Nothing seems easy about artificial intelligence. It’s expensive to build and implement, be it with skills to build it or commercial software offerings. It’s biased and spits out hallucinatory results. It’s going to take away jobs. It’s going to require new jobs. It’s a big, big deal.
Then, there’s another way to introduce AI in a more digestible way — as bite-size nuggets, or agents that zero in in specific tasks. That’s one, less-invasive way to ease AI’s path into the workplace.
AI bots, if you will, serve as digital “coworkers” that won’t necessarily displace people out of their jobs, but serve as assistants that help get things done. There’s even a relatively new term to describe the phenomenon: “agentic” workflows.
We’ve already seen AI agents emerge everywhere: Siri, Alexa, and Google Assistant as intelligent, and now, AI-powered agents are taking on roles as diverse as automating appointment scheduling, IT service desk inquiries, automated negotiation systems, corporate chatbots, and autonomous vehicles.
The levels of AI agent expertise range from “reflex agents,” such as stock trader bots or office productivity boosters, to “learning agents” that can oversee and adapt to customer purchasing patterns, to “hierarchical agents” that operate in robotics or autonomous domains to oversee more complicated processes, as outlined by Thomas Latterner as well as within an AWS tutorial.
At this point, only about 10% of organizations have functioning AI agents within, but 82% intend to implement them within the next one to three years, a recent survey of 1,100 organizations published by Capgemini finds.
AI agents mirror the promise of AI overall, the Capgemini survey also shows. Almost three-quarters (71%) anticipate that AI agents will facilitate automation. A majority also believe that they will relieve human operators of repetitive tasks, allowing them to focus on value-added functions such as customer experience.
“The emergence of AI agents marks a shift, with potential to enhance automation and productivity across sectors, business processes, and along the entire value chain,” the report’s team of authors, led by Pascal Brier, group chief innovation officer of Capgemini, state. “AI agents have evolved from supportive tools to autonomous entities capable of executing tasks independently.”
Importantly, unlike Big AI and all its baggage, we tend to actually trust AI agents, Brier and his team found. “There is a level of trust in AI agents for specific tasks, such as generating work emails, coding, and data analysis,” they write.
Seven in ten respondents (70%) would trust an AI agent to analyze and synthesize data, and 50%would trust an AI agent to send a professional email on their behalf.
At the same time, AI agents, just like Big AI, need “guardrails to validate AI-made decisions, ensuring transparency and accountability,” they add.
About three-quarters of respondents (75%) intend to deploy AI agents to tackle tasks such as generating and iteratively improving code. Other tasks include generating draft reports, incorporate feedback, and iterating (70%) and researching, writing, and optimizing website content (68%).
In examples of actual implementations, Brier and his colleagues cite Swedish payments company Klarna, which “uses an AI assistant to handle tasks equivalent to the workload of nearly 700 employees. This AI assistant addresses service requests, manages refunds, and handles returns in various languages.”
