Agentforce Agent and Unified Knowledge
Below are experiences of the authors, and referencing features in pilot and actively updating every month
One of the most exciting innovations for me in the last 12 months has been Salesforce Unified Knowledge, because of the power it has to truly revolutionise AI automation in Customer Service.
I figured since I’m writing some blogs about Agentforce, this was an obvious candidate. Just a hunch, but I imagine we’ll see hundreds of companies going live and seeing value just from this one action alone, so it’s worth checking out!
What is it?
Knowledge content that agents use to get answers can be spread all over the place. Online help docs, a Google drive of pdf manuals, Articles in Salesforce or another tool, and much more. The Unified Knowledge connectors allow a company to merge all of that into a single searchable repository for Service Cloud users.
Why do we care?
Don’t get me wrong, it’s great to have good knowledge articles, but the impact this will have on Agentforce is what’s really exciting. Public API’s like ChatGPT are trained on google, and could be using data from anywhere in their answers, including a competitor, but an Agentforce Service agent will be trained off of a Companies knowledge base, allowing it to speak to the products, policies, brand, and tone of the company. So the more data we can get it, the better. Coupled with Einstein Knowledge Creation, which documents an article every time an agent has to answer a new question, we can get pretty fast to a world with an Ask-Me-Anything Service Agent.
What is a good Example?
I recently built a unified knowledge repository off of all the content on help.salesforce.com and unofficialsf.com. I spent an hour asking every setup question I could think of, on web chat and sms, and barely got a bad answer. Though I think I need to add an instruction to shorten the responses 😀
c
But how is this different from the Knowledge action on Bots?
Really key question, we shipped a Knowledge Action for Einstein Bots a couple of months ago, what’s different about Agentforce?
Both of the above examples are now possible by adding the Gen AI Knowledge action to an Einstein Bot, so it’s pretty powerful, but what makes an Agentforce Agent different is the ability to be conversational.
I’ll flag two examples:
- I can ask back and forth questions, get clarification, dig into certain sections, or ask a follow up but with some different data.
- It can use the context of the entire conversation, and data pulled back from actions, to inform the response that’s generated from the Knowledge base
Let’s say I have a knowledge base for a Travel Agency, with articles on every airline, and where every policy has a different answer based on whether I’m a Platinum, Gold, or Silver member. With Agentforce, if I ask about the cancellation policy for me, and it knows I’m a platinum member from my contact record, or earlier in the conversation, it’ll give me a personalised answer. What’s more, I can ask follow-up questions, and it’ll know the context of the previous answer and respond in kind.
See an example of a conversation I had below, and the Knowledge base it searched against to do it, separating out info about the airline, the Loyalty level, and the time of the flight.
Have a try yourself and see what fun conversations you can have???
And then maybe have a try yourself and see what fun conversations you can have? 🚀
————
