A survey of 2,552 U.S. consumers published this week suggests that expectations of the tasks and responsibilities assigned to artificial intelligence (AI) agents are rapidly evolving as more individuals are exposed to interact with them.
Conducted by Salesforce, the survey finds roughly half of respondents (53%) want an AI agent to break down complex information in a way that makes sense to them. In contrast, 22% of respondents want an agent to simplify decision-making and handle tasks, while 16% want an agent to maximize efficiency, multitask and act as a personal project manager. Another 15% simply want an AI agent to provide personalized, curated recommendations for shows, restaurants and products that align with their unique tastes and preferences.
Overall, the survey makes it clear that AI agents are already starting to have a societal impact, with younger survey respondents exhibiting more willingness to rely on AI agents to execute tasks on their behalf, says Vala Afshar, chief digital evangelist at Salesforce.
It’s still too early to say how widely AI agents will be trusted to perform those tasks and, at any different time, the level of trust in an AI agent is going to depend on the use case and previous experience. It’s important to remember AI agents are based on large language models (LLM) that are able to leverage reasoning engines to make a probabilistic guess based on the data exposed to them. The challenge is many tasks are deterministic in the sense that they need to be performed the same way every time, which is something an AI agent is not going to do given its probabilistic tendency.
However, there is no shortage of tasks that don’t have to be done precisely the same way every time, including, for example, creating snippets of code or generating an article. For those types of tasks, AI models, and soon agents, are proving themselves indispensable.
In fact, AI agents will inject a level of joy into performing and completing tasks that right now is all too absent, notes Afshar. Manual processes today often contribute to making work more tedious than most anyone really enjoys, when that time could be put toward far more emotionally and financially rewarding activities, he adds.
Regardless of how anyone might feel about AI agents, they will soon be pervasively employed. Some jobs might even be eliminated by them, but at the same time new ones that would never have otherwise been needed will be created. The challenge and the opportunity will be determining when and how best to orchestrate a set of AI agents to perform and manage workflows in collaboration with humans in a way that generates a trusted result, says Afshar.
Individuals that embrace AI agents will ultimately fare much better than those that resist a massive societal change that is about to occur with or without them, he adds. “In the long term, the future is defined by the optimists,” Afshar says.
The only thing left to determine now is to what degree the proverbial AI agent glass at any given moment is half full or half empty.
