Facing enterprise software giants: OpenAI seeks "disruption," Anthropic chooses "symbiosis"

OpenAI and Anthropic are taking very different strategic stances in the enterprise software market—one loudly declaring disruptive intentions, the other deliberately emphasizing partnerships. Analysts believe this divergence is reshaping market expectations about AI’s impact on traditional software industries.

On February 24, Anthropic announced new features for its Claude Cowork AI software, demonstrating how enterprises can access and invoke data stored in business applications like DocuSign, LegalZoom, and Salesforce through the tool.

Following the announcement, enterprise software stocks previously pressured by AI disruption expectations generally rebounded—Figma rose 10%, ServiceNow edged up 1.4%, and Salesforce increased 4%. The market’s positive response indicates that Anthropic’s actual positioning is not as aggressive as some investors feared.

Meanwhile, last week, OpenAI clearly stated at an investor conference that its AI agents and future products will have the capability to replace software from companies like Salesforce, Workday, Adobe, and Atlassian, presenting revenue data from these firms to contrast with OpenAI’s own projected 2030 revenue.

This private stance sharply contrasts with Anthropic’s public posture and further heats up the defensive stance within the enterprise software industry.

Anthropic bets on “replacing human labor” rather than “replacing software”

Tech media outlet The Information reports that when Anthropic launched the Claude Cowork new features, it deliberately focused its narrative on replacing human labor rather than software tools. The reasoning is:

The new AI tools will continue to call on existing enterprise software, and companies will still pay for these software licenses; the revenue streams for software vendors will not disappear.

At the launch event, Anthropic specifically invited its Chief Economist Peter McCrory, who stated that AI’s impact on the labor market will be “highly uneven”—high-skilled workers will leverage AI to boost productivity, while low-skilled workers involved in basic data entry face the risk of being replaced.

Derek Hernandez, an analyst at PitchBook Data, further pointed out that Claude’s capabilities aimed at the financial services sector could threaten jobs in that industry—“particularly investment banking seats, equity research positions, especially entry-level white-collar roles.”

The report notes that it’s worth mentioning that Anthropic employees are still using various traditional enterprise applications. While this detail doesn’t fully eliminate concerns from software vendors—whose business models depend on large enterprise user bases—it does send a market signal quite different from OpenAI’s.

OpenAI shows its “disruption card” to investors

In contrast, OpenAI’s ambitions in the enterprise software market are more overt.

According to reports, at last week’s investor conference, OpenAI explicitly listed giants like Salesforce, Workday, Adobe, and Atlassian as potential targets for replacement, and showcased their revenue scales against OpenAI’s own projected 2030 revenue, aiming to demonstrate the market potential.

OpenAI also disclosed an estimate:

Using ChatGPT, average employees save about 50 minutes of work daily, roughly $50 per person per day. The enterprise ChatGPT subscription starts at only $25 per user per month, and OpenAI believes it currently captures only a tiny fraction of the value created. This estimate partly references data from OpenAI’s shareholder Ark Invest.

These private statements align with the broader interpretation of OpenAI’s recent launch of the new “Frontier” AI product—OpenAI is positioning its technology as a gateway for enterprise applications, gradually influencing corporate software and AI procurement decisions through control of enterprise data access.

Traditional software vendors build defenses as customer behaviors subtly shift

Faced with the encroachment of AI labs, traditional enterprise software providers are actively building defenses.

Companies like ServiceNow and Microsoft emphasize to customers that their software is more reliable and compliant than experimental products from AI labs. HubSpot is considering charging extra for customers who want to invoke their system data via AI agents.

However, whether traditional software vendors can truly block AI agent infiltration remains uncertain. The core design logic of AI agents is to take over user computers and operate various applications manually.

Analysts point out that the most powerful moat for traditional enterprise software vendors currently lies in their long-standing experience with global data compliance and privacy regulations—capabilities that experimental AI products are unlikely to replicate in the short term.

Although most enterprise clients have yet to fully replace existing software with AI, AI agents are changing how employees interact with software and are generating tangible cost substitution effects in specific scenarios.

According to reports, a cybersecurity executive said he avoided over $100,000 annually in CrowdStrike subscription fees by integrating AI agents. CrowdStrike’s product was originally used to automate employee account management—flagging suspicious logins, identifying and locking dormant accounts of former employees.

He instead adopted an AI agent from startup Torq—powered by OpenAI and Anthropic models—that directly connects to raw login data collected via Microsoft software, achieving the same functions at a lower cost.

This case reveals a potential pathway for AI to disrupt the enterprise software market: even if companies don’t immediately abandon their current software, AI agents could make some software products seem “redundant” in employees’ eyes, gradually establishing AI itself as the core tool of work. CrowdStrike stated that it now allows AI agents to access its software to enable integrated use.

Risk warning and disclaimer

Market risks exist; investments should be cautious. This article does not constitute personal investment advice and does not consider individual users’ specific investment goals, financial situations, or needs. Users should consider whether any opinions, viewpoints, or conclusions herein are suitable for their particular circumstances. Invest at your own risk.