Question 1

Our data is a mess. Is it too early for us?

Accepted Answer

That is the normal starting point, and it is what the first two steps are for. Connect a source and every column is graded A to F for completeness and consistency before you do anything - so you can see exactly what is wrong rather than guessing. Transform then fixes it by pointing and clicking: filter, join, group, repair, and build new columns worth analysing, with the grade updating after every step so you can watch it improve. You do not need clean data to start. You need it to be clean before you trust an answer, and that is the part Edilitics does with you.

Question 2

What if the AI gets the answer wrong?

Accepted Answer

Three things make that checkable rather than something you have to take on faith. Every answer shows the exact query that ran against your database - you can copy it, run it yourself in any database client, and get the same number back. Every answer states the sample size and the statistical test behind the figure. And every column carries a quality grade, so you know what the answer is standing on. The AI writes the query; the numbers are computed in code, not generated by the model. When your data cannot support a conclusion - a trend that fails its significance test, for example - the number is withheld and the reason is given, instead of a confident figure with nothing underneath it.

Question 3

We already have a BI tool. Why would we use this instead?

Accepted Answer

Most BI tools start after the hard part. They assume something else already connected your sources, cleaned them, and decided what each column means - typically a separate ingestion tool, a separate transformation tool, and a data engineer to run both. Edilitics covers that whole path in one place: connect, clean, track, and ask. If your dashboards already work and the only gap is ad-hoc questions, keep them and use AskEdi alongside. If you are still assembling the layer underneath them, that is the larger problem this solves.

Question 4

Do we need a data team to set this up?

Accepted Answer

No, and that is the point. There is no SQL to write, no pipeline code, and no query language to learn - transformations are built by pointing and clicking through a list of operations that stack top to bottom, so anyone can read what happened to the data. The one task worth a careful eye is approving what each column means, which takes under an hour per source and is usually done once by whoever knows the business best. After that, people ask questions themselves instead of filing requests.

Question 5

What does it cost to get a team started?

Accepted Answer

One Team licence at $89 per month covers a full builder seat and includes five viewer seats at no extra cost, so six people can be working on day one - the builder connects sources, cleans data and creates dashboards, while the five read those dashboards and ask their own questions in plain English. Individual plans start at $49 per month for a single user. Every plan includes all four parts of the platform rather than gating the AI behind a higher tier, and there is a 14-day evaluation with no card required.

Question 6

How long until this is actually working?

Accepted Answer

About fifteen minutes to a first real insight - that is our own measured run, end to end: connect a source or upload a spreadsheet, let the profiling grade the columns, apply the suggested cleanup, build a dashboard, and ask a question. Sample data is already connected the moment you sign up, so you can do the whole loop before connecting anything of your own. Larger schemas take longer to review, but you can approve a handful of tables and start there rather than waiting for all of them.

Question 7

Does Edilitics copy our data or move it anywhere?

Accepted Answer

No. Your rows are never extracted, synced, staged, or stored on our infrastructure. A question becomes a query that runs live against your own database, and the result travels to your screen. What Edilitics keeps is the shape of your data and nothing from inside it: column names, types, quality statistics, the descriptions your team approved, your dashboard layouts, the totals each chart displays, and the query behind each answer. If you cancel, your data is exactly where it was, because it never went anywhere else.

Question 8

Do our actual records ever reach an AI provider?

Accepted Answer

No, on every plan and in every privacy mode. The AI receives your column names, data types and quality statistics - enough to write a correct query - and never a value from inside those columns. Edilitics runs the query it produces; the results go from your database to your screen without passing back through the AI. Three privacy modes control how much structure is shared: Private replaces column names with placeholders before sending and restores them before the query runs, Balanced sends real column names, and Full Context adds the values that appear most often. None of the three sends a row of your data.

Question 9

What happens when the data cannot support an answer?

Accepted Answer

You are told, in plain words, along with what was missing. A forecast is only shown if the trend clears a statistical significance test - if it does not, the number is withheld and the reason is stated rather than a confident figure being invented. If a query genuinely returns nothing, you get a sentence saying so instead of an empty chart. And if your question is outside what the selected table can answer, you are turned back before any query runs, told why, and given questions that would work instead - with no credit charged.

Question 10

Can we check the working behind a number ourselves?

Accepted Answer

Yes, without asking anyone for database access. Every answer has a one-click view showing the exact query that ran - copyable and runnable in any database client, so a second person can independently reproduce the figure. Alongside it sits a methodology note stating the sample size and which statistical test produced the result. In the strictest privacy mode you see two panels: the anonymised query the AI wrote, and the rewritten query with your real column names that actually executed.

Question 11

How does Edilitics decide our data is good enough to trust?

Accepted Answer

Every column is profiled the moment a source connects and scored A to F on completeness, uniqueness and type compliance, with ID and timestamp columns weighted more heavily because more depends on them. A second score covers whether the data is ready for AI specifically: half from that structural health, half from whether a person has confirmed what each column means. A clean table still scores poorly if nobody has said what its columns are, because an AI cannot reason correctly about a field it does not understand. Low grades raise a warning before you start asking questions.

Question 12

What stops two people getting two different revenue numbers?

Accepted Answer

Metrics are shaped once and reused everywhere rather than redefined per chart. Most BI tools let anyone type a formula into a visualisation, which is how one company ends up with four versions of revenue - Edilitics has no formula box in the charting layer at all. A new metric is built once in Transform, lands in your own database, and every dashboard and every answer reads that same table. Definitions your team approves are locked and no later automatic refresh overwrites them.

