Getting started with Agentforce

One of the first points of discussion was how Agentforce is viewed.

Many people see Agentforce as an AI tool, similar to a chatbot. Agentforce is much more than this; it’s not just automation. An agent is not a chatbot – it’s the step between artificial intelligence and human capabilities.

So the first step to getting started with Agentforce, is understanding it’s potential.

To understand more about Agentforce, check out our introduction to Agentforce blog.

When it comes to AI as a whole, the panel agreed that it shouldn’t be seen as a new feature as it creates a new way of working. When embracing AI, you will need to look at how your organisation is currently running and where AI can be implemented. Your organisation will need to consider new workflows and outcomes.

When it comes to experimenting, you can’t simply jump straight in and start building agents.

Unlike Chat GPT, you can’t give Agentforce a few sentences of context for it to generate a solution. Agentforce needs to be fed data and detailed context.

Having detailed and defined processes is essential when getting started with Agentforce. This means providing a full end-to-end explanation of how work is being conducted – be explicit.

Based on your defined processes, you can see where gaps exist that agents can support.

When reviewing business processes, look out for bottlenecks, error hotspots, repetitive tasks, and human frustration.

Agentforce is brilliant at supporting with tasks that are limited by human capabilities. For example, AI can read and review documentation incredibly quickly, which would take a human hours. There may be tasks in your organisation where Agentforce can support with interpreting and referencing back to lengthy documentation.

There is often confusion between deployment risk and complexity. You can start with solutions that have low deployment risk, which shouldn’t cause major issues.

For example, you might look to deploy a sales agent that can be used to support training for staff. If something went wrong, this isn’t going to have major repercussions.

Whereas a task like deploying an agent to answer license-based queries on your public-facing website might not seem complex, but if the agent makes mistakes, it can cause significant problems.

Implementing Agentforce has its costs, not just licensing, but the costs for ongoing maintenance and investments in improving the work of agents.

So, once you have worked out your use cases, it’s important to define your KPIs and how you will measure the effectiveness of Agentforce.

The panel agreed that it’s important to have a combination of qualitative and quantitative KPIs:

• Quantitative KPIs might be time saved, number of cases resolved with Agentforce, revenue or cost savings.
• Qualitative KPIs include changes in efficiency, staff feeling, and customer satisfaction. You can measure these with tools like CSAT and NPS.

An important thing to measure is accuracy – whilst Agentforce might be answering things quickly, is it answering correctly?

It can be a daunting prospect to implement Agentforce, but it can have many benefits for your organisation.

You don’t have to roll out Agentforce across the whole organisation straight away. Start with a small project to get used to defining your processes and the work involved. This doesn’t have to be a project that you measure; it is a start towards understanding the possibilities with Agentforce.

The first step to getting started is to just get started!