The Answer to Dreamforce 2026: AI Agent, Has It Finally Turned from Demo to Revenue?

Those who sell shovels during a gold rush make a fortune, but if miners can’t find gold, shovels will eventually stop selling.

Written by: Jim, MSX Maitong

Edited by: Frank, MSX Maitong

Has AI finally started to "make money" after spending so much?

In the past two years, the most thrilling narrative in the U.S. stock market surrounding AI has been almost entirely about counting money on the infrastructure side: NVIDIA printing money, TSMC running at full capacity, Broadcom quietly profiting from custom chips, with funds flowing into the entire physical world through electricity and optical modules.

The entire investment logic is simple and crude— as long as large models continue to grow, computing power is the most certain hard currency.

However, standing in the autumn of 2026, after hundreds of billions of dollars in capital expenditures have been poured into data centers, the market's patience is visibly narrowing, beginning to question when this computing power will actually translate into real revenue on financial statements.

This watershed moment has already opened a gap in the U.S. stock market in mid-September. On September 14, influenced by discussions about a potential slowdown in AI development, chip stocks that had been on a run for two years collectively faced pressure, while security giants CrowdStrike and Palo Alto surged over 13% in a single day, and even long-stagnant Salesforce and ServiceNow saw a welcome return of buying interest.

At this critical juncture, Dreamforce 2026 also kicked off.

Setting aside the dazzling keynotes and conceptual packaging, the entire conference was essentially answering the same core question for everyone: Can AI Agents really turn demos into cold hard cash?

1. Saying Goodbye to "Chat Partners", AI Begins to "Get the Job Done"

The enterprise AI of the past two years has mostly been just an advanced version of Copilot.

Employees ask a question, and AI helps summarize documents, generate emails, write code, or organize meeting notes; it can improve efficiency, but a live person must sit in front of the computer, feeding instructions step by step.

This year, Salesforce has finally stopped telling this "co-pilot" story.

For example, the Hunter responsible for customer acquisition, previously, Copilot could at most help polish a cold email; now the design allows it to formulate plans, analyze data, and follow up continuously for weeks. Just before Dreamforce, it expanded the Agentforce product line all at once, including customer service agent Casey, IT and HR agent Paige, e-commerce shopping agent Carter, sales agent Hunter, and supply chain backend process agent Marshall.

It is designed to work continuously across systems, steps, and even time, thereby connecting the entire chain: finding potential customers, formulating plans, following up continuously, adjusting strategies based on new customer information, requesting sales personnel approval when needed, and then continuing to push tasks forward.

This process can last for weeks or even longer. Although Hunter is still in the pilot phase and is expected to officially go GA in November this year, it is not yet a fully mature commercial product, but the underlying business logic has already changed.

In the past, the relationship between humans and AI was more like constantly issuing prompts; in the future, what enterprises want to buy may be a set of digital labor that can read data, call tools, execute tasks, and leave audit trails.

Once the product shifts from "function" to "digital labor", the ROI calculation becomes completely transparent.

How well a chatbot performs can easily remain a subjective experience; but if a customer service agent can directly reduce tickets, a sales agent can generate pipelines, and a backend agent can reduce manual operations, enterprises can directly calculate how much they paid for it and how much work it accomplished for them.

AI only reaches this point when it truly touches the threshold of business.

2. Saying Goodbye to Seats: Charging by "Work Done"

This is also one of the most noteworthy aspects of this year's Dreamforce.

Demos at the launch event can always be edited to look spectacular, but to judge whether commercialization has landed, there are only two indicators— is there real cash flow, and is there real usage?

Salesforce's latest quarterly report provided a set of very noteworthy data, with Agentforce ARR surpassing $1.5 billion, a year-on-year increase of 240%; the combined ARR of Agentforce and Data 360 is close to $3.9 billion, with a year-on-year growth of over 210%. Meanwhile, Agentforce and Slack have cumulatively completed 7 billion Agentic Work Units, with 3.2 billion of those in the second quarter alone, a quarter-on-quarter growth of 97%.

Although this $1.5 billion ARR includes Slackbot and other AI assets, which somewhat carries statistical packaging, it at least indicates that enterprises are no longer just "tasting".

In customer cases disclosed by Salesforce, about 50% of chat inquiries for Engine can now be fully resolved by its agents; about 60% of the sales pipeline for Perk is established by sales agents; about 70% of administrative requests for Autism Queensland are handled by agents; Hibbett's AI has participated in about 90% of core shopping processes; and when Anthropic uses Fin, about 79% of customer service conversations Fin encounters can be resolved automatically.

These statistics vary in their definitions and all come from Salesforce's official disclosures, so they cannot yet be used to prove that the entire agent industry has matured, but they at least represent that the core focus has genuinely shifted from "how smart is it" to "how much work has it helped me carry".

