AI is more than just a productivity hack
getty
For technical and software engineering teams, agent AI is the best productivity tool ever, increasing productivity by light years. However, for business end users, agent AI is still in its infancy and its outcomes are uncertain.
Those are the words of Box CEO Aaron Levie, who recently joined CXOTalk’s Michael Krigsman for a discussion about the potential and dangers of AI and agent AI. “We’re still in the early stages of what agent work looks like in the enterprise and what deployment looks like,” Levie said. “We have an interesting dynamic, kind of a tale of two cities.” Agent coding is proving to be a game-changer, and then there’s business-focused agent AI, which still has uncertain benefits.
The key to leveraging AI in business, Levy argues, is to see it as more than just a productivity hack, but as a “technology for abundance.” He explained this with a thought exercise. Look at the difficult parts of your business and ask: What would you do if you had unlimited ability to “review information, make decisions, and access data?”
Until now, such capabilities have tended to be limited by the number and skills of human workers. Instead of manipulating spreadsheets or ERP system data, decision makers can leverage computing power to address problems.
For example, if B2B companies could deploy agents to comb through their customer base, they would have better insight into the right time to communicate the right message to customers. With unlimited computing, you can gain greater insight into your customers and how they respond to your marketing campaigns. Alternatively, agents can comb through LinkedIn data to identify potential talent. Agents can comb through marketing and sales results to identify where budget and resources are being wasted.
At this point, Levie emphasized, “the agent is probably the most technical solution ever deployed to a non-technical person.” “You’re putting non-deterministic intelligence into the hands of every knowledge worker.” The danger here is that agents can “run amok and get the wrong data and generate the wrong reports.”
While software developers may love AI and business leaders may ponder replacing more expensive and less available human labor with AI, reality soon hits. Those who work closely with agents quickly realize that human supervision is essential. Levy warns that there’s “always a chance that the agent will do something wrong, that the flavor of what they serve will change, that there will be bugs.”
What is needed now is a mechanism to ensure that agents are working in the right context, at the right time, and with the right guardrails, especially when less technical business users are involved. Verification is key.
The processes adopted to validate software development are not applicable to knowledge work. “We’ve seen what the promise of an agent is in coding, and we’ve seen what the promise of a kind of chatbot is in knowledge work. The question now is: What is the full promise of an agent across knowledge work? I think this is definitely going to be a defining topic within the enterprise for years to come.”
