Your bot queries live business data

A well-built knowledge base is the floor of a useful chatbot. It captures your tone, your policies, your product catalog at a snapshot, your shipping zones, your warranty terms. NebulaHex’s knowledge base handles that side. But there is an entire class of question no static document can ever answer correctly.

“Where is my order?” depends on the order. “Is this in stock?” depends on what shipped this morning. “When is my appointment?” depends on a calendar your customer can’t see. “What’s our spec for the X-2 SKU?” depends on whichever revision your operations team last filed. None of these answers belong in a knowledge base. They belong in the system of record, and they need to be fetched when the question is asked.

Live data connectors take a different approach: the bot doesn’t store your operational data at all. It queries the source on demand, returns what’s true now, and forgets it the moment the conversation ends.

Two surfaces are shipped today, a third is on the roadmap. You configure them in the Live Data picker on your bot’s knowledge tab.

You can run multiple surfaces on the same bot. Most growing businesses connect Google Sheets for the operational data they already maintain there (member directories, internal lookups, custom catalogs), plus Shopify if they run a store, plus the regular knowledge base for tone and policies.

Most small businesses run on a spreadsheet. Order tracking, member directories, inventory counts, appointment slots, internal contact lists, project status — somewhere in your business there is a tab in a Google Sheet that is the single source of truth for something operational. The Google Sheets connector lets your bot read that sheet in real time so it can answer questions grounded in whatever is in the cells right now.

You connect a Google account on the Live Data picker. Authorization happens through Google’s own login screen, with read-only scope— NebulaHex never gets write access to your sheets, and you can revoke access at any time from your Google account settings. Once connected, you paste a spreadsheet URL into the connector card, click Refresh from Sheet, and the dashboard pulls in the column headers from your first row. From that point, you’re configuring the connector, not the data.

There are four sections on the connector card, and each one corresponds to a decision you want to make once and forget. The defaults are sensible. The controls are explicit. The dashboard helps with the parts that are easy to get wrong.

Most teams get a connector configured in five to ten minutes. The longest step is usually deciding which columns count as sensitive — the dashboard’s warnings catch the common mistakes, but the judgment call about “is this column safe to share” belongs to you.

A visitor on your site asks for their order. The bot recognizes the lookup, asks the verification question, waits for the answer, runs the lookup, and replies with the row. End-to-end, the conversation looks like this.

If the row doesn’t exist, the bot says so honestly instead of inventing one. If the verification answer doesn’t match, the bot doesn’t show the row — it apologizes, prompts again, and after a small number of failed attempts it stops responding to the lookup so a casual probe doesn’t become a brute-force check. Sensitive lookups are throttled by built-in rate limiting so a single visitor can’t hammer the connector with guesses.

None of this requires you to write any safeguards yourself. The connector ships with the locks engaged.

Shopify is the other shipped live data surface, but it works through the dedicated Shopify integrationon your bot’s Integrations tab rather than the Google Sheets-style configuration card. You connect via Shopify’s official OAuth flow, choose which of the eleven store actions you want the bot to perform, and you’re live.

Once connected, the bot can pull a specific order by number and read back the status, ship date, tracking link, and items inside. It can search your product catalog by name, type, or attribute. It can check whether a SKU or variant is in stock right now. It can pull a customer-level record when the visitor identifies themselves with the email on file.

The point is the same as with Google Sheets: the answer the bot gives is the answer that’s true in your store right now. Stock levels are read from the live store. Order status reflects the current state of fulfillment. The bot doesn’t keep a shadow copy of your catalog — it reads the source.

There is a meaningful gap between “trained on your data” and “queries your data.” Most chatbot vendors describe the first and quietly hope you won’t ask for the second. NebulaHex ships both. The static knowledge base does what static knowledge bases do: captures your policies, your tone, your product details. The live data layer fills in the dynamic answers — the ones that change every minute and have no business being embedded.

The combination is what separates “a search engine over your help center” from “a real first-line support agent.” A bot that can read your shipping policy and look up the customer’s actual order is doing the same work an agent would do. A bot that can quote your sizing guide andcheck current stock has finished the conversation that the agent would otherwise have to finish. That’s the productivity gain — and it does not exist if the bot is restricted to static knowledge alone.

Live data lookups touch real customer data, so the security posture matters. Connection credentials — the OAuth tokens that let the bot read your sheet or your store — are encrypted at rest. They’re never displayed back in the dashboard once saved, never written into transcripts, and never used to retrain models.

The bot reads only the specific record relevant to the question being asked; it does not sweep your sheet or your store and ingest the contents. Access is read-only by default, and the visible-fields whitelist is the second layer of defense. The bot cannot share a column you didn’t tick, cannot search by a column you didn’t tick, and cannot guess at fields that aren’t in the schema.

Verification questions are a real second factor for sensitive lookups, and the lookup flow is rate-limited so brute-force attempts on a verification field hit a wall quickly. Disconnect from Google or Shopify at any time, from inside Google/Shopify or from the connector card, and the bot loses access immediately. Role-based access control inside your team governs who can configure connectors. See our privacy policy and terms for the full posture.

If your support team is repeating the same lookup all day, a live data connector is the lowest-effort, highest-value thing you can do with NebulaHex. You don’t need to migrate anything. You point the bot at the spreadsheet or the store you already maintain, decide what visitors can see, and the bot does the rest.