More importantly, Salesforce's own charging method has also begun to adapt to this change.

Agentforce has launched Flex Credits. In the current public pricing, 100,000 Credits cost $500, and a standard Agent Action consumes 20 Credits, which is about $0.10. Enterprises can directly pay according to the usage of agents updating a record, processing a process, or executing an action.

This may be more important than the $1.5 billion ARR itself.

After all, the core business unit of traditional SaaS is Seats; for 1,000 employees, you sell 1,000 accounts.

However, the problem that agents ultimately want to solve is precisely to allow fewer people to accomplish more work. If a business that originally required 10 people can now be handled by just 3 people plus a batch of agents, then software companies relying solely on seat fees may face an awkward problem: the more successful AI becomes, the fewer human seats there will be.

Therefore, Salesforce is eager to implement usage-based billing, essentially trying to find a new pricing unit for the post-SaaS era.

3. Underlying Currents: Security, Interfaces, and Profit Margin Challenges

If this change holds, it will not only affect Salesforce.

It is well known that the most important keyword in the AI market over the past few years has been CapEx.

Training models requires GPUs, GPUs require data centers, and data centers require electricity, networks, optical modules, and storage. Therefore, as long as hyperscalers continue to increase capital expenditures, the infrastructure chain can continue to benefit.

Once agents are truly commercialized, AI will generate another value chain, which is from models entering enterprise workflows, then connecting enterprise data, identities, permissions, security, and real business systems.

For example, companies like Salesforce and ServiceNow master workflows; platforms like Snowflake connect enterprise data; while security vendors like CrowdStrike, Palo Alto Networks, SailPoint, and Varonis will face increasingly complex problems as the number of agents increases.

In the future, an enterprise may not only have tens of thousands of employees but also operate thousands or even tens of thousands of "non-human identities". These agents can read internal documents, call APIs, modify CRMs, send emails, operate code, and even participate in transaction processes.

At that time, the problems enterprises face will become who are they? What can they see? What can they do? Who do they represent? Can issues be traced? Therefore, as agents get closer to real production environments, capabilities that previously seemed to be backend, such as Identity, Permission, Data Governance, and Runtime Security, will increasingly approach the core of AI applications.

This is also why the rotation of cybersecurity stocks on September 14 is worth paying attention to. The attack surface of AI is expanding from human accounts, devices, and servers to an increasing number of agents with autonomous execution capabilities. CrowdStrike and Palo Alto are not just reacting to short-term sentiment; they are facing the severe spillover of AI's attack surface.

Meanwhile, this commercialization path is still far from being fully realized, and there are two hard hurdles ahead:

• Gross margin erosion: Every time an agent acts, there is a real cost of reasoning, retrieval, and cloud resource consumption behind it. After accounting for these costs, can it still maintain the attractive high gross margins of traditional SaaS?
• Self-cannibalization: Can the incremental usage from agents outpace the decline in human seats? If not, it will just be a transfer from one hand to the other.

Dreamforce 2026 also introduces a change that may make this test more interesting.

Salesforce's newly launched AIforce is opening up data, workflows, business logic, permissions, and governance capabilities that were originally locked within the CRM interface to new AI interfaces like Claude, Slack, and Agentforce Coworker. The first batch of Claudeforce offers 37 pre-built sales skills; Agentforce Coworker gained 100,000 user activations within 35 days of its launch—of course, activation numbers cannot be directly equated to active users or paying customers.

This means that competition in enterprise software may undergo another change.

Previously, Salesforce's most important aspect was that CRM interface; in the future, users may not even need to open Salesforce but can directly let AI call the data and workflows behind Salesforce within Claude or Slack. In other words, the interface is disappearing, but the underlying data, permissions, processes, and business logic may become more valuable.

If we look at this round of AI market trends over a longer period, the past few years have actually gone through two very clear stages.

• The first stage is training AI, where the biggest beneficiaries are GPUs, advanced processes, and ASICs;
• The second stage is building AI, where data centers, electricity, optical communications, and storage begin to bear huge capital expenditures;
• And the signals released by Dreamforce 2026 so far mark that the market has officially hit the wall of the third stage: turning computing power into productivity.

This path has not yet been fully realized, and it is premature to say that software has "taken over" semiconductors. But at least, enterprise AI has for the first time shown a more complete business closed loop—someone buys agents, agents start to get the job done, work generates usage, and usage turns into revenue for software companies.

Overall, the business closed loop has already begun to take shape, although it is still quite bumpy.

But the most honest logic in the business world will never change—whoever can truly convert the anxiety enterprises spend on computing power into profits on the balance sheet will be the winner of the next round of the game.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.