Executive Summary and Context
Artificial intelligence (AI) continues to be a central topic for investors in the application software sector. The past 12 months have brought a wave of speculation around AI-driven disruption, particularly with the emergence of large language models (LLMs) and AI agents. While some commentators predict existential threats to incumbent software vendors, we believe this outlook is overstated. Our research suggests that while AI will influence enterprise software, the disruption will be incremental and manageable for companies with deep, multidimensional moats.
This report outlines the four core arguments of the AI bear case and offers our perspective on why entrenched vendors like Workday remain well positioned. Meanwhile, software vendors that rely more heavily on user interface familiarity and are less embedded in enterprise workflows, face greater near-term risk.
Key themes examined include:
1. AI agents and seat-based pricing pressure
2. AI-native coding tools and barriers to entry
3. Application design for human users vs. AI workflows
4. AI disintermediation of business logic and data access
Each of these points introduces specific threats, but they also highlight the resilience of enterprise software models and the pragmatic pace of AI adoption.
Evaluating The First Two AI Bear Arguments
AI Agents Replacing White-Collar Workers
The argument that AI agents will replace large numbers of white-collar workers is currently unsubstantiated by employment data. While LLMs are enhancing productivity in roles like software development, customer service, and recruiting, the evidence points to augmentation rather than replacement. Software developers, for instance, remain in demand, albeit with a shift in job composition1.
Workday’s Recruiter Agent exemplifies AI’s role in freeing up time (2.5 hours/day) rather than eliminating jobs. Even with these efficiency gains, full automation is constrained by tasks AI cannot yet perform, including compliance, mentorship, and decision-making requiring judgment.
Importantly, application software vendors are protected by multi-year contracts and are already adapting seat-based pricing models. For example:
• Workday, and other large software vendors, integrates CPI-based and innovation-linked pricing escalators.
• Usage-based revenue models are being tested, offering flexibility if seat counts decline.
AI-Native Coding Tools and Competitive Entry
The idea that LLM-based development tools like GitHub Copilot or Claude Code will reduce barriers to entry in application software overlooks the broader picture. While these tools cut coding time by up to 50%, actual coding represents only ~5–8% of the Opex required to build an enterprise SaaS platform. In other words, cost savings from faster coding are marginal in the context of building a large-scale software business.
Moreover, incumbents also leverage these tools to accelerate their own innovation. Workday’s expanding portfolio of AI-driven agents demonstrate the ability of established firms to stay ahead.
Crucially, startups face headwinds including:
• Lack of historical customer data, which limits model fine-tuning.
• Disjointed go-to-market channels.
• Preference among CIOs for integrated suites due to compliance, cost, and complexity considerations.
As a result, AI-native startups are often complementary rather than directly competitive with system-of-record vendors like Workday
Structural Resilience of Enterprise Software
Application Design and Human-Centric Interfaces
The third AI bear argument, that enterprise applications will become obsolete as they were built for humans, assumes an unrealistic pace of automation. Most application software is structured in three layers: user interface, business logic, and data schema. While user interfaces may evolve with AI prompting capabilities, replacing the underlying business logic and data models is a non-trivial task.
User interface familiarity also provides friction to switching, especially for power users. For example, graphic design products typically enjoy strong user lock-in, but they are more vulnerable to interface disruption than vendors like Workday, whose moat relies more heavily on complex business logic, regulatory, and compliance functionality.
More importantly, business logic and data schemas have been designed to serve human workflows. Replacing these workflows with AI agents would require significant redesign, which is costly, risky, and often unnecessary. Workday’s single-codebase and clean data architecture allow it to iterate quickly in response to new demands, offering a structural advantage over more fragmented competitors.
Data Access and the Role of AI Agents
The fourth and most speculative bear case contends that AI agents will eventually replicate the business logic encoded in software, leaving software vendors as mere data repositories. While conceptually appealing, this thesis overlooks several real-world constraints:
• Immaturity of AI agents: Current systems struggle with multistep workflows and error accumulation.
• Data complexity: Each vendor’s data schema is unique. AI agents would need to replicate proprietary logic just to interpret data meaningfully.
• Control over access: Incumbent vendors can and do restrict API access to protect their IP and client data. Glean, for example, has already faced such throttling by multiple software vendors.
• Regulatory and audit requirements: Mission-critical workflows (e.g., SOX compliance) demand deterministic outcomes and robust access controls.
For now, enterprises will remain cautious about replacing trusted applications with unpredictable AI systems. The “blast radius” of errors in multi-agent systems accessing sensitive data is too high a risk for most firms to accept, especially in mission-critical back office functions like HR and Finance.
Implications and Outlook
While generative AI is already reshaping aspects of software usage and development, claims of wholesale disruption are premature. We believe the more likely outcome is one of augmentation, not replacement. Enterprises will adopt AI tools to boost productivity and enhance workflows, but they will do so with measured caution, especially in regulated or data-sensitive domains.
As such, we view Workday as particularly well positioned:
• Strong product suite across HR and Finance
• High gross retention (~98%)
• Single data model and fast innovation cycles
• Deep business logic integration and enterprise trust
The emergence of LLMs and AI agents introduces new dynamics to enterprise software markets. However, the structural advantages held by incumbent vendors including contractual lock-in, data control, embedded business logic, and regulatory compliance offer significant protection against disruption. We believe firms investing in AI proactively, while maintaining trust and security, will likely emerge stronger, not weaker, in the AI era.
Hotchkis & Wiley Research