Question 13

Can we clean and reshape data without writing SQL?

Accepted Answer

Yes. Twenty-five point-and-click operations cover filtering, joining, grouping, deduplicating, handling nulls, casting types, pivoting, window functions and more, stacked top to bottom in the order they run so a non-technical colleague can read the pipeline. You can describe what you are trying to build and get back suggested steps with the reason for each, applied only when you approve them. When something genuinely needs custom logic, a Python step can be dropped in at any position and you carry on clicking afterwards. Building and previewing costs nothing; the finished table is written to a database you choose.

Question 14

Where does the transformed data end up?

Accepted Answer

In your own database, not ours. During a run your data is pulled into isolated memory, transformed, written to the destination you selected, and the machine is destroyed - leaving Edilitics with the recipe and never the ingredients. Source and destination can be different databases entirely, and column types are converted to whatever the destination supports without you mapping any by hand. While you are building, previews run against an encrypted sample that is purged after 30 minutes of inactivity.

Question 15

Do we still need dashboards if we can just ask questions?

Accepted Answer

They do different jobs. Dashboards cover the numbers your team checks on a schedule - the standing view everyone reads the same way each week. Asking covers the question that just came up in a meeting and that nobody built a chart for. Both read the same approved tables, so the two cannot disagree. You can describe what you want to understand and approve the charts that answer it, have the whole board summarised in plain language, and compare this week against any saved earlier version to see what changed.

Question 16

Can we share dashboards outside our company?

Accepted Answer

Yes, three ways. Share inside your workspace with colleagues, send an external viewer a link they open with a one-time email code, or embed a dashboard in your own product with no login at all - protected by a signed token and a strict list of domains allowed to display it. Shared viewers can read everything and export it, but cannot change what you built or spend your credits. An embed can be switched off in one click.

Question 17

What are we charged for, and can the bill surprise us?

Accepted Answer

You are charged for completed analyses, not attempts, and the balance stops at zero rather than running over into an overage. Nothing is deducted if the AI provider fails, if it returns something unusable, or if your question was outside what the table could answer. Transform is billed by the seconds a pipeline actually runs - no charge per row, no charge per operation, and nothing at all while you build and preview. On Team plans everyone draws from one pooled balance, so a busy week for one person is not blocked by a quiet week for someone else.

Question 18

Can the AI show someone data they are not allowed to see?

Accepted Answer

No, and the reason is structural rather than a policy setting. A documented risk with AI analytics tools is that the AI queries the database through its own connection and bypasses the permissions that apply to human users. Edilitics does not work that way: the AI never receives rows at all, and the query it writes runs against your database using your own credentials and the permissions already enforced there. There is no separate AI access path with broader rights. Each conversation is also fixed to one table and the columns you chose, so nothing reaches across into data it was never pointed at.

Question 19

Why not just paste our data into ChatGPT, Claude or a spreadsheet AI?

Accepted Answer

Because none of them connect to your live database - they analyse a file you upload, which is already a sample rather than your real data, and it goes stale the moment the source changes. Whatever explanation you give about what your columns mean lives in that one chat thread, so the next person starts over and nobody can verify how a number was produced. Edilitics queries the live database directly, reasons against column descriptions your team approved once and everyone reuses, keeps them current through automatic profiling, and shows the exact query behind every answer so anyone can check it.

Question 20

What does Edilitics replace in a typical data stack?

Accepted Answer

Three layers with one product: the connector tool that moves data, the transformation tool that cleans it, and the dashboard tool that displays it. It connects to a warehouse you already run rather than replacing it. Assembling those layers separately commonly runs into six figures a year before anyone answers a business question, and each one needs somebody to keep it working. The market is consolidating the same way for the same reason - the ingestion and transformation vendors have been merging into single products.

Question 21

Why not hire an analyst instead of paying for a platform?

Accepted Answer

An analyst is worth hiring for judgement, not for being a queue. Without tooling underneath, most of their week goes on maintenance - fixing schemas, chasing data quality, patching pipelines and answering one-off requests that should not need an engineer. That is the work Edilitics removes: sources stay connected and graded automatically, cleaning is point-and-click, and business users answer their own routine questions. The analyst then spends their time on the analysis that actually needs a person. The two are complementary, not alternatives.

From a raw table nobody trusts
to a decision you can defend.

The Question Everyone Asks First

So how does our data get to you? It doesn't.

Every number lives in five places. So does every disagreement.

Where Everyone Is Stuck

Three ways to be locked out of AI. One reason underneath.

Step One - Integrate

Know what your data is worth before you trust
a single answer from it.

Read structure only

Grade every column

Describe every field

Only people get you to A

Everything agrees

Step Two - Transform

A bad grade tells you what is wrong. This is where you fix it.

Step Three - Visualize

The number your team checks every Monday, without rebuilding it.

Step Four - AskEdi

Stop asking where a number came from. It already tells you.

When It Doesn't Know

The most useful thing it does is refuse to answer.

The Cycle, Inverted

Test the hunch in seconds. Deep-dive only what earns it.

How You Get Started

Everyone starts by connecting. What comes next is up to you.

The Same Number, Twice

Revenue on the dashboard. Revenue in the answer. One number.

Where Your Data Stays

Pass your security review without an exception.

Private Mode

Balanced Mode

Full Context Mode

Everything you need to know before you decide.

Start with the table you trust least. End with a decision you can defend.

From a raw table nobody truststo a decision you can defend.