Current assumptions suggest generative AI will make SaaS solutions obsolete. This is already observable in the valuations of the SaaS leaders(e.g., Salesforce, $CRM). Feedback from senior software engineers indicates a return to custom software development assisted by AI codegen. When AI automates repetitive coding tasks, it allows developers to create custom apps that accurately reflect the business needs, which is in contrast with the generic SaaS solution, which has bloated significantly over the last decade.
Note: Analysis based on Aaron Levie’s insights from a recent a16z podcast episode. His original thoughts are provided at the end as a transcript summary at the end of the article.
The Problem: SaaS Constraints and Complexity
Enterprise SaaS systems frequently burden users with excessive features, inflexible workflows, and poor UX. Salesforce is the poster child for feature bloat and obsolete interfaces. Businesses must adapt critical processes to conform to SaaS ontologies and business logic, bringing unnecessary complexity and losing agility in competitive industries like insurance.
New Approach: Accelerating Custom Software Development with AI
Future enterprise software will likely use targeted, AI-enhanced custom applications that link specific datasets to defined business goals. This model combines human architects with AI code-generation capabilities:
Instead of devaluing the work and expertise of human software engineers, AI reshapes their focus toward high-impact, strategic tasks. Key competencies include:
AI can accelerate the delivery of strategic software projects by producing highly specialized software assets that have been fine-tuned for the needs of the business. Instead of training employees on the new SaaS system and its quirks, custom software delivers productivity directly to the end-users of the business system.
The SaaS industry’s reliance on rigid multi-year contracts may indicate vulnerability rather than resilience. CTOs will rely increasingly on AI-supported, custom-developed solutions, rather than SaaS. For highly regulated sectors such as insurance, precise alignment of domain expertise and tailored software may significantly enhance competitive capabilities.
Aaron Levie, co-founder and CEO of Box, offers a practical view on how enterprise software is evolving in the age of AI. He challenges the binary narrative between fully standardized SaaS and fully AI-generated, one-off tools. In his view, most users prefer custom software that works without bloat but also without configuration or prompting. They expect systems that present the relevant information and interactive elements without asking, while keeping traditional SaaS platforms relevant for non-differentiating functions like HR or ticketing.
He sees significant potential in the underdeveloped areas of enterprise software, where internal tools are often left undeveloped due to time and resource constraints. AI-enabled development platforms, such as Replit, are beginning to unlock this long tail of use cases by making custom tool creation faster and easier for IT teams.
Levie maintains confidence in vertical SaaS. The strength of these platforms lies less in their technology and more in their domain expertise. For example, a team of pharma industry veterans sitting with engineers to shape workflows delivers more value than a generic, prompt-based AI system attempting to solve the same problem.
He also pushes back against the idea that interfaces will disappear in favor of AI agents and API-only systems. People still want structured dashboards and predefined views, particularly in enterprise settings. Even if agents generate pages dynamically, they will likely recreate many familiar patterns.
In boardrooms and executive workflows, AI is already influencing strategic thinking. Levie notes that some companies use AI to surface discussion points during board meetings. He personally uses AI tools to analyze drafts of earnings scripts, identify gaps, and generate anticipated questions from analysts. This has improved both clarity and preparedness.
The impact of AI on developers is especially strong. Levie believes AI amplifies the performance of skilled developers by taking over repetitive tasks like boilerplate coding, syntax cleanup, and unit test generation. Developers now act more as reviewers and architects, applying their expertise to validate and guide AI-generated output. The relationship has shifted. Instead of AI correcting human mistakes, humans now correct minor AI inaccuracies, while output volume increases substantially.
On enterprise adoption, Levie explains that AI spending can be absorbed within typical budget dynamics. Instead of hiring ten additional engineers, a company might hire five and invest in AI licenses. These tools provide productivity gains that offset the reduction. Most organizations have enough flexibility within a one- to two-year planning cycle to adapt without major disruption.
Levie’s conclusion is clear: AI is increasing the volume and speed of custom software development. Teams with strong domain understanding and strategic focus will use it to deliver much more value.