Salesforce announced Monday the latest version of its AI agent platform as the company looks to lure enterprises to its AI software in an increasingly crowded market.

The customer relations manager giant unveiled the new platform, branded Agentforce 360, ahead of its annual Dreamforce customer conference that kicks off October 14. This newer version of Agentforce includes new ways to instruct AI agents through text, a new platform to build and deploy agents, and new infrastructure for messaging app Slack, among others.

A notable aspect of Agentforce 360 is its new AI agent prompting tool, called Agent Script, which will be released in beta in November. Agent Script gives users the ability to program their AI agents to be more flexible and better respond to “if/then” situations. This allows AI agents to be programmed to be more predictable in less rigid situations like customer questions.

Users can tap into “reasoning” models, which claim to think before responding as opposed to responding based on patterns. Anthropic, OpenAI and Google Gemini power these “reasoning” agents.

Salesforce also announced it is releasing a new agent building tool, Agentforce Builder, which allows users to build, test and deploy AI agents from a singular spot. This tool, which will be released in beta in November, includes Agentforce Vibes, an enterprise-grade app vibe coding tool that Salesforce announced earlier this month.

The company also announced a broader integration between Agentforce and Slack. Salesforce said its core apps, including Agenforce Sales, IT and HR, among others, will surface directly in Slack starting this month and expand through the beginning of 2026.

Slack is piloting a new version of its Slackbot chatbot that is meant to be more of a personalized AI agent that learns about its user and will offer insights and suggestions.

Salesforce wants Slack to serve as an enterprise search tool in the future too and plans to launch connectors with platforms like Gmail, Outlook, and Dropbox in early 2026.

This latest update from Salesforce comes at an interesting time for the enterprise AI market. Companies continue to release AI features aimed at their enterprise customers while enterprises struggle to see a return on investment for these tools.

Last week Google announced Gemini Enterprise, a suite of tools — many of which were already available — for building enterprise-grade AI agents, that counts Figma, Klarna and Virgin Voyages as early customers, among others.

Anthropic also started to show traction for its enterprise product, Claude Enterprise. The company announced it struck a deal with consulting giant Deloitte to bring its Claude chatbot to Deloitte’s 500,000 global employees — its largest enterprise deal yet. Anthropic announced a strategic partnership with IBM the next day.

Salesforce touts that Agentforce has 12,000 customers — significantly higher than any of its competitors, according to its Agentforce press release. Early pilot customers of its Agentforce 360 upgrades include Lennar, Adecco, and Pearson.

This is all despite a recent MIT study found that 95% of enterprise AI pilots fail before they reach production as companies still struggle to justify spending money on these AI tools.

For years, Big Tech CEOs have touted visions of AI agents that can autonomously use software applications to complete tasks for people. But take today’s consumer AI agents out for a spin, whether it’s OpenAI’s ChatGPT Agent or Perplexity’s Comet, and you’ll quickly realize how limited the technology still is. Making AI agents more robust may take a new set of techniques that the industry is still discovering.

One of those techniques is carefully simulating workspaces where agents can be trained on multi-step tasks — known as reinforcement learning (RL) environments. Much like labeled datasets powered the last wave of AI, RL environments are starting to look like a critical element in the development of agents.

AI researchers, founders, and investors tell TechCrunch that leading AI labs are now demanding more RL environments, and there’s no shortage of startups hoping to supply them.

“All the big AI labs are building RL environments in-house,” said Jennifer Li, general partner at Andreessen Horowitz, in an interview with TechCrunch. “But as you can imagine, creating these datasets is very complex, so AI labs are also looking at third party vendors that can create high quality environments and evaluations. Everyone is looking at this space.”

The push for RL environments has minted a new class of well-funded startups, such as Mechanize Work and Prime Intellect, that aim to lead the space. Meanwhile, large data-labeling companies like Mercor and Surge say they’re investing more in RL environments to keep pace with the industry’s shifts from static datasets to interactive simulations. The major labs are considering investing heavily too: according to The Information, leaders at Anthropic have discussed spending more than $1 billion on RL environments over the next year.

The hope for investors and founders is that one of these startups emerge as the “Scale AI for environments,” referring to the $29 billion data labelling powerhouse that powered the chatbot era.

The question is whether RL environments will truly push the frontier of AI progress.

What is an RL environment?

At their core, RL environments are training grounds that simulate what an AI agent would be doing in a real software application. One founder described building them in recent interview “like creating a very boring video game.”

For example, an environment could simulate a Chrome browser and task an AI agent with purchasing a pair of socks on Amazon. The agent is graded on its performance and sent a reward signal when it succeeds (in this case, buying a worthy pair of socks).

While such a task sounds relatively simple, there are a lot of places where an AI agent could get tripped up. It might get lost navigating the web page’s drop down menus, or buy too many socks. And because developers can’t predict exactly what wrong turn an agent will take, the environment itself has to be robust enough to capture any unexpected behavior, and still deliver useful feedback. That makes building environments far more complex than a static dataset.

Some environments are quite robust, allowing for AI agents to use tools, access the internet, or use various software applications to complete a given task. Others are more narrow, aimed at helping an agent learn specific tasks in enterprise software applications.

