Make way, copilots! It's time to make way for the AI agents.
That has been the message from the software industry in recent days, as among the largest firms have thrown their weight behind the newest idea of how generative artificial intelligence could change into an integral a part of working life.
Microsoft, Salesforce and Workday put agents at the middle of their AI plans this week, while Oracle and ServiceNow also used the industry's annual user conference this month to advertise the concept.
AI assistants, often known as copilots – a term first popularized by Microsoft – have change into the software industry’s primary response to the generative AI unleashed with the launch of ChatGPT nearly two years ago.
The latest wave of AI agents goals to go even further and act on behalf of users. While agents are the newest front within the battle between tech giants like OpenAI and Google, also they are the software industry's latest try and sell generative AI to business customers.
This development reflects each an advancement within the underlying technology and a brand new marketing pitch from an industry trying to capitalize on a highly hyped technology that has not yet had a significant impact on its revenues.
If the industry's claims prove true, the evolution of AI assistants into AI agents could also open the door to a much more disruptive phase in the event of generative AI – each for the employees affected by the technology and for the software firms themselves.
Behind the proliferation of agents – also known as “agentic” systems – lie plenty of advances within the underlying technology because the first generative AI chatbots.
With more memory, systems can higher understand context and their planning capabilities have improved. Agents also often hook up with other systems through application programming interfaces (APIs), meaning they will perform actions on a user's behalf relatively than simply returning information.
The latest generation of agents is meant to enhance existing copilots, not completely replace them. According to Microsoft CEO Satya Nadella, his company's copilot software is evolving into an “enterprise orchestration layer,” a conversational interface that permits employees to create and use agents to perform specific tasks.
Initially, AI agents are primarily advertised as tools that perform easy routine tasks corresponding to filling out an expense report.
But some firms are already boasting about their ability to tackle more complex tasks and even take over some tasks entirely. One focus is on automating customer support systems, which could potentially replace large numbers of call center agents.
So far, generative AI has had little impact on increasing revenue growth for software firms.
The entire software industry remains to be in “show me” mode in the case of copilots or AI agents, said Jim Tierney, a growth equity investor at AllianceBernstein. “It's still an open query how exactly that is going to be monetized,” he added.
Salesforce CEO Marc Benioff said there was not enough traction with copilots, telling the FT: “Microsoft has misled customers with its AI strategy. They don't must do it themselves. We construct it into our platform. Customers shouldn’t be forced to coach their models over and all over again.”
Software firms expect customers to achieve direct productivity advantages from agents who can take over entire tasks.
According to Microsoft's Nadella, as such AI systems change into more powerful, “the models themselves have gotten an increasing number of of a commodity, and all the worth is created by the way you anchor, control and optimize those models with your online business data and workflows.”
As agents tackle more of the work, firms like Apple, with its dominant smartphone platform, and Microsoft, with its desktop productivity apps, could have one of the best likelihood of success, Tierney says.
For now, the total impact of this shift is prone to be muted, because the tendency of generative AI systems to “hallucinate” makes users reluctant to permit them to perform unsupervised actions.
“I'm skeptical and even slightly nervous” concerning the widespread use of agents, says Barry Briggs, a former chief technology officer at Microsoft and now an analyst at Directions on Microsoft, an independent research firm.
The probabilistic nature of the technology signifies that customers cannot use the technology for essential tasks but must construct it into work processes where staff have the ultimate say, he added.
However, some firms are already claiming to be considering the technology through to its logical conclusion. Last month, Sebastian Siemiatkowski, CEO of Swedish fintech company Klarna, said his company was within the means of halving its workforce using AI.
Siemiatkowski also caused a stir within the software industry when he said Klarna would ditch Salesforce and Workday entirely and as an alternative use AI to construct the software it must run its business. That statement is widely seen as an outlier given the present state of technology within the tech world, though it does point to a future that some consider will probably be much more disruptive.
However, most software investors are betting that the massive winners will probably be the present powers of the software industry – even when it remains to be unclear how and when the advantages of the technology will unfold.
As an extension of the copilots, agents are only the newest step in the trouble by today's leading software firms to stake out their ground and prepare for the time when generative artificial intelligence is advanced enough to deliver real productivity gains, says Kevin Walkush, portfolio manager at Jensen Investment Management.
The current generation of agents is unlikely to contribute much to the revenue of software firms, he added, but: “It's now only a matter of creating their beachheads and positioning themselves for the long run.”
Even if incumbents have one of the best likelihood of winning, the move to agent-based systems could still disrupt the way in which they do business. Most charge customers a licensing fee based on the variety of employees using their software – a model that’s in danger if AI agents seriously reduce headcount.
In response, most software firms have begun testing usage-based pricing, which ties their revenue to question volume. Salesforce, for instance, plans to charge $2 for every “conversation” with its AI agents. Many are also talking about moving to outcome-based pricing, which might allow them to share in among the profits customers make from using the software. However, it’s unclear how this might work.
“It’s still early and we don’t know the way the pricing models will impact,” said Byron Deeter, partner at Bessemer Venture Partners and investor in young software firms.
Similar to the move to the cloud, where a change in software firms' revenue recognition methods caused turmoil within the industry, the transition to a brand new pricing model for AI “could possibly be bumpy for publicly traded (software) firms,” he added.