Klarna, a financial services company, announced earlier this year that its artificial intelligence agent, powered by OpenAI, had taken over two-thirds of customer chats and was performing tasks equivalent to 700 full-time agents. This was after only one month of use.
Alexander Kvamme, CEO of customer engagement startup Echo AI, told CNBC that Klarna’s announcement in February could have been the first indication of AI agents “having their ChatGPT moment.”
OpenAI made the ChatGPT chatbot available to the public in late 2022, giving them a taste of how new generative AI chatbots could provide much more thorough, creative, and conversational responses to web queries than traditional search, which was how consumers sought online information for the previous 25 years. Google, Microsoft
Others followed with competing products.
The industry quickly moved beyond text responses to AI-generated images and videos. Now comes the rise of artificial intelligence agents.
Agents are designed to be productive and complete tasks, rather than simply providing answers, as chatbots and image generators do. They’re AI tools that can make decisions, for better or worse, “without a human in the loop,” Kvamme explained.
Grace Isford, a partner at venture firm Lux Capital, said there has been a “dramatic increase” in interest among tech investors in startups focused on developing AI agents. They’ve raised hundreds of millions of dollars and seen their valuations rise alongside the larger generative AI market.
According to PitchBook, generative AI exploded in 2023, with $29.1 billion invested in nearly 700 deals, representing a more than 260% increase in deal value over the previous year. Meanwhile, the non-AI investing landscape has been in a prolonged lull for more than two years, following record financings during the Covid pandemic.
If 2023 was the year of peak AI hype, then 2024 is the year of early deployments.
“It has really been a torrent of innovation that has hit the market since the introduction of ChatGPT,” Jared Spataro, Microsoft’s corporate vice president of AI at Work, told CNBC. Microsoft is OpenAI’s biggest backer, having invested billions of dollars in its own generative AI models and products, in addition to billions in the ChatGPT developer.
The term AI agents is not well defined in the technology sector. Industry experts who spoke with CNBC about the emerging trend generally saw agents as a step up from chatbots in that they are typically designed for specific business functions and can be customized on large AI models. Consider J.A.R.V.I.S., Tony Stark’s multifaceted AI assistant from the Marvel universe.
AI agents are frequently described as advanced generative AI tools capable of performing multistep, complex tasks on behalf of users and creating their own to-do lists, eliminating the need for users to walk them through the process step by step.
“An assistant is not just giving you the answer, but automating a series of steps,” explained Francois Ajenstat, chief product officer at Amplitude, a digital analytics company.
How are Microsoft and Google playing?
Microsoft CEO Satya Nadella stated on an earnings call earlier this year that he wants to offer an AI agent that can complete increasingly more tasks on a user’s behalf, but that there is “a lot of execution ahead.” Meta and Google executives have highlighted their efforts to improve the productivity of AI assistants.
Google announced Project Astra, the company’s latest advancement toward its AI assistant, at Google I/O in May. Google’s DeepMind AI unit is developing it.
In Google’s demo video, the assistant used video and audio to help the user remember where they had left their glasses, review code, and answer questions about an object shown. It is currently only a prototype, but Alphabet CEO Sundar Pichai hopes to make it available to users later this year.
The demo came a day after OpenAI demonstrated a similar audio back-and-forth conversation with ChatGPT, positioning it as an AI assistant capable of serving as a conversationalist, language translator, math tutor, and code co-writer.
At its Build developer conference, Microsoft announced a partnership with Cognition AI, which will bring Cognition’s own AI agent, Devin, to customers. Devin, according to Cognition, is the “first AI software engineer.”
Devin quickly gained traction on social media for its ability to handle multistep processes. Devin develops a problem-solving process before writing, testing, and shipping the code.
According to Martin Kon, the operating chief of enterprise AI startup Cohere, AI agents could perform tasks such as booking and expensing plane tickets, providing loan interest rates, and updating Salesforce with customer arrival times.
Until now, the tools have primarily been used to assist in the writing of code. In an early 2023 blog post, CEO Thomas Dohmke stated that approximately 46% of all code “across all programming languages” on Microsoft’s GitHub was generated by AI.
