The AI Agent Paradox: Tech CEOs Say They’re Unfazed, but Their SEC Filings Tell a Very Different Story

="">

In boardrooms and on earnings calls, the message from some of the biggest names in software has been remarkably consistent: AI agents won’t eat our lunch. Figma CEO Dylan Field and HubSpot CEO Yamini Rangan have both projected calm confidence about the rise of autonomous AI systems that can perform tasks once handled by their products. But buried in the fine print of their companies’ regulatory filings, the tone shifts dramatically.

The disconnect is striking. And it raises a question investors should be asking more aggressively: Are software executives downplaying the most significant competitive threat in a generation?

The Information reported that both Figma and HubSpot have added or strengthened language in their SEC filings explicitly warning that AI agents pose material risks to their businesses — even as their CEOs publicly minimize those same risks. It’s a pattern that extends well beyond these two companies, touching a fundamental tension running through the entire enterprise software sector right now.

Figma, the collaborative design platform that nearly merged with Adobe in a $20 billion deal before regulators intervened, went public in its IPO filing with risk factors that specifically cite AI agents as a competitive threat. The company warned that AI-driven tools could reduce the need for traditional design software — the very product that made Figma a Silicon Valley darling. Field, meanwhile, has framed AI as an opportunity rather than a threat, positioning Figma’s own AI features as additive to the core product.

HubSpot’s story follows a similar arc. Rangan has spoken publicly about how AI agents will complement rather than replace the company’s marketing, sales, and customer service platform. Yet HubSpot’s risk disclosures now include language about AI agents potentially disrupting the demand for its software. The filings don’t hedge. They warn.

When the Legal Team and the CEO Aren’t Reading From the Same Script

This isn’t unusual in corporate America, of course. Risk factors in SEC filings are drafted by lawyers whose job is to enumerate every conceivable threat, no matter how remote, to shield the company from shareholder lawsuits. CEOs, by contrast, are in the business of selling confidence. But the AI agent risk isn’t some hypothetical scenario cooked up by overzealous attorneys. It’s happening now.

Companies like Anthropic, OpenAI, and Google are building AI agents capable of performing multi-step tasks autonomously — browsing the web, filling out forms, writing code, managing customer interactions, even designing user interfaces. Salesforce has bet heavily on its Agentforce platform. Microsoft’s Copilot agents are being embedded across its product line. The ambition is clear: software that doesn’t just assist humans but replaces entire workflows.

For companies like HubSpot, whose core value proposition is automating marketing and sales processes, the threat is existential in nature. If an AI agent can autonomously manage a sales pipeline — identifying leads, sending personalized outreach, scheduling meetings, updating CRM records — what’s the value of a traditional SaaS platform built around manual and semi-automated workflows? The same logic applies to Figma. If AI agents can generate design assets, iterate on layouts, and produce production-ready interfaces based on natural language prompts, the role of a human designer using collaborative design tools necessarily shrinks.

None of this means these companies are doomed. Far from it. Both Figma and HubSpot have significant advantages: massive user bases, deep integrations into enterprise workflows, network effects, and brand loyalty. But the gap between what executives say publicly and what their lawyers write in filings suggests an internal awareness that the ground is shifting faster than the public narrative admits.

The pattern extends across the software industry. ServiceNow, Workday, and other major SaaS players have similarly updated their risk disclosures to address AI-related competitive threats. The language varies, but the message is consistent: AI could reduce demand for existing products, attract new competitors, or force costly pivots in strategy. Meanwhile, on earnings calls, these same companies talk almost exclusively about AI as a growth driver.

The Investor’s Dilemma: Which Version Do You Believe?

Wall Street, for its part, has largely bought the optimistic narrative. Software stocks have rallied on AI enthusiasm, with investors rewarding companies that announce AI features, AI integrations, and AI partnerships. HubSpot shares have performed well, buoyed in part by the company’s messaging around its own AI capabilities. Figma, still private, has maintained its lofty valuation through its IPO process.

But some analysts are starting to ask harder questions. The bull case for incumbent software companies rests on the assumption that they can successfully integrate AI into their existing platforms and that customers will continue to pay for those platforms even as standalone AI agents become more capable. That’s a big assumption. And it gets bigger every quarter as the capabilities of frontier AI models improve.

Consider the pace of change. A year ago, AI agents were largely a research curiosity — impressive in demos but unreliable in production. Today, Anthropic’s Claude can operate a computer, Google’s Gemini agents can handle complex multi-step research tasks, and OpenAI is building agents that can write and execute code autonomously. The gap between demo and deployment is closing fast.

For HubSpot specifically, the threat vector is direct. The company’s bread and butter is helping small and mid-sized businesses manage customer relationships, automate email campaigns, and track sales pipelines. These are precisely the kinds of tasks that AI agents are becoming good at. A startup armed with an AI agent framework could potentially offer a fraction of HubSpot’s functionality at a fraction of the cost — without requiring users to learn a complex software platform at all.

Figma faces a different but related challenge. The rise of AI-generated design threatens to compress the time and skill required to produce high-quality user interfaces. If a product manager can describe what they want in plain English and an AI agent produces a polished design in minutes, the collaborative design workflow that Figma enables becomes less central to the product development process. Not irrelevant. But less central.

Both companies are responding. Figma has invested heavily in AI features, including its controversial Make Design feature that generated plagiarism concerns before being pulled and reworked. HubSpot has introduced AI-powered tools across its platform, including content generation, predictive lead scoring, and automated customer service responses. These moves are smart and necessary. But they also represent a defensive posture — an acknowledgment that standing still isn’t an option.

The broader question is whether incumbent software companies can adapt quickly enough. History offers mixed guidance. Microsoft successfully pivoted to cloud computing under Satya Nadella, turning what looked like an existential threat into a growth engine. But plenty of other companies — Blackberry, Nokia, Yahoo — failed to make similar transitions despite seeing the threat clearly.

What makes the AI agent threat particularly tricky is that it doesn’t just compete with existing products. It potentially eliminates the category. Traditional software automates specific tasks within a defined interface. AI agents promise to automate entire workflows without requiring a dedicated interface at all. That’s not a competitive threat in the conventional sense. It’s a structural one.

So when Dylan Field says he’s not worried about AI agents, and then his company’s S-1 filing says investors should be, the rational response is to weight the filing more heavily. Not because CEOs are dishonest — they genuinely believe they can adapt. But because legal disclosures are written with the benefit of internal data, board-level discussions, and a fiduciary obligation to be candid about risks.

The software industry is entering a period where the gap between public optimism and private concern may widen further. Companies will continue to announce AI features and tout their AI strategies. Executives will continue to project confidence. And buried in the back pages of their SEC filings, the lawyers will continue to write what the CEOs won’t say out loud.

Investors would do well to read both.