Ask It in a Sentence, Check the Definition
A question typed in plain words becomes a report definition you can read, run through the same engine as every report you build by hand, under the same permissions. The answer comes with its working, so you can check what was counted before you quote it.
The operations director of a pharmaceutical distributor in Nakuru wants one number before the 10 o’clock call: what did the top ten hospital customers buy last quarter? The report builder can answer it, in about seven clicks she has never learned. The analyst who knows the clicks is in Eldoret. So she types the question into a box, and the interesting part is not that an answer comes back. It is that the answer comes back with the definition that produced it.
A question becomes a definition
Ask your data, under Reports, takes a sentence and turns it into the same kind of definition the custom report builder writes: a dataset, the fields to show, filters, a grouping, a measure, a period, a sort and a row limit. That definition is shown beside the result in plain words. “Customer invoices, sum of total, grouped by customer, last quarter, top 10, highest first.” If the director meant net of credit notes, she can see that it did not, before the number goes into the call.
| You type | It reads as |
|---|---|
| Top 10 customers by sales last quarter | Ranked group: sum of invoice total by customer, last quarter, 10 rows |
| Invoice total by status this year | Measure by dimension: sum of total grouped by status, this year |
| How many open tickets this month | One figure: count of tickets, status open, this month |
Starter questions are listed under the box, grouped by module, and only for data you can read. The words it understands cover about thirty datasets across sales, POS, procurement, payables, expenses, helpdesk, projects, HR, payroll and assets, with the vocabulary people actually use: “turnover” and “billed” for invoices, “requisitions” for purchase requests, “headcount” for employees.
What the model is shown, and what it is not
Where your workspace allows third-party AI, a language model reads the question and proposes the definition. It is shown the schema of the datasets you may read: field names, labels and types. It is never shown a row of your data. And the schema it sees is narrower than the builder’s: only a dataset’s declared columns, never one marked sensitive, never a field whose own permission you lack, and joined fields only through a declared relation that carries no row rule.
Its reply is treated as untrusted input. A validator checks every dataset, field, join, operator and period against that list, and refuses the whole definition if anything is outside it, rather than quietly dropping the part it did not like. Nothing becomes a database query except through the report builder’s own payload. Why the builder works only from declared datasets, and never from raw SQL, is in who is allowed to ask a new question.
The model writes a proposal. The validator decides whether it is allowed. The report engine decides what it returns.
When AI is off
Some workspaces switch third-party AI off, some switch it off for one module, and every workspace has a monthly allowance. None of those stops the box working for the common shapes. A rule parser reads the three shapes in the table above without any model, produces the same kind of definition, and passes it through the same validator, so the two paths cannot differ in what they are allowed to read. If a question is not one of those shapes, it says so and offers the builder. It never picks a dataset it found no word for.
The same fences as a report built by hand
- Row-level security applies. A branch manager who sees only her branch’s invoices in a report sees only those here.
- The query timeout and the plan’s row cap apply, and an answer shows at most 200 rows, 50 unless the question asks for more.
- Asking needs the report view permission; opening the definition in the builder or saving it as a report needs the permission to create reports.
- Every question is logged with its definition, who asked, which path answered and how long it took. The rows are never logged.
That last point matters for the audit question nobody asks until later: who looked at payroll by department in March? The log answers it without having kept a copy of the payroll. If a figure from an asked question differs from the monthly pack, the definition beside it is where to start; we wrote about the usual reasons two reports disagree in the same report, different rows.
From a question to a saved report
An answer worth keeping can be opened in the custom report builder with the definition already filled in, or saved straight away as a report under a name you choose. The question becomes something the analyst can refine, schedule and share like any report.
The same service sits behind the API, and the in-app assistant uses it to recognise when a chat message is really a question about your own figures, the kind a report answers. More on the screen on the Ask your data page.
Three questions to ask any “ask your data” tool
Can I see exactly what was counted?
What you will probably hear
The AI worked it out.
How to read it
An answer without its definition cannot be checked, and an unchecked number travels further than a checked one.
Does it respect what I am allowed to see?
What you will probably hear
It uses an admin connection.
How to read it
Then it is a way round your permissions. It should run under the asker’s own rules.
What does the model see of our data?
What you will probably hear
It needs the data to answer.
How to read it
It needs the shape of the data. The rows can stay on the server.
