One of the major emerging challenges for 2024 will be getting all the digital assistants, also known as copilots, that invoke various artificial intelligence (AI) models to automate various tasks to interoperate with one another.

As AI continues to evolve, every provider of an application, platform or service is investing in AI capabilities that can be invoked via some type of digital assistant that has been trained to automate a specific task. The challenge is that most business processes span multiple applications, services and platforms, so inevitably there will be a need for various digital assistants to cooperate with one another to achieve a specific outcome.

In fact, Salesforce is already looking at this issue within the context of its technology alliances program, says Nick Johnston, senior vice president for strategic partnerships and business development at Salesforce. “Copilots will need to be interoperable,” he says. “There will be an ecosystem for assistants.”

It may be a while before various digital assistants are, for example, cooperating across application programming interfaces (APIs), but at this point it’s all but inevitable.

Less clear is how digital assistants that are assigned competing missions might be reconciled. A forthcoming fourth wave of AI will see the rise of digital assistants that will more proactively manage tasks in ways within a set of parameters defined for them. For example, a supplier of a product might instruct an AI agent to monitor an online market to determine when, based on demand and supply, is the optimal time to sell a specific product.

Of course, it’s just as likely a buyer of that product will program an AI agent to monitor that same market to determine when, based on supply and demand, is the best time to buy that product. Ultimately, it may require some human intervention to resolve the inherent conflict between agents assigned diametrically opposing missions, but at the very least, much of the time and effort that would otherwise be required to get to that point will have been eliminated.

In the meantime, organizations should expect that copilots and agents of varying types and capabilities will be spreading like wildfire across the enterprise in 2024. Many of them will essentially function as personal assistants that will perform rote tasks in the background, but that’s just the beginning. End users will inevitably start to combine prompts to enable multiple agents to perform various related tasks in sequence. That might require some expertise, but it won’t be long before organizations create libraries of reusable prompts to automate tasks using multiple co-pilots that will each be invoking a large language model (LLM) optimized for specific tasks.

The challenge and the opportunity now is not only determining what existing workflows to optimize, but also identifying new opportunities to automate tasks that previously would have required too much coding effort to achieve. The issue will be waiting to see to what degree providers of copilots will be willing to play nice enough with each other to enable organizations to achieve that goal.