the Ask Casey box beside the question box, where the button then reads Ask. The sidebar button and Ctrl+K open the Casey panel, which keeps one conversation as you move between screens; a question asked with the tick box is answered in the chat instead, and does not join the panel's conversation. Casey never sees your query results. Casey is only available if your administrator offers it, and questions to Casey use your firm's AI provider account. For everything else, including what Casey can read and its limits, see Working with Casey.
No row of your firm's data is ever sent to the AI provider.
Your firm uses its own account with one of four AI providers: OpenAI, Anthropic, AWS Bedrock or Microsoft Foundry. Your administrator chooses which.
The service keeps what your firm writes (rules, definitions, notes, settings), its people and their roles, and an audit journal of what was done and by whom. It never keeps rows or results from your firm's database: your results pass through it to your browser and are not stored. The only values from your database it keeps are codes your firm chooses to keep in a note, definition or rule. Nothing about a database's data is read until your administrator has saved its Schema rules.
No. Your firm's connector opens the database read-only and checks every query before it runs, and anything other than reading is always blocked.
No. Open your firm's CaseQuery web address in a browser and sign in.
Sessions end after 30 minutes without use, and after eight hours in any case. Sign in again. Your chat history is kept in your browser and is still there.
Each answer is worked out afresh from your wording, and a question that can be read more than one way may be read differently next time. Check How this was measured and any Worth checking panel. Answers marked Verified query are the exception: they run the same checked query every time.
Your administrator has set one level for everyone, so there is nothing for you to change.
Either your administrator has turned the result table off, or the question returned no rows. If the table is there but folded, click its heading to open it.
Yes, if your Explanation choice is open to you. Set it to Expert — SQL.
No. It goes to your firm's settling list. Once a lead settles it, future answers change.
History is kept only in the browser. It cannot be recovered from the service, because it was never stored there.
Those are for leads and administrators, and your role decides what you can do. Use the option that passes the question on instead. If you think your role is wrong, ask your administrator.
Questions use your firm's own AI provider account. Answers from a Verified query cost nothing.
| What you see | What to do |
|---|---|
| A figure you know is wrong | Read How this was measured, then use Not right? on the item that is off (section 7). |
| An amber Worth checking panel | Open it. If the reading used is not what your firm means, use Help me sort this out… and pass it on. |
| The answer seems to answer a different question | Reword it more precisely (section 4). If it keeps happening, report it with Not right? and pick It answered a different question. |
| Database not connected, or a message that the connector is not connected | Tell your administrator. Nothing on the chat screen will fix it. |
| A message that your firm has no AI provider set up | Tell your administrator. |
| This question cannot be sent | Reword the question without the specific value (section 4). |
| The sign-in page, saying your session has ended | Sign in again (section 2). |
| Signing in is refused although your password is right | Wait a while after several failed attempts, and check you are on the office network or VPN if your firm requires it. Otherwise ask your administrator. |
| Questions fail only when you are away from the office | Your firm may allow its data to be read only from its own network. Connect to the office network or VPN. |
| "Request could not be completed" | Try again once. If it happens again, tell your administrator and give them the question, the time and the version number from the sidebar. |
| A figure across linked databases says not shown | It covers too few records, one record carries most of it, or with an answer given in the last 30 days it would reveal a small group. Ask more broadly (section 9a). |
| A choice described in this guide is missing | Your administrator has not switched it on, or has set it for everyone. See section 10. |
Whenever you report a problem, include the version number shown at the bottom of the sidebar.