Salesforce is arguing that AI is an expansion, not an extinction event
Salesforce is trying to make a difficult case to Wall Street: that the rise of AI agents will not hollow out the enterprise software business that made the company powerful, but instead create a new growth engine for it. According to the supplied report, CEO Marc Benioff is pushing back directly against the idea that AI could make traditional software seats less valuable or even obsolete. His answer is a combination of product strategy, customer anecdotes, and a newly invented measurement system designed to show that AI can be counted, priced, and managed inside the Salesforce ecosystem.
The pressure behind that argument is real. The report says Salesforce stock is down 28 percent since the start of the year, reflecting investor concern that AI agents might reduce headcount at customer companies and weaken the logic of per-seat licensing. Another fear runs alongside that one: if generative AI makes it easier for companies to “vibe code” their own tools, then the premium attached to large software suites could come under pressure as well.
Salesforce is not denying the disruption. It is trying to claim a privileged position inside it.
‘Agent Albert’ is meant to push the company past today’s limitations
Benioff’s most concrete response, based on the supplied text, is a new AI product codenamed “Agent Albert,” expected to launch by the end of the year. The description is brief but ambitious: the platform is designed to analyze users automatically and take actions on its own.
That phrasing matters because it implies a move beyond basic chatbot behavior toward higher-autonomy workflows. If Salesforce can persuade customers that AI agents can operate inside enterprise guardrails while drawing on business data and existing workflows, the company can argue that AI intensifies the need for a trusted software layer rather than replacing it.
Benioff is also making that case on security and compliance grounds. The report says he argues homegrown AI solutions are too risky when sensitive enterprise data is involved. That is a familiar enterprise-software defense, but it may be more potent in the age of autonomous agents than it was in the age of simple cloud applications. As AI systems are given more discretion, the value of governance, auditing, and policy enforcement rises.
Agentforce adoption shows both momentum and limits
The challenge for Salesforce is that its current AI track record is mixed. The report says Agentforce, launched in late 2024, has been adopted by 23,000 of Salesforce’s 150,000 customers. That is meaningful uptake, but it also indicates that the majority of customers have not yet adopted the product.
The case studies cited in the source text show why the picture is uneven. Pearson reportedly saw a 40 percent increase in customer inquiries resolved automatically, suggesting that routine, high-volume tasks are a natural fit for the current generation of AI tools. But Pandora, the jewelry maker, said Agentforce struggled when customer requests became vague and required reliable recommendations.
That split is important because it reveals where enterprise AI still has friction. Structured tasks with clear parameters can generate measurable value. Ambiguous, higher-judgment interactions remain harder. Salesforce’s future in AI may depend less on whether agents are useful in general than on how quickly the company can push them from narrow automation toward dependable performance in messy real-world contexts.
Why the ‘Agentic Work Unit’ matters
Salesforce has also introduced what may be the most revealing part of its strategy: a new metric called the “Agentic Work Unit,” or AWU. The idea, according to the source, is to quantify AI’s impact by linking its capabilities to concrete outcomes such as resolved inquiries.
This is more than a branding exercise. Enterprise software companies need a way to turn AI from a flashy feature into an operational and financial model. If seats become a weaker proxy for value in an AI-heavy world, vendors need new units of measurement. AWU appears to be Salesforce’s attempt to create one.
Whether customers and investors accept that metric is another question. New units succeed only when they are intuitive, auditable, and meaningfully tied to business outcomes. Still, the introduction of AWU signals that Salesforce understands a central problem of the AI era: companies will not simply buy “intelligence.” They will want to buy measurable work, reduced handling time, increased resolution rates, or some other trackable output.
The broader stakes for enterprise software
The bigger issue is not just Salesforce’s quarterly performance. It is whether large enterprise software vendors can redefine themselves before AI changes the basis of competition. The “SaaSpocalypse” theory described in the report imagines a world where agents reduce demand for seats and lower the barriers to custom software creation. Benioff’s rebuttal is that enterprise complexity, security, and compliance still favor established platforms.
Both arguments can be partly true at once. AI may compress some kinds of software value while expanding others. Standalone features may become easier to replicate, but trusted data layers, orchestration, workflow management, and governance may become more important. In that environment, the winners would not be the companies that deny the shift, but the ones that redefine what exactly they sell.
Salesforce is trying to do that in public. “Agent Albert” is the product signal. AWU is the pricing-and-proof signal. The mixed performance of Agentforce is the reminder that the company is still in the transition, not beyond it.
For now, the most defensible conclusion from the supplied reporting is straightforward: Salesforce sees AI agents as existential to its future, but not necessarily existential in the way critics mean. The company is betting that the next generation of enterprise software will be judged not by how many human seats it supports, but by how much machine-assisted work it can securely deliver. That is a plausible strategy. The harder part will be proving it in customer results rather than conference rhetoric.
This article is based on reporting by The Decoder. Read the original article.
Originally published on the-decoder.com
