Is 80% of Your Enterprise Knowledge Invisible to Agentforce? 🙈 External AI is the Key

Working smarter today means tapping into all your knowledge, not just the neatly structured data living in your Salesforce/CRM. We’re talking about the 80-90% of organizational wisdom that’s currently lurking in the shadows – the unstructured stuff. Think documents, web content, presentations, even that insightful (but unsearchable!) video from last year’s sales kickoff. 🙈.

Historically, wrestling this unstructured data into a usable format has been, well, hard.

Enter Agentforce in Salesforce and you might be looking at a winner. Tools like Einstein Data Library and Unified Knowledge make it super easy to bring in external documents to Data Cloud. All that knowledge is just one hop away to be used by Agentforce 🕵️‍♀️.

The Challenge: Unstructured Data, Meet Salesforce #

Let’s be real for a moment. Imagine your organization’s digital library - the GB/TB’s of data. Now imagine uploading all of that into Salesforce or Data Cloud. The “enticing” promise of unlocking knowledge starts to feel more like the promise of T2 being the last Terminator, and yes, there is also the matter of duplication, hefty storage and AI processing bill.

This is where external AI strides in, like a knight in shining (and surprisingly cost-effective) armor. The beauty of this approach lies in its simplicity and its ability to leave your data where it is - at least for the most part. Think of it as building a super-powered librarian who can access and understand your existing document repositories without needing to physically move all the books.

Here’s the magic formula:

• The Connector (Middleware/Custom Program): We create a smart little bridge. This automation syncs your documents from their source (think Google Drive, SharePoint, even your website!) and converts them into “embeddings” within a Vector database. Don’t worry about the tech jargon – just think of it as creating a super-efficient index of your document content.
• The AI Oracle (External AI Service): We bring in the AI brains! Any reputable AI service will do the trick. This service takes a question (your prompt) and dives into the vector database, pulling out relevant information from those embeddings. It’s like asking a super-smart search engine that understands context, not just keywords.
• Salesforce Integration + Topics /Actions in Agentforce: Now, the seamless part! We set up a topic or action within Salesforce (Agentforce is perfect for this!) to call this external AI service whenever needed.

The Flow in Action:

1. User Asks: A sales rep needs to quickly answer a customer question about a specific product feature, buried somewhere in a product manual.
2. Agentforce Detects: Agentforce recognizes that this question might be beyond standard Salesforce knowledge.
3. External AI Summoned: Agentforce calls upon our external AI service, passing along the relevant question.
4. AI Oracle Responds: The external AI sifts through the document embeddings, finds the answer, and sends it back.
5. Agentforce Delivers: Agentforce elegantly summarizes the AI’s response, presenting it in a clear, Salesforce-friendly format to the sales rep. Boom! Customer question answered, knowledge unlocked. 🚀

e recently created a bot that answers questions about our Salesforce practice directly from our website content and a few PDFs in Google drive.

1. Pre-requisite: N8N service to sync the documents from Google Drive to Vector embeddings on Pinecone.
2. Create a service to leverage Gemini Flash 2.0 model and Pinecone API to get the answer for the question.

Was it perhaps a tad over-engineered for a proof of concept? Maybe. But hey, someone had to push the boundaries of external AI integration, right? 😉

Now, for “The Catch” – Because nothing is ever truly free lunch (even in the AI world):

• Data Privacy Dance: Crucially, the external AI isn’t directly accessing your Salesforce data. Agentforce is the gatekeeper, carefully deciding when to tap into external knowledge. This is a win for security, but trust in your chosen external AI service becomes paramount. (More on trust principles another day!)
• Wild West of AI Trust : External AI models operate outside the cozy confines of Einstein Trust Principles. We’re venturing into the open frontier, but rest assured, we’re in good company. This is a challenge being tackled at scale across numerous platforms.
• AI Agent Army: Yes, you’re now running multiple AI agents in your ecosystem. Embrace it! This is the future. Think of it as building a well-orchestrated team, each AI agent playing to its strengths.

What does this mean for you and your business? #

For Professionals: Embrace the change! Skills in AI integration, data strategy, and understanding the nuances of both Salesforce and external AI ecosystems will be hot commodities. Start experimenting, learn about vector databases, and play around with AI APIs.

For Businesses: Don’t let your unstructured knowledge be a hidden liability. Explore external AI solutions. Start with a pilot project. Imagine the possibilities: faster onboarding, empowered agents, richer customer interactions, and finally, unlocking the true potential of that 80% of knowledge you’ve been sitting on!

The world of AI in CRM is evolving at warp speed. Innovation and adaptability aren’t just buzzwords; they are survival skills. By embracing solutions like external AI for unstructured data, we can move beyond simply managing data, and truly unlock the knowledge that drives real business value.