If you are a CRM leader or architect in Manufacturing, higher education or public sector, wondering about how the new Agentforce product by salesforce will be of value to you, this article will provide you guidelines to evaluate it and make decisions and have some key considerations.

What is agent force and how it works?

If you look at the above picture, you can see the evolution of gen AI where we are getting to a point on using it to reason and infer human inquires. The key thing to understand is that agents are more about taking actions on your data using inbuilt automations built on the salesforce platform like flows, apex triggers etc. In a normal customer service chat , where it can only answer few questions and show content links, agent force takes one level more where you can quickly display order status, explain why your order is delayed with a human touch.

How does it Work?

Thanks to the architect keynote, the above picture says it all. You will have Agentforce preconfigured with topics which are mostly objectives with guardrails. Based on the inquiry by a human, Agent force infers and matches a topic and does actions which are exisiting automations like flows, apex and integrations. The key to remember is the Atlas resoning engine which infers based on the topic what action to execute. The last thing is the response to the user and how do we make it human as possible? Enter your LLM (chatgpt, claude AI). Agentforce sends the data with your CRM (RAG) to LLM gets a human response and displays to the user. Bottom line is Automated actions based on a human inquiry with a human response powered by LLMs.

Factors to consider to make this work

In order for you to make this succesful, here are the following key factors.

1. Data – This can be your CRM data with Salesforce or unstructured data which can be used from your vector databases or Salesforce Data cloud is a possiblity.
2. Actions- These are your pre existing automations with flows, apex and integrations.
3. Guardrails – Given Agent force will use LLM, it is important to set Guardrails on what data should not be sent to LLM, what content should not be displayed to the human user to prevent compliance issues.
4. Role – This is the one which makes it interesting and architects will love it. Each Agent will need to have a purpose , clear instructions on what it should do or not do. Make them unique and avoid duplicating agents.
5. Channel – Where do you want them to work and engage with your users which can be inside your org or outside on communities.

Common Usecases in Manufacturing, Higher Education and Government.

For Manufacturing users, here are the top 2 usecases where I see this would help.

1. Service your low tier customers with Agent force on your customer service portals where it can display order status, explain order delays.
2. Answer questions on your partner networks (dealers, suppliers) on invoices and there statuses with your partner portal network

For Higher Education users, here are the top 2 usecases where Agent force will help in increasing enrollment.

1. Answering questions on the status of applications with explanation of financial aid benefits using agent force will drive prospective students closer to decision and keep parents like me happy as well!!
2. Helping students answer questions on non credit courses with meaningful human responses will help in upselling non credit courses.

For Government users, here are the top 2 usecases where Agent force will help reduce citizen inquiries resolution time.

1. Leveraging Agent force on Claim based forms to answer questions on filling rigid claim forms will help citizens to submit claims quicker with right information.
2. Automating quick scan of documents uploaded with Agent force will help citizens upload the right documents to there claims which will help citizens get benefits quickly and get compensated.

As always feel free to let me know your questions or you can email me at buyan@eigenx.com for any further questions.