While RL environments are the hot thing in Silicon Valley right now, there’s a lot of precedent for using this technique. One of OpenAI’s first projects back in 2016 was building “RL Gyms,” which were quite similar to the modern conception of environments. The same year, Google DeepMind trained AlphaGo — an AI system that could beat a world champion at the board game, Go — using RL techniques within a simulated environment.

What’s unique about today’s environments is that researchers are trying to build computer-using AI agents with large transformer models. Unlike AlphaGo, which was a specialized AI system working in a closed environments, today’s AI agents are trained to have more general capabilities. AI researchers today have a stronger starting point, but also a complicated goal where more can go wrong.

A crowded field

AI data labeling companies like Scale AI, Surge, and Mercor are trying to meet the moment and build out RL environments. These companies have more resources than many startups in the space, as well as deep relationships with AI labs.

Surge CEO Edwin Chen tells TechCrunch he’s recently seen a “significant increase” in demand for RL environments within AI labs. Surge — which reportedly generated $1.2 billion in revenue last year from working with AI labs like OpenAI, Google, Anthropic and Meta — recently spun up a new internal organization specifically tasked with building out RL environments, he said.

Close behind Surge is Mercor, a startup valued at $10 billion, which has also worked with OpenAI, Meta, and Anthropic. Mercor is pitching investors on its business building RL environments for domain specific tasks such as coding, healthcare, and law, according to marketing materials seen by TechCrunch.

Mercor CEO Brendan Foody told TechCrunch in an interview that “few understand how large the opportunity around RL environments truly is.”

Scale AI used to dominate the data labeling space, but has lost ground since Meta invested $14 billion and hired away its CEO. Since then, Google and OpenAI dropped Scale AI as a customer, and the startup even faces competition for data labelling work inside of Meta. But still, Scale is trying to meet the moment and build environments.

“This is just the nature of the business [Scale AI] is in,” said Chetan Rane, Scale AI’s head of product for agents and RL environments. “Scale has proven its ability to adapt quickly. We did this in the early days of autonomous vehicles, our first business unit. When ChatGPT came out, Scale AI adapted to that. And now, once again, we’re adapting to new frontier spaces like agents and environments.”

Some newer players are focusing exclusively on environments from the outset. Among them is Mechanize Work, a startup founded roughly six months ago with the audacious goal of “automating all jobs.” However, co-founder Matthew Barnett tells TechCrunch that his firm is starting with RL environments for AI coding agents.

Mechanize Work aims to supply AI labs with a small number of robust RL environments, Barnett says, rather than larger data firms that create a wide range of simple RL environments. To this point, the startup is offering software engineers $500,000 salaries to build RL environments — far higher than an hourly contractor could earn working at Scale AI or Surge.

Mechanize Work has already been working with Anthropic on RL environments, two sources familiar with the matter told TechCrunch. Mechanize Work and Anthropic declined to comment on the partnership.

Other startups are betting that RL environments will be influential outside of AI labs. Prime Intellect — a startup backed by AI researcher Andrej Karpathy, Founders Fund, and Menlo Ventures — is targeting smaller developers with its RL environments.

Last month, Prime Intellect launched an RL environments hub, which aims to be a “Hugging Face for RL environments.” The idea is to give open-source developers access to the same resources that large AI labs have, and sell those developers access to computational resources in the process.

Training generally capable in RL environments can be more computational expensive than previous AI training techniques, according to Prime Intellect researcher Will Brown. Alongside startups building RL environments, there’s another opportunity for GPU providers that can power the process.

“RL environments are going to be too large for any one company to dominate,” said Brown in an interview. “Part of what we’re doing is just trying to build good open-source infrastructure around it. The service we sell is compute, so it is a convenient onramp to using GPUs, but we’re thinking of this more in the long term.”

Will it scale?

The open question around RL environments is whether the technique will scale like previous AI training methods.

Reinforcement learning has powered some of the biggest leaps in AI over the past year, including models like OpenAI’s o1 and Anthropic’s Claude Opus 4. Those are particularly important breakthroughs because the methods previously used to improve AI models are now showing diminishing returns. 

Environments are part of AI labs’ bigger bet on RL, which many believe will continue to drive progress as they add more data and computational resources to the process. Some of the OpenAI researchers behind o1 previously told TechCrunch that the company originally invested in AI reasoning models — which were created through investments in RL and test-time-compute — because they thought it would scale nicely.

The best way to scale RL remains unclear, but environments seem like a promising contender. Instead of simply rewarding chatbots for text responses, they let agents operate in simulations with tools and computers at their disposal. That’s far more resource-intensive, but potentially more rewarding.

Some are skeptical that all these RL environments will pan out. Ross Taylor, a former AI research lead with Meta that co-founded General Reasoning, tells TechCrunch that RL environments are prone to reward hacking. This is a process in which AI models cheat in order to get a reward, without really doing the task.

“I think people are underestimating how difficult it is to scale environments,” said Taylor. “Even the best publicly available [RL environments] typically don’t work without serious modification.”

