Answer from Salesforce during a chat

If your business has reached the point where you’re running Salesforce, you’ve made a decision about where the truth lives. Account hierarchies, opportunity stages, custom objects modeling your specific workflow, reports that leadership reads every week — that infrastructure isn’t easily moved, and it shouldn’t be.

The cost of any new tool you adopt isn’t just the tool itself. It’s the cost of the data drift between that tool and Salesforce. Every system that doesn’t talk to your CRM creates a reconciliation tax that your RevOps team eventually pays.

That’s the design constraint NebulaHex’s Salesforce integration is built around. We don’t ask you to mirror your CRM inside another database. We don’t ask reps to learn a second tool. We connect the conversations to the records you already trust, in the system your team already uses. The bot becomes a participant in the same data fabric, not a parallel one.

Once you connect a Salesforce org to a bot, the Configure Tools panel lists ten CRM tools; this page shows seven of them, in three categories. Four run during a chat today, all of them reads, and like every integration tool they run only when your agent answers with a Claude model. The three writes are refused even when switched on; they are coming soon, each for a visitor who proves their email address with the code we email them. The other three reads, which this page leaves out, are locked off and stay off. A connected org with every tool disabled does nothing. That’s by design.

The bot uses these lookups inside live conversations. When a customer asks about their case, the bot can look it up once they confirm a detail on the record. It does not drop Leads into Salesforce or open Cases there today: captured leads stay on your NebulaHex Leads page, and escalations go to your team in the NebulaHex inbox with the whole conversation, so the agent who picks one up isn’t starting from zero.

Salesforce shops take change management seriously. A serious org doesn’t make changes in production without testing them somewhere first. Salesforce sandboxes are isolated replicas of your org used to validate changes before they touch live data. NebulaHex connects to a Sandbox as well as to Production: you choose one when you click Connect, and the same lookups run in either.

The same OAuth flow handles both. The Salesforce card shows the hostname of the org you connected to and a Sandbox badge if applicable, so there’s never any ambiguity about which environment the bot is touching.

The connection flow is short by design.

Disconnection is just as simple. Click Disconnect on the card and the bot stops accessing your CRM immediately. Disconnecting deletes nothing in Salesforce: your records stay there, because that data belongs to your org, not to NebulaHex. Conversations the bot already had stay in your NebulaHex inbox. The link between the two systems is the only thing that goes away, and you can re-establish it any time.

Different roles get different value from the integration. Here’s how it lands for the teams who actually use Salesforce day-to-day.

Salesforce is the largest customer relationship management platform in the world. Sales Cloud has been the leading sales CRM for over two decades, and the broader Salesforce platform has grown into a suite that covers sales, service, marketing, commerce, analytics, and a developer platform that thousands of companies extend with their own logic.

Two things make Salesforce a good target for an AI bot integration. First, the object model is well-defined and predictable: Leads, Contacts, Accounts, Opportunities, Cases, Knowledge Articles. Each has a clean, predictable schema; today the bot reads Contacts, Cases and Knowledge Articles. Second, Salesforce’s reach inside revenue and support orgs is wide enough that answering from it is leverage — every question answered from a live record saves time across whatever workflow your team has already built.

What NebulaHex adds is the front door. Your knowledge base, your products, your past Cases — combined into a chat-shaped agent that can answer most questions and steps aside when a visitor taps Talk to a human in your website widget, on a bot with live chat handoff switched on (it is off by default).

Plenty of teams have considered building their own Salesforce-aware support agent. The shape of the project always looks the same: stand up a chat surface, plumb it into the Salesforce API, build a knowledge layer, manage authentication and refresh tokens, handle conversation state, train people on the new internal tool, maintain it forever. It’s months of engineering before the first useful conversation, and the result is a bot only your team knows how to extend.

NebulaHex is shipped. Connecting Salesforce to a knowledge-grounded bot is a few clicks, not a quarter. You get a chat surface that’s already polished, a knowledge ingestion pipeline that already handles URLs, files, and structured Q&A, a widget that’s already brandable, an inbox where your humans can take over conversations, an analytics view that shows which conversations are converting, and integrations with the rest of your stack that work today.

You also get honest scope. The bot resolves what it can automatically, your team handles the conversations a visitor hands over by tapping Talk to a human in your website widget, on a bot with live chat handoff switched on (it is off by default), and customer satisfaction goes up because your team isn’t drowning in repetitive questions. The Salesforce integration fits inside that frame: it makes the bot’s automatic work better. Coming soon: a Salesforce Case for each of those handoffs, to make the human handoff cleaner too.

Your Salesforce org contains some of the most sensitive data your business owns. We treat the connection with that in mind.

Connections use OAuth 2.0 against Salesforce’s own login flow — you authenticate against Salesforce, not against us, and your credentials never pass through our system. The access tokens we receive are encrypted before they’re stored. All data is encrypted at rest and in transit. Disconnecting from the Salesforce card revokes access immediately on our side; revoking from the Salesforce side does the same.

Your conversations and your Salesforce data are never used to train AI models. Each customer’s data is isolated to their own workspace. See our privacy policy and terms for the full posture.

If your business runs on Salesforce, your AI agent should know it. Pick a bot, open the Integrations tab, click Connect on the Salesforce card, choose Production or Sandbox, sign in to Salesforce, approve, and check which lookups are on. You’re done.

Start free. Build the bot, ingest your knowledge, and connect Salesforce, in Production or in a Sandbox.