While shares of Nvidia stock have soared amid demand for chips that train artificial intelligence models, software companies are still waiting for their AI moment to kick in. Generative AI "copilots" have failed to generate meaningful revenue for software makers, but there's hope that the next generation of autonomous, goal-driven AI "agents" will make it happen.
Since the emergence of generative AI in late 2022, Microsoft, Salesforce and other software makers brought to market conversational chatbot interfaces often called copilots. They've offered them as add-ons to existing software products. However, while many customers tested copilot technology, they've been unwilling to pay more for using it.
Now Microsoft, Salesforce, Workday, HubSpot and others are betting on AI agents as moneymakers. Salesforce introduced Agentforce at its Dreamforce customer conference in September.
The iShares Expanded Tech-Software Sector ETF has climbed only 14% in 2024 while Microsoft stock has advanced 11%. Meanwhile, the S&P 500 is up 23%. Nvidia stock has soared 172%.
The revenue lift from Microsoft 365 Copilot has disappointed Wall Street. Microsoft recently rolled out new healthcare AI agents.
Cybersecurity companies are among those developing AI agents, which perform tasks on behalf of users.
AI Agents: Acting, Not Assisting
"I think the difference is assisting versus acting," Nikesh Arora, chief executive of Palo Alto Networks, told IBD. "Copilots assist; agents act."
In October, Nvidia partnered with IT services giant Accenture for corporate services built around "agentic AI." Alphabet's Google demonstrated the potential uses of AI agents at its recent "Gemini at Work" event. And Facebook-parent Meta Platforms has many AI agent initiatives underway.
TD Cowen analyst Derrick Wood said AI agents could finally enable software companies to create revenue streams from generative AI.
"I think there's a lot more value software companies can unlock with agents versus co-pilots," he told IBD. "Agents can leverage LLMs (large language models) to tap into private data, reason on how to solve a problem, and make decisions or take action in an autonomous way. Software companies should be able to better monetize agent networks better than co-pilots."
Generative AI leader OpenAI, which was recently valued at $157 billion in a new funding round, aims to spawn the age of AI agents with a new training model capable of enhanced reasoning skills.
Code-named "Strawberry," it's officially called OpenAI o1. The new model is said to lower the risk of "hallucinations," the Achilles' heel of chatbot technology.
Research firm Gartner is upbeat on "intelligent agents in AI."
Gartner Projects Big Growth
"Today's AI models perform tasks such as generating text, but these are 'prompted' — the AI isn't acting by itself. That is about to change with agentic AI," said a Gartner report.
"By 2028, 33% of enterprise software applications will include agentic AI, up from less than 1% in 2024, enabling 15% of day-to-day work decisions to be made autonomously."
However, some Wall Street analysts are skeptical.
"Recently, we have seen many software companies rebrand AI functionality, or introduce new features messaged as 'Agents,'" said RBC Capital analyst Rishi Jaluria in a report. "While we are believers in agentic AI (autonomous AI agents that don't require human intervention), we understand why some may question the reality of these new features."
Jaluria added: "In addition, we caution some companies may not be able to meet the expectations they are currently setting with customers. Having said that, we believe GenAI is heading in an agentic direction."
Copilots aim to improve worker productivity but still rely on human prompts, notes Wood. Copilots help users answer questions. But agents execute tasks on behalf of humans by solving problems and taking action.
Salesforce Stock: Shift To AI Agents
Salesforce's new AI agent platform is expected to be available on Oct. 25. The company has been touting artificial intelligence since it introduced predictive Einstein tools for salespeople in 2016. But Salesforce has never disclosed AI revenue.
Based on new generative AI, which responds to prompts, Salesforce last year introduced an Einstein copilot feature — a conversational AI assistant integrated within the user interfaces of Salesforce apps.
At its Dreamforce conference, the software maker demonstrated AI agents performing customer service tasks, sales lead prospecting and order fulfillment.
The software maker's new Agentforce business model will be "consumption" based. That means customers will be charged based on usage. List pricing is $2 per interaction.
Meanwhile, Salesforce stock has gained 10% in 2024.
As with copilots, companies will need to clean up their data to use them.
To build gen AI software applications, companies need to harness proprietary data to train and deploy AI models.
The rub is that gen AI chatbots are known to "hallucinate," or make up incorrect answers when they can't find an accurate response.
OpenAI's New Model Key
Before deploying AI agents, companies will need to be confident they're not hallucinating, either.
"There's going to be a very careful transition from assisting to acting," said Palo Alto's Arora. He says companies will need to decide when "to take their hands off the steering wheel" and let AI agents do their thing without human interaction.
Meanwhile, OpenAI's new large language model is expected to boost the decision-making capability of AI agents. The program, OpenAI o1, breaks apart complex problems into more manageable pieces before answering and, through reinforcement, it learns to recognize and correct mistakes to try a different approach when the current approach isn't working.
OpenAI has raised $6.6 billion in new funding, valuing the startup at $157 billion, up from $86 billion early this year. New investors include Nvidia and Japan-based SoftBank.
Further, Salesforce hopes that AI agents will create demand for its Data Cloud products. The platform helps organizations manage and analyze their data effectively.
"Agents are the next wave of AI interface in software and an interface that we believe has a better likelihood of success and adoption than co-pilots," said KeyBanc Capital analyst Jackson Ader in a report. "What we want to know is whether customers that express interest even have their data estate in a place for the agents to deliver value."
Nvidia Stock Is Still An AI Play To Watch
A wave of AI agent startups, meanwhile, is getting funding from venture capital firms, including Sierra, NinjaTech AI, Hippocratic AI and Agency. Nvidia's venture arm is among investors in Hippocratic AI. Nvidia is among the top AI stocks to watch. Another is Sierra, co-founded by Bret Taylor, chairman of generative AI leader OpenAI.
Taylor served as co-CEO of Salesforce, founded Quip and was chief technology officer of Facebook.
A Sierra blog noted: "Chatbots continue to fall short of customer and company expectations, struggling to solve problems, communicate with clarity, or self-improve over time. Sierra's AI agents offer a superior alternative by handling complex tasks, communicating using natural language and continuously improving."
Meanwhile, Sierra plans a fundraising round valuing it at $4 billion, according to a report.
Follow Reinhardt Krause on Twitter @reinhardtk_tech for updates on artificial intelligence, cybersecurity and cloud computing.