Where this leaves you
A question becomes a report definition you can read, validated against the fields you may use, run by the report engine under your row rules and row cap, and logged without its rows. With AI off, the common shapes still work. Anything worth keeping becomes a saved report in one step.
What AWRA OpsHub does today
- Plain-language questions turned into a report definition shown beside the answer.
- A schema-only prompt: declared, non-sensitive fields you have permission to read, and never a row.
- A validator that refuses any unknown dataset, field, join, operator or period outright.
- The report engine’s own fences: row-level security, query timeout and the plan’s row cap.
- A rule parser for ranked lists, measures by dimension and single counts when AI is off.
- Every question logged with its definition, asker and path, without the rows.
- Open in the builder or save as a report, on the web and through the API.
More we can add to your workspace
- A chart chosen from the shape of the question, alongside the table. Commission it
- Follow-up questions that refine the last answer, such as “now only Nairobi”. Commission it
- Scheduled delivery of an asked question without saving it as a report first. Commission it
- Swahili phrasing in the rule parser, for teams who ask in the language they work in. Commission it
- One question across two datasets, such as invoices against payments by customer. Commission it
Where we point you to a specialist
- An answer is only as good as the records behind it. If an invoice was never raised, no question will find it, and the definition beside the answer is the way to see what was and was not counted.
Follow-up questions are the extension asked for most, because the definition the last answer used is already on the screen.
Anything above that you need, we can build for you
Everything listed above as something we can add describes what ships in the standard product today — it is a starting point, not a limit on what AWRA OpsHub can do for your organisation. Kenya's eTIMS integration and its maintained payroll engine are both in the product because clients needed them and commissioned them; neither appeared by itself, and the same door is open for whatever you just read about. One qualification so this is worth what it claims: a small number of things on this blog we deliberately leave to a specialist rather than build — a statutory ledger we will not sign our name to, a rule that would decide a tax question for you, a clinical or member-funds record that belongs in a regulated system — and where that is true the post says so in those words. Everything else is a scope, a timeline and a price.
The operational work, which is what most commissions actually are
An extra approval stage in a chain that does not match the standard one, a custom field set on employees or assets that only your sector needs, an expiry that has to block an order rather than send an email, a report your board asks for in a shape nothing produces, or a scanner or weighbridge feeding the goods-in door. These are the commissions we are asked for most often and the smallest ones we quote — and unlike a revenue-authority pipeline, none of them waits on a regulator.
The module-shaped additions, which are the ones readers ask for most often
A price list with real discount authority, a customer-facing quotation that expires, a bill of materials or recipe costing, a staff advance that is issued, acquitted and chased, a member or unit ledger, a matching rule that holds a payment. Each of these is a build rather than a setting, and each has been quoted before — a bigger piece of work than a custom field, with a written spec and a date instead of a roadmap slide.
The report, document or pack nothing currently produces
The board pack in the shape your board actually asks for, a donor or funder layout, an invoice or receipt template carrying what your regulator or your customer expects, a dataset the report builder cannot reach yet. Usually the fastest thing on this list to deliver, because the data is already in the system.
Systems, rails and hardware you already run
The accounting package, CRM, online store, core banking or custom database you intend to keep — connected through our API so a fact is entered once and appears everywhere it is needed. Plus the physical edge: a scanner, a scale, a weighbridge or a till peripheral feeding the door it belongs to.
How it works: you describe the requirement, we return a written scope, timeline and cost, and once agreed it is built into your environment and maintained as part of the product. Nothing here waits on a regulator or a published specification, which is why operational builds are the ones we quote fastest. Tell us the requirement that would otherwise rule us out — that is a better first conversation than a demo.
Tell us what your operation needsSee Ask your data
Type a question, read the definition, check the answer, keep it as a report.
Explore Ask your dataFrequently asked questions
Does the AI see our business data?
No. It is shown only the names, labels and types of the fields you are allowed to read. The rows stay on the server and are fetched by the report engine.
Does Ask your data work with AI switched off?
Yes, for the common shapes: a ranked list such as top ten customers by sales, a total grouped by a field, and a single count for a period.
Can someone see figures they are not allowed to see by asking a question?
No. Questions run through the same report engine with the same row-level security, and a field you lack permission for is not offered to the model and is refused by the validator.
How many rows can an answer show?
Up to 200, with 50 as the default, and lower if your plan sets a lower report row cap.
Can I keep an answer?
Yes. With permission to create reports you can open the definition in the report builder or save it straight away as a named report.