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.

AI companies are making their much-anticipated enterprise plays, but the results are wildly inconsistent. Just this week, Deloitte announced it’s rolling out Anthropic’s Claude to all 500,000 employees. On the very same day, the Australian government forced Deloitte to refund a contract because their AI-generated report was riddled with fake citations. It’s a perfect snapshot of where we are: companies racing to adopt AI tools before they’ve figured out how to use them responsibly. 

On this episode of Equity, Kirsten Korosec, Anthony Ha, and Sean O’Kane dig into the messy reality of AI in the workplace, plus funding news and regulatory drama across tech and transportation. 

Listen to the full episode to hear more news from the week, including: 

• Zendesk’s claim that its new AI agents can handle 80% of customer service tickets autonomously, and what happens in the other 20% 

Equity is TechCrunch’s flagship podcast, produced by Theresa Loconsolo, and posts every Wednesday and Friday.  

Subscribe to us on Apple Podcasts, Overcast, Spotify and all the casts. You also can follow Equity on X and Threads, at @EquityPod. 

Meanwhile, employees using personal AI tools succeed at a 40% rate.

The disconnect isn’t technological—it’s operational. Companies are struggling with a crisis in AI measurement.

Three questions I invite every leadership team to answer when they ask about ROI from AI pilots:

1. How much are you spending on AI tools companywide? 
2. What business problems are you solving with AI?
3. Who gets fired if your AI strategy fails to deliver results?

That last question usually creates uncomfortable silence.

As the CEO of Lanai, an edge-based AI detection platform, I’ve deployed our AI Observability Agent across Fortune 500 companies for CISOs and CIOs who want to observe and understand what AI is doing at their companies.

What we’ve found is that many are surprised and unaware of everything from employee productivity to serious risks. At one major insurance company, for instance, the leadership team was confident they had “locked everything down” with an approved vendor list and security reviews. Instead, in just four days, we found 27 unauthorized AI tools running across their organization.

The more revealing discovery: One “unauthorized” tool was actually a Salesforce Einstein workflow. It was allowing the sales team to exceed its goals — but it also violated state insurance regulations. The team was creating lookalike models with customer ZIP codes, driving productivity and risk simultaneously. 

This is the paradox for companies seeking to tap AI’s full potential: You can’t measure what you can’t see. And you can’t guide a strategy (or operate without risk) when you don’t know what your employees are doing. 

‘Governance theater’

The way we’re measuring AI is holding companies back. 

Right now, most enterprises measure AI adoption the same way they do software deployment. They track licenses purchased, trainings completed, and applications accessed. 

That’s the wrong way to think about it. AI is workflow augmentation. The performance impact lives in interaction patterns between humans and AI, not solely on tool selection.

The way we currently do it can create systematic failure. Companies establish approved vendor lists that become obsolete before employees finish compliance training. Traditional network monitoring misses embedded AI in approved applications such as Microsoft Copilot, Adobe Firefly Slack AI and the aforementioned Salesforce Einstein. Security teams implement policies they cannot enforce, because 78% of enterprises use AI, while only 27% govern it.

This creates what I call the “governance theater” problem: AI initiatives that look successful on executive dashboards often deliver zero business value. Meanwhile, the AI usage that is driving real productivity gains remains completely invisible to leadership (and creates risk).

Shadow AI as systematic innovation

Risk doesn’t equal rebellion. Employees are trying to solve problems. 

Analyzing millions of AI interactions through our edge-based detection models proved what most operating leaders instinctively know, but cannot prove. What appears to be rule-breaking is often employees simply doing their work in ways that  that traditional measurement systems cannot detect.

Employees use unauthorized AI tools because they’re eager to succeed and  because sanctioned enterprise tools succeed in production only 5% of the time, while consumer tools like ChatGPT reach production 40% of the time. The “shadow” economy is more efficient than the official one. In some cases, employees may not even know they’re going rogue.

A technology company preparing for an IPO showed “ChatGPT – Approved” on security dashboards, but missed an analyst using personal ChatGPT Plus to analyze confidential revenue projections under deadline pressure. Our prompt-level visibility revealed SEC violation risks that network monitoring completely missed.

A healthcare system recognized doctors using Epic’s clinical decision support, but missed emergency physicians entering patient symptoms into embedded AI to accelerate diagnoses. While improving patient throughput, this violated HIPAA by using AI models not covered under business associate agreements.

The measurement transformation

Companies crossing the “GenAI divide” identified by MIT, whose Project Nanda identified the remarkable struggles with AI adoption, aren’t those with the biggest AI budgets; they’re those who can see, secure, and scale what actually works. Instead of asking, “Are employees following our AI policy?” they ask, “Which AI workflows drive results, and how do we make them compliant?”

Traditional metrics focus on deployment: tools purchased, users trained, policies created. Effective measurement focuses on workflow outcomes: Which interactions drive productivity? Which creates genuine risk? Which patterns should we standardize organization-wide?

The insurance company that discovered 27 unauthorized tools figured this out. 

Instead of shutting down ZIP code workflows driving sales performance, they built compliant data paths preserving productivity gains. Sales performance stayed high, regulatory risk disappeared, and they scaled the secured workflow companywide—turning compliance violation into competitive advantage worth millions.

The bottom line

Companies spending hundreds of millions on AI transformation while remaining blind to 89% of actual usage face compounding strategic disadvantages. They fund failed pilots while their best innovations happen invisibly, unmeasured and ungoverned.

Leading organizations now treat AI like the biggest workforce decision they’ll make. They require clear business cases, ROI projections, and success metrics for every AI investment. They establish named ownership where performance metrics include AI results tied to executive compensation.

The $8.1 billion enterprise AI market won’t deliver productivity gains through traditional software rollouts. It requires workflow-level visibility distinguishing innovation from violation.

Companies establishing workflow-based performance measurement will capture productivity gains their employees already generate. Those sticking with application-based metrics will continue funding failed pilots while competitors exploit their blind spots.

The question isn’t whether to measure shadow AI—it’s whether measurement systems are sophisticated enough to turn invisible workforce productivity into sustainable competitive advantage. For most enterprises, the answer reveals an urgent strategic gap.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

Fortune Global Forum returns Oct. 26–27, 2025 in Riyadh. CEOs and global leaders will gather for a dynamic, invitation-only event shaping the future of business. Apply for an invitation.

When Salesforce recently rolled out an AI agent on its website, the agent started to hallucinate and wasn’t giving consistent results.