As more of the workforce becomes digital labor, who is going to manage it?  Jensen Huang – CEO of Nvidia – suggested that as agents were technology, then IT should manage them. Firstly what do we mean by manage?  If we draw the parallel between agents as digital labor and humans then it makes it easier to understand the JTBD when you have a team to manage:

BTW The 1st steps are reversed for humans

• Train (build agents, clean data, update knowledge)
• Onboard (deploy into production and integrate with human labor processes)
• Coach (update if improvements are identified or for regulatory non-conformance)
• Monitor (audit and governance of changes)
• Salary (pay agent platform which requires tracking)
• Promote (add new skills to broaden the power of the agent)
• Demote (remove skills from the agent if they are not performing)
• Off sick (take down if the agent goes rogue)
• Sack (supersede by newer agent)

Who is going to manage the agents?

Here is Jensen Huang’s video where he suggested that as agents were technology, then IT should manage them. I don’t believe Jensen’s view makes sense for 3 reasons:

• Many IT systems in organizations are successfully managed by business units – CRM, marketing – rather than IT. 
• Think of agents as digital labor, then surely the business unit they work for should manage them.  
• Building agents can be done by the business using process diagrams and drag/drop tooling for backend workflows and integrations, so it is not IT building them.

IT is a supplier or advisor to the business units. IT doesn’t need to be the owner of the agents. They can help with procurement with AI platform and ensure there is consistency across the business units. This could be through an AI COE – more on this later. IT also needs to get involved where there are integrations, architecture and security considerations. 

HR (Human Resources) is not going to become AR – Agent Resources.  HR doesn’t manage people. It is the team leaders, supervisors and managers in the business. HR administers the HR programs on behalf of the business units and ensures that the organization is in compliance. HR works for the organization not the workforce. 

Someone needs to own agents

It is such early days we haven’t got many good examples of how to build, manage, and monitor agents. But it is better to put a sub-standard approach in place and modify it as you learn more, than to have no approach. Our recommendation is that the business is the owner.  

Why? Because the operational and regulatory issues are far greater than the technical issues. Embedding agents into the workforce alongside existing teams will not be straightforward. If you remember back when outsourcing was big, there was a huge pushback from teams who were reluctant to train the outsourced team just so that they could take their job. Why would it be any different for agents? 

Also, agents may need to hand back to humans when they cannot complete a task or there is an issue with the agent performance. That integration needs to be planned so that the experience for the customer is seamless. The agent needs to summarize the history of the conversation so far and hand it to the human as context. This is easier said than done. Is it synchronous with a seamless hand-off from agent to human?  Or async where the human follows up later? From a user’s perspective, the discussion needs to start from where you left off. It cannot start at the beginning again. 

Lean in and learn

At the moment, learning and experimenting is the most important part. For highly regulated organizations you need to have a governance framework that your teams, internal auditors and external regulators are happy with. Again it may not be ideal, but it is a start. And with agents’ capabilities expanding so quickly you have a baseline to be able to innovate from. Taking a “wait and see” is not an option with technology moving so quickly.

I’ll finish with an excerpt of the statement about AI and Agentforce from our COO to the entire company

We began our internal Agentforce journey just a few months ago with two key objectives. First, we aimed to understand Agentforce thoroughly, allowing us to speak about it with authority and explore how Elements could be relevant within the evolving Salesforce landscape. Second, we wanted to leverage our learnings to promote and market Elements. While some of our initial topics have been exploratory ‘demoware,’ others have already delivered tangible value. 

The team has surpassed my expectations, and we now have a structured approach that is gaining traction. Our integration of agent-building into our normal workflow using Elements is proving effective. We are utilizing Requirements and Stories, developing in Sandboxes, and deploying changes via Gearset. This structured approach is now being documented into Solution Guides, which will feed into training and certification programs. 

A key takeaway is that Agentforce is now considered Business As Usual (BAU). While we will continue to push the boundaries and explore new use cases, the priority is now on delivering meaningful impact. Thank you for your contributions and enthusiasm—let’s keep the momentum going!

Learn more

We’ve developed some content to help you accelerate your journey based on our experiences

EBOOK > Ultimate Guide to Creating Agents

https://elements.cloud/whitepapers-ebooks/ultimate-guide-to-creating-agents/

SOLUTION GUIDE

https://support.elements.cloud/en/articles/10470159-building-agentforce-agents-with-elements-clou

VIDEO – demo of our complex Sales Coaching agent