AskEdi

AskEdi lets you ask questions about your connected data in plain language and get instant answers, charts, and recommended actions. You choose what context the AI receives, Edilitics runs the query against your source, and the result comes back as a narrative, a visualisation, or a structured recommendation. No SQL required.

---

Before You Start

AskEdi requires AI Column Insights to be generated for any table you want to chat with. Column insights give the AI the semantic understanding of your data it needs to produce accurate, relevant answers.

---

How a Chat Works

A chat is a conversation scoped to one table. You choose the table, the columns to include, the privacy mode, the AI provider, and the language once at creation. Every question you ask in that chat works within those settings.

When you submit a question, AskEdi first determines what kind of answer is needed. Questions that can be answered from data produce a query, which Edilitics executes against your source in real time. The AI then receives the result, along with the context your privacy mode approved, and writes the response. Questions that fall outside what the selected table and columns can answer come back as a plain-language explanation with guidance on what to ask instead.

The AI never connects to your database directly. Queries are validated as read-only before they run.

---

Response Types

Every response has an action bar beneath it. What appears in the bar depends on the type of response.

Narrative responses cover questions about trends, comparisons, drivers, and summaries. The response is structured text with a heading, supporting analysis, and a conclusion.

Decision Intelligence responses appear when you ask what to do or where to focus. Instead of a descriptive answer, AskEdi drills into the strongest opportunity or risk in your data and ends with a concrete recommended action, supporting evidence, a scenario-based upside or risk, a confidence level, and a validation metric.

Chart responses appear when a visualisation communicates the result better than text. AskEdi generates the chart automatically and includes a brief description of what it shows. You can open the data table behind the chart, export the chart as an image, or export the underlying data as CSV.

Out-of-scope responses appear when the question cannot be answered with the selected table and columns. AskEdi explains why and suggests what to ask instead. No analysis credit is consumed.

Response action bar

Follow-up suggestion

At the bottom of every response, AskEdi suggests one related question based on what it just found. Clicking it submits that question immediately and consumes one analysis credit, the same as typing it yourself. You can ignore it and type your own question at any time.

---

Starter Suggestions

When a chat opens, AskEdi generates four suggested questions tailored to the table and columns you selected. The suggestions appear under the Suggestions button in the input area. Two are narrative-type questions focused on trends, comparisons, or summaries. Two are chart-type questions suited to visualisation.

The suggestions are grounded in your actual schema and generated at chat creation time using your organisation's industry context and your role. They are a starting point. You can click one to submit it immediately, or ignore them and type your own question.

---

Persona-Aware Responses

AskEdi tailors its language, framing, and suggested questions to the role and industry on your profile. A CFO gets responses framed around margin, forecasting, and capital allocation. A Data Analyst gets analytical terminology and interpretability-focused answers. A Sales Manager gets pipeline and conversion framing.

The role used is the Job Title set on your user profile, combined with your workspace's industry. Both are set during onboarding and apply to every chat you create.

---

How to Start a Chat

---

Asking Questions

The input field accepts up to 800 characters. Type your question in plain language.

Type @ in the input field to open the column picker. Each column is shown with its name, data type, and the description written when column insights were generated. You can filter by column name or by description text, which is useful when you remember what a column contains but not what it is called. Selecting a column inserts it directly into your question. This matters because multiple columns can hold similar-sounding data: inserting the exact column name removes ambiguity and gives the AI a precise field to query rather than having to infer from your words. In Private mode, the picker also shows the anonymized alias alongside the real name so you always know what you are referencing.

You can also use voice input. Click the microphone icon to start recording. AskEdi transcribes your speech into the input field. Review the transcription before submitting. Voice input never submits automatically.

---

The Chat Card

Every chat you create or have access to appears as a card on the AskEdi home screen. The card shows:

• The AI provider used for the chat
• The number of messages in the conversation
• The language the chat is in
• A preview of the last message or the most recent follow-up suggestion
• The source integration and table the chat is connected to
• The privacy mode the chat was created with
• When the chat was last active
• The creator's initials (on shared chats you did not create)

Hover over a card to reveal the action menu.

Card actions

---

Continuing a Previous Chat

AskEdi preserves every chat. Open a previous chat from the home screen to read through the full conversation. If you are the owner, click Continue Analysis to pick up where you left off. The table, columns, privacy mode, provider, and language settings remain exactly as you set them when the chat was created.

---

Privacy and Governance

The AI never accesses your database. It receives only the metadata and context your privacy mode approved. Your source credentials are decrypted only at query execution time and are never sent to the AI.

Every query is validated as read-only before execution. Queries must begin with SELECT or WITH. Any query containing a mutation keyword or a semicolon is rejected with an error. This applies even if your underlying database connection has write permissions.

Data never leaves your source systems. Queries run live against your source. Results are returned directly to your session and are not staged, copied, or extracted.

One table per chat. Scoping each chat to a single table prevents cross-source leakage and makes every query fully auditable.

Every action is logged. Chat creation, questions submitted, mode selections, provider choices, and shares are all recorded with timestamp and user ID for governance review.

---

Analysis Credits

Each question that produces an analysis consumes one credit. Clicking a follow-up suggestion also consumes one credit. Out-of-scope responses do not consume credits.

Credits reset at the start of each billing cycle. On Team plans, credits are pooled across the workspace.

Adding your own API key (BYOK) doubles your workspace's credit allocation the first time a key is configured. Additional providers do not double again.

You can purchase non-expiring add-on credit packs from your workspace billing settings at any time. Once your monthly allocation runs out, AskEdi draws from purchased packs automatically.

See Analysis Credits for the full breakdown including add-on pack pricing.

---

Sharing and Collaboration

You can share any chat with workspace members as a view-only link. Shared viewers can read the full conversation, view all charts, download the chat as PDF or PPT, and export chart data as CSV. They cannot submit new questions.

---

Frequently Asked Questions

---

Next Steps

Need help? Email support@edilitics.com with your workspace, job ID, and context. We reply within one business day.