While the distinction between an AI coding tool and a true AI agent is hazy, most experts who spoke with CNBC said the defining feature of an agent is that it goes far beyond a single use case and begins to resemble an all-capable personal assistant.
Anthropic and other startups are already working towards this goal. The first step is to enable their chatbots to interact with external tools and services on behalf of customers.
Microsoft’s Spataro stated that developing his company’s Copilot coding agent has been “kind of like being strapped to a rocketship.” He stated that Microsoft is shifting from one- or two-step tasks to multistep tasks. This could include looking at a user’s calendar and providing a 30-second overview of what to prioritize for the day.
In a recent note to investors, Macquarie’s head of U.S. AI and software research, Fred Havemeyer, stated that the firm anticipates seeing more AI agents.
“We think agentic AI, which can self-direct towards achieving tasks, will be the tools that unlock the value of GenAI for everyday users,” Havemeyer stated.
Romain Huet, OpenAI’s head of developer experience, told CNBC that the concept of AI agents first emerged last year, but people quickly realized that more work needed to be done to make the tools more autonomous.
“We have the models that become more and more powerful, so we can now capture user intent much better than before, but we’re also still pretty early on that journey at building agents,” Huet stated.
The significant advancement, he stated, will be when an AI agent can learn your preferences and “take action on your behalf” without your permission.
Startups raise big money.
AI agent startups are raking in huge sums of money from investors. They aren’t the billion-dollar-plus financings that have been flowing into AI model companies, but valuations remain far ahead of business fundamentals.
Last year, Adept, founded by OpenAI and Google alumni, was valued at more than $1 billion. The company’s website states that its technology “navigates the complexity of software tools so you don’t have to.”
In May, H, a French AI agent startup, raised a $220 million seed round from investors such as Amazon, Samsung, UiPath, and former Google CEO Eric Schmidt. Artisan AI, a Y Combinator-backed startup developing AI agents marketed as “AI employees for enterprise,” recently raised $7.3 million in seed funding and claims to have onboarded over 100 companies.
Artisan AI’s founder and CEO, Jaspar Carmichael-Jack stated that it would not be possible to begin working on true AI agents until 2022, when chatbots such as ChatGPT first enabled the average consumer to interact with these tools.
“People talk about how the VC market is down in general,” Carmichael-Jack explained. “But for us it’s like 2021 in AI startups.”
Braden Hancock worked for Facebook Research and Stanford’s Artificial Intelligence Lab before co-founding Snorkel AI in 2019. He claimed that the market is in a “similar hype cycle” to that of self-driving cars. And broader AI agents will take a long time to reach the mainstream, he predicted.
Hancock stated that agents must be “many times” better before people are “willing to accept putting something on autopilot.” He went on to say that having technology sign your name and make money transfers on your behalf “sets a really high bar.”
Imbue, Kanjun Qiu’s three-year-old startup, has been valued at more than $1 billion and backed by Amazon’s Alexa Fund and Eric Schmidt. Based on the company’s own user research, Qiu stated that the current characterization of AI agents — as generally intelligent personal assistants that handle delegated tasks — is not what users actually want because, by design, they are “not fully trustworthy.”
“Even as CEO, it’s hard for me to delegate things to my executive assistant,” Qiu stated. “I’ve had her for two years, and she’s amazing.” For new things, Qiu said, “It’s still hard for me to fully know, ‘Okay, is this going to come back the way I expected?'”
Imbue is working on ways for people to create their own AI software agents — without coding — to run in the background for their specific needs, such as tracking the news or building a bot to book travel. Because each use case is unique, these AI models do not require user data to train.
Instead of delegating tasks to an agent created by OpenAI or Google, which would be centralized and controlled by those companies, Imbue envisions agents putting control in the hands of their users.
“There’s a way of thinking about agents as enabling every person to make software,” Qiu stated. The individual is “asking the agent to write code on the computer, to make the computer do what I want to do.”
