Microsoft and Salesforce recently announced AI agents. Now Google will join a crowded but lucrative space. Code-named Project Jarvis, the technology will conduct consumer functions to minimize mundane tasks.
What makes these AI agents from Google and Microsoft more interesting is their history with automating advertising in platforms that require lots of information and precise actions to find and target messages using specific data sets and modeling.
Expertise in search and discovery is required to complete the tasks. And it appears that these companies are using their experience with advertising to expand across other areas and business workflows.
Project Jarvis, an AI agent for Chrome, has the potential to support web-based task automation, but introducing computer agents that take over browser tasks will surely introduce scrutiny among privacy advocates. If not protected, data could leak.
advertisement
advertisement
These agents would follow through with tasks by interpreting screenshots and performing actions such as clicking and typing in response to contextual prompts. Testing is expected to begin in December.
Some suggest that Project Jarvis will make some APIs obsolete, while others believe it could reduce the need for some in certain situations.
Michael Birkeland, who lives in Norway and describes himself on Medium as a tech enthusiast, amateur coder, and AI aficionado, earlier this year wrote about AI agents and the end of APIs.
For example, Birkeland told ChatGPT to make a weeklong meal plan for dinner for healthy and nutritious choices for someone on a specific budget and to ensure the products can be found in Oda, a Norwegian online grocery store, and have it delivered.
With a few additional requests, the agent found seven recipes that met the criteria. It set up a shopping list, opened the web browser, navigated to the Oda online grocery store, navigated to the products on the shopping list, and added them to the cart. It found a delivery day and paid for it using his card or account.
All of these interactions and more described in his post could be performed without using an API. It would simply navigate, look at the page and click, similar to the way a human would do it.
These AI agents, software applications that autonomously perform complex tasks on behalf of users, could theoretically save consumers time and companies millions of dollars by automating repetitive tasks, according to Microsoft.
In fact, Microsoft says agents can save companies as much as $50 million annually or offer the equivalent value of adding 187 full-time employees by automating business processes.
Microsoft's ten new autonomous agents function inside Microsoft's Dynamics 365 back-office suite. Similar to Salesforce's agents, Microsoft's AI agents can qualify sales prospects, automate time and expense tracking, among other things.
Google could roll out Project Jarvis with the release of the next Gemini Large Language Model, according to The Information.
OpenAI also reportedly is working on something similar called computer-using agent (CUA). It makes sense that it will likely have the ability to conduct research on any chosen subject.