OpenAI’s Head of Engineering for its API business, Sherwin Wu, said in a recent podcast that he was “short” on RL environment startups. Wu noted that it’s a very competitive space, but also that AI research is evolving so quickly that it’s hard to serve AI labs well.

Karpathy, an investor in Prime Intellect that has called RL environments a potential breakthrough, has also voiced caution for the RL space more broadly. In a post on X, he raised concerns about how much more AI progress can be squeezed out of RL.

“I am bullish on environments and agentic interactions but I am bearish on reinforcement learning specifically,” said Karpathy.

These agents, to many in the tech industry, are the next evolution in our new AI-powered future, where artificial intelligence not only acts as an assistant, but can autonomously complete complex multi-step actions with just some human intervention in sensitive sectors like healthcare, and no human intervention in lower-risk areas.

But at least in the short term, the real battle between AWS and agentic AI competitors may depend less on technology differentiation, and more on who employs the most quality talent to help guide large corporations on where to even begin with AI agents.

Businesses “are frustrated because they want someone to tell them what to do and how to do it,” Dave Nicholson, chief technology advisor at The Futurum Group, told Fortune. “There isn’t enough [talent] to go around. Humans are the bottleneck.”

Nicholson added that AWS and other cloud and large tech companies will need to heavily lean on partner companies to assist with customer education and implementation too.

The business case for agents was pushed into the forefront last year by Salesforce, with the announcement of a new division it calls Agentforce. Google, OpenAI and other cloud and technology players have since rushed to announce AI agent tools and services geared toward corporations. On Thursday, a day after AWS’s showed off its agent tools, OpenAI announced a new, general purpose agent for users of its ChatGPT product.

Fear of missing out

With just about every CEO these days under pressure to craft an AI strategy, the incoming AI agents may be poised to capitalize on the situation.

“This is the highest level of ‘fear of missing out’ ever among behemoths in the IT industry right now,” Nicholson said. “These are existential decisions being made at Microsoft, Google, and Amazon.”

In an interview with Fortune after his keynote presentation announcing a new in-house collection of agent-building services dubbed AgentCore as well as a marketplace for agents, AWS VP of agentic AI, Swami Sivasubramanian said that Fortune 500 execs whose companies don’t start experimenting with the technology risk missing out on a transformational moment as pivotal as the creation of the internet.

“Agents are fundamentally going to change how we work and how we live,” Sivasubramanian said when asked how execs at Fortune 500 companies can be sure that their investments in building or deploying AI agents isn’t supplanted by a new shiny technology of the moment next year. The executive provided an example of how AI technologies will make it feasible for an agent to, for example, not only plan an itinerary for a trip, but do all of the bookings too.

“You can give it a high level objective, like, ‘Hey, create me a 10 day itinerary in December to visit Australia,’” he said. “It actually understands the objective. Breaks it down into…I need a flight, I need activities to go see in these cities, and then, based on my preferences, it creates a customized itinerary, and actually also secures reservations by calling APIs.”

That’s the type of personal, tangible, example that gives this AWS executive and other proponents of AI agents, the belief that many customer experiences can be overhauled, or created from scratch, with this technology — in ways that might even be hard to envision now.

Agentic rolemodels needed

Slick as some of these scenarios may sound however, the reality is that there are currently few examples of corporations using agents at massive scale. The green field of opportunity is sure to be attractive for some, but it’s also a big challenge for the companies selling agentic products and tools since there are not many real-world examples to guide or inspire.

Amazon Web Services’ market leadership in cloud computing should serve as some advantage, providing a large existing customer base to sell to. And because those companies’ operations are already dependent on AWS, they have more patience for any bumps Amazon experiences as it refines its AI agent business.

“They’re more likely to get two or three strikes,” Nicholson said of AWS and its AI agent rollout.

But it’s an open question whether AWS’ initial focus on heavily marketing its new agentic tools to software developers versus the executives with the purse strings will prove problematic.

“They have disjointed messaging,” Mark Beccue, an analyst at the research firm Omdia, told TechTarget. “When talking about agents, you must have the complete story.”

AWS’ Sivasubramanian said that most C-suite customers that he meets with naturally look inward to how their own organization runs when considering where and how to deploy AI agents first to help automate, or reduce the time to complete, boring, repetitive tasks.

This, of course, raises the question of when and how AI agents will disrupt or displace jobs and in which areas. Amazon CEO Andy Jassy recently weighed in on the overall AI boom in an employee memo, saying that while these technologies will both eliminate current roles while creating new ones, “we expect that this will reduce our total corporate workforce [over the next few years] as we get efficiency gains from using AI extensively across the company.” On Thursday, a day after AWS’ agent-focused summit, the company carried out layoffs of at least hundreds of employees.

A day earlier, Sivasubramanian, perhaps not surprisingly, struck an optimistic tone when discussing a new world full of AI agents that now Amazon — and many rivals — are rushing to bring to fruition.

“Yes, in the short term, if you look at [past] transformations, there were actually changes on the specific job categories [in which people worked], “but then we as humans have really adapted to these changes and then started working on different things. You don’t find people who are doing Y2K engineering anymore.”