Forget the one-dimensional chatbot. Agentic AI is a real contender to become the next big thing in customer contact. Agentic AI is AI that can perform (software) tasks and is seen as the third wave in AI. That begs the question: what will be the fourth wave? In the words of Mick Costigan of Salesforce Futures. 

 

The last time we met Mick Costigan, head of Salesforce Futures spoke, the tech fair CES in Barcelona had just ended: the time when gadget freaks can marvel at the latest technology. The topic was 'when machines become customers', following the rise of personal virtual agents such as Rabbit. In the meantime, Salesforce (not the only software house) has launched its own army of AI agents.

 

'Machines become customers' – have any serious steps been taken in this area or is it still a pipe dream?

“We talked about a world where individual consumers would have their own virtual agent. It is now one of the arguments for companies to get started with Agentforce, so that they can prepare for the massive arrival of agents on the side of those consumers. I am still convinced that we are moving towards a situation where machines act on behalf of customers.”

 

Salesforce is now talking about the third wave of AI, and by that it means AI agents. Has that brought the machine customer one step closer?

“Yes, various developments contribute to this, such as Apple Intelligence, Anthropic's latest innovations or new applications from the OpenAI stable. It is all moving in the direction of AI agents with a brain, a memory, which can also perform actions.

 

Instead of the machine customer, all attention is now focused on the machine employee or AI agent.

“We, as a leadership team, also started working on the Agent Builder and within a few hours we had hundreds of different use cases. Roughly a dozen Salesforce customers are running pilots. It’s not just about increasing efficiency by offloading human tasks to AI agents, but also about new tasks that weren’t possible before.”

 

Composing emails and conversation summaries and offering knowledge snippets or conversation suggestions are some of the most common use cases of generative AI right now. What’s the next, most promising use case after this?

“It will be about 'narrow use cases' – for example in the field of logistics. Think of agents who make API calls to a logistics service provider. Furthermore, a lot of development capacity is being put into large action models. Agentforce is only just available, so many applications still have to find their way. The American department store Saks went live last September with a pilot with Sophie. "

 

A common view in the Netherlands is: why would you automate conversations instead of improving processes so that customers no longer have to contact you?

“One does not exclude the other. Customers continue to need personal contact, also for questions that you would rather automate. Few people still contact you for a password reset. But if they want to contact you, how can they reach you?”

 

Looking at Agentforce, one could say that Salesforce’s CRM functionality is a foundation, but also a commodity. What are Salesforce’s ambitions with Agentforce?

“I can’t give you any numbers. But where we’re going is that people and AI agents are working together in the platform that is Salesforce. Those who already have Salesforce as a platform can get started with AI agents faster.”

Note: While the rates for Salesforce Service Cloud or Salesforce Marketing Cloud are based on the number of users per month, AgentForce pricing model based on a fee per conversation. An Agentforce Service conversation is when a customer sends at least one message or selects at least one menu option or choice within a 24 hour period.

We also put the question to Costigan: If Agentic AI is the third wave of AI – also referred to as autonomous agents – what will be the fourth wave?

“In any case, it concerns physical robots – in my opinion, these are AI agents with a ‘body’ and their development runs parallel to that of agentic AI. The difference: robots perform actions in the physical environment, agents perform actions within a software environment. Ultimately, after the autonomous agent, general artificial intelligence or AGI is the next step. As a futurist within Salesforce, I often visit R&D labs. There I often hear: AGI will be there in 2030. When I ask for an explanation, they point to the short time between GPT2 and GPT4. If they repeat that trick, you will end up with AGI, they say. It is a scenario to take into account. We will have to think more about what it means for people to work with AI. Even if that AI is much more powerful than what we have today. I assume that within 25 years we will have something close to AGI.”

 

If you look closely at the demo of Saks, then you can already talk about autonomous AI agents. Will there soon be more AI agents than regular employees?

“It’s not very complicated to sketch dystopian prospects around AI. You can also look at it from another angle. No company would employ hundreds of employees to do something extra for the customer. With AI agents, that becomes possible.” (Ziptone/Erik Bouwer)

Featured, Technology Salesforce