Investors Fear AI Is Going to Eat Software — What That Means for AI in Recruiting

A CNBC Television segment featuring Jefferies analyst Brent Thill explored why investors are increasingly convinced that AI will "eat" the software industry. The conversation covered a wave of selling across application software, the concentrated winner-takes-most nature of AI infrastructure, and how misconceptions about rapid AI-driven automation are reshaping valuations. This article breaks down the discussion, explains the practical differences between infrastructure and applications, and applies the lessons directly to one fast-evolving use case: AI in recruiting.

Why investors think AI will eat software

The market has increasingly priced a scenario where a handful of infrastructure companies capture the vast majority of value from AI, while traditional application software vendors suffer multiple compression. Stocks such as Atlassian and major CRM vendors have fallen more than twenty percent year-to-date in response to the narrative that AI will commoditize application-level functionality.

Two core beliefs underlie this investor behavior. First, the winners are thought to be the infrastructure providers that enable large-scale models to run: GPU makers, cloud platforms, and enterprise data-stack specialists. Second, many investors believe that once large models become available, the marginal cost and complexity of building applications will shrink dramatically—making product differentiation and recurring SaaS economics harder to sustain.

Infrastructure winners vs. application casualties

The transcript makes a clear distinction: companies like NVIDIA, Oracle, and Microsoft are beneficiaries today because the application vendors need them to scale AI. Oracle’s backlog growth and cloud-related demand illustrate how infrastructure vendors are capturing new enterprise budgets tied to AI initiatives.

By contrast, many application companies have yet to monetize AI meaningfully. Salesforce and other application vendors are growing, but not at the breakneck rates that would justify current multiples if investors believe those multiples will be under threat. This mismatch—where infrastructure captures immediate spending and applications lag in monetization—helps explain the divergent market performance.

The misconception: "One prompt will build everything"

A recurring theme in the CNBC segment is a misunderstanding about how software is built and maintained. The notion that a simple prompt will enable anyone to conjure a fully functional, secure, audited, and integratable business application is an oversimplification.

Software development entails data modeling, workflow design, integration, testing, role-based access control, maintenance, compliance, and change management. Generative models can assist many of these tasks, but they do not replace the architecture, project management, and ongoing support required by enterprises. That gap is especially relevant for vertical or domain-specific applications where nuance, context, and regulatory constraints matter.

Where the market sees opportunity: discounted quality

The sell-off among application names has created potential entry points. The segment highlights companies like Monday.com and Intuit as examples of businesses that were punished harshly—sometimes overreacting to short-term headwinds or go-to-market changes. Investors who separate transient execution issues from durable product-market fit see value in certain application names.

Several factors make these names worth scrutinizing:

• Underlying product strength and customer retention.
• The ability to layer AI features into existing workflows, creating upsell and stickiness.
• Large addressable markets where domain expertise still matters.

Which companies look vulnerable?

Not every application vendor will be insulated. The firms most at risk are those that cannot effectively embed AI into product experiences or that rely on commoditized workflows that a large model can mimic without deep integrations. The segment flags IBM as an example where services-led AI work is strong, but questions remain about transitioning to software-led revenue that captures recurring value.

In the race among AI-focused enterprise players, the contrast between Palantir and C3 AI was used to demonstrate how execution and product-market fit matter. Palantir’s traction in operational AI deployments gives it an edge, while C3 AI’s metrics have shown stress. Adobe also stands out as a specific worry: consumer-grade generative image tools raise investor concerns about how Adobe will defend and monetize its creative software franchise.

Applying this to AI in recruiting

The recruiting function is a useful lens for understanding the difference between infrastructure and applications and for assessing where AI will augment versus replace. AI in recruiting is often framed as an immediate wave of automation: scour the web, rank candidates, generate outreach, and let hiring managers pick from an AI-curated list. But the reality is more nuanced.

AI in recruiting can and will improve sourcing, screening, matching, and candidate engagement. Modern applicant tracking systems (ATS) and talent platforms can integrate models for resume parsing, candidate scoring, interview scheduling automation, and candidate experience personalization. Still, there are several reasons why AI in recruiting is likely to augment rather than entirely subsume recruiting software in the near term:

• Data integration: Recruiting workflows require connections to HRIS systems, payroll, background checks, and calendar systems. Those integrations are not trivial and demand engineering investment beyond prompt-driven prototyping.
• Compliance and fairness: Hiring decisions are subject to legal and regulatory constraints. AI outputs need explainability, audit trails, and guardrails to mitigate bias—features that mature recruiting platforms are built to support.
• Domain nuance: Different industries, roles, and seniority levels require distinct evaluation criteria. Off-the-shelf models can provide a baseline, but customization and continuous tuning are required to achieve reliable outcomes.
• User experience and change management: Recruiting teams depend on workflows and reporting—applying AI without careful UX and governance often leads to limited adoption.

Therefore, AI in recruiting will likely be an enhancement to established recruiting platforms rather than a wholesale replacement. Vendors that embed model-driven capabilities into their core product—while preserving integrations, compliance, and workflow orchestration—will retain value and pricing power.

Practical implications for buyers and investors

For enterprise buyers:

• Prioritize vendors that demonstrate operational AI—models that are integrated, monitored, and governed within existing workflows—over those that merely showcase generative demos.
• Assess how potential AI features affect total cost of ownership, hiring velocity, and candidate quality, not just the number of automated tasks.
• Require vendor commitments on explainability, bias mitigation, and data privacy when evaluating AI in recruiting solutions.

For investors:

• Differentiate between infrastructure plays (beneficiaries of raw compute and model hosting) and application plays (companies that must monetize AI through product-led upgrades).
• Look for application vendors that are already monetizing AI or have credible paths to convert infrastructure spending into product-led revenue growth.
• Recognize that market overreactions can create buying opportunities in high-quality software businesses that simply need time to layer in AI and demonstrate monetization.

Key takeaways

Brent Thill’s comments crystallize a few important realities: AI is reshaping the enterprise technology landscape, but it is not a single monolithic force that instantly removes the need for domain-specific software. Infrastructure is winning the immediate spending war, while many application vendors face valuation pressure as markets price in potential disruption.

When applying these insights to the specific case of AI in recruiting, the likely outcome is augmentation rather than wholesale replacement. Recruiting systems that embrace AI will gain productivity and intelligence, but the complexity of integrations, compliance, and domain-specific tuning will sustain the value of well-engineered software.

Conclusion

The debate about whether AI will "eat" software reflects both rational capital allocation and oversimplified thinking. Markets have rewarded a concentrated set of infrastructure leaders while punishing many application names for perceived vulnerability. However, the path from model to reliable, monetized application is long and nuanced. AI in recruiting is a concrete example where added intelligence will transform workflows, but existing platforms that offer governance, integration, and continuous improvement will remain essential. Investors and buyers should distinguish between short-term hype and durable product-led value when evaluating companies in the AI era.