State Of AI Agents At Big Tech Companies

How are the key players differentiating and what bets are they making?

Happy Monday and welcome to Investing in AI. I’m co-hosting an AI Summit in New York on March 19th. If you are (or know) a startup that should come present in our startup showcase, they can apply here. The event is not open to the public, but is invitation only.

Today I want to talk about the big 4 tech companies working on agents, and where they stand. How do Google, Amazon, Salesforce, and Microsoft seem to be playing in this space? If you want the TL;DR version, here it is:

• Google is betting on democratization. They want AI agents to be as easy to deploy as a Google Form.

• Microsoft sees AI agents as orchestrators—digital conductors linking fragmented apps. By embedding AI into Azure, Teams, and Dynamics 365, they’re turning agents into glue for enterprise workflows.

• Salesforce is all about digital labor. Their “Agent Force” initiative aims to flood businesses with a billion hyper-specialized bots, automating everything from customer service to HR.

• Amazon is building a society of agents. AWS Bedrock lets companies create teams of specialized bots—a supervisor agent delegates tasks to HR, finance, or compliance specialists.

When you think through the assets and core business strategies of each of these companies, their approach to the agentic ecosystem makes sense. While these big four all offer a lot of similar functionality, where they focus and prioritize is different.

Google wants data, and so they want people to build and use as many agents as they can. Google is focused on democratization of agent building. They seem more focused on real-time data and open APIs. Their bet seems to be that people will want to use agents first and foremost for operational efficiency.

Microsoft has its co-pilot product, which has had mixed reviews from the public. It makes sense, as the office task bellwether, that Microsoft would see agents as work coordinators, and offer more fully finished products and less prioritization of build-your-own agent capabilities. I’m sure Microsoft would love to see models more commoditized and would prefer that the end relationship with the user and access to work documents for agents as the main drivers of value.

Salesforce, which is probably the weakest overall in AI of the big 4, has the most interesting positioning of agents as digital labor. Salesforce is very focused on quick win applications like agents for HR, and sees agents as the next version of software - very task focused. So agents seem to tie to individual tasks.

Amazon, through AWS, has built an interesting agent coordination infrastructure where they see a future of multi-agent coordination. This fits their view of themselves as an infrastructure provider first and foremost.

The primary strategic points of differentiation of these companies lies in these areas:

• centralized vs modular architectures

• quick accessible deployment vs customization and personalization

• ready to go agents for quick wins today vs long term infrastructure transformation

I’m sure all four will play a major part in our agentic future, but we are a long way from understanding the winning approach.

Thanks for reading.