Why SaaS Is Still Broken — and Why the Age of AI Must Fix It

For decades, software-as-a-service (SaaS) products have revolutionized how businesses operate. Giants like Salesforce, HubSpot, Workday, and ServiceNow have transformed workflows, enabled remote work, streamlined operations, and created billions in value. Yet despite all this progress, SaaS remains frustratingly difficult to use for most people — especially when compared to the simple conversational experiences many now enjoy with tools like ChatGPT or Claude.

The difficulty isn’t that SaaS doesn’t work. Rather, the day-to-day experience of using these tools is still “terrible” — not in terms of functionality, but in terms of usability. Too often, SaaS feels like a maze of menus, configuration screens, and workflow builders that require specialist training, certification programs, or consultants to navigate. This complexity has become normalized in business — but that doesn’t mean it should be accepted as the end state.

A typical enterprise might run 130+ SaaS applications, each with its own mental model, navigation paradigm, settings buried deep within interfaces, and unique configuration quirks. Instead of saving time, these tools impose a tax on internal teams: hiring expensive admins or consultants to implement, integrate, and maintain systems; draining precious hours from talented employees; and creating knowledge silos when a specialist leaves the company.

The cumulative cost — in time, money, and opportunity — is staggering. A seemingly straightforward purchase becomes a major operational project that requires months of implementation and months more of training and support. And when the goal was simply to get outcomes (like automatic lead follow-ups or streamlined customer support), what businesses actually get is interfaces.

When AI entered the picture, many hoped it would be the long-awaited solution to SaaS complexity. After all, large language models can now understand natural language intent and generate responses that feel effortless. But what many vendors shipped instead are AI assistants bolted onto existing complex UIs — essentially making complex systems slightly faster to navigate, but not eliminating the underlying complexity.

Today, SaaS vendors such as Salesforce, HubSpot, Zendesk, and ServiceNow offer AI agents or copilots. These are genuinely powerful — they can reason, execute tasks, and automate workflows. But each agent is tied to its parent application, has its own configuration needs, and often doesn’t communicate with other agents. The result? Users now have 130+ tools plus 130+ AI agents to manage, doubling the cognitive and operational burden rather than reducing it.

This situation highlights a central paradox of modern SaaS: AI is being used to optimize interactions with software interfaces that shouldn’t exist in the first place. It’s like adding a GPS to a horse-drawn carriage — a technological improvement, yes, but missing the bigger opportunity to build a truly new way of interacting.

The real promise of AI in business software isn’t about speeding up clicks. It’s about eliminating the clicks altogether — enabling users to speak outcomes into existence. Imagine telling your system in plain language: “Set up a follow-up sequence for prospects who missed their demo,” or “Show me deals at risk this quarter with reasons.” And then having the system automatically configure the workflows, notifications, and dashboards that deliver that outcome — without any menus or builders.

In this vision, the configuration layer becomes a conversation, and institutional knowledge becomes accessible through natural language instead of hidden in labyrinthine settings. Enterprise automation no longer requires a team of specialists; it requires clear goals and concise instructions.

Despite current limitations, the technology already exists to support this future. Large language models can understand intent, translate requests into actions, and refine workflows based on feedback. What’s missing are robust bi-directional integrations and guardrails that allow AI to act in systems securely and reliably, not just suggest actions. Those who build this layer — where AI can truly operate across platforms — stand to redefine how SaaS is used.

Traditional SaaS interfaces won’t disappear overnight. Vendors are understandably reluctant to cannibalize their existing products or training ecosystems. But the industry is now in a transition state — one where AI can finally deliver on the original promise of business software: outcomes without barriers.

For founders and product leaders, the challenge is not incremental AI improvements. It’s rethinking products for the age of conversational intent. Instead of building yet another set of configuration screens, ask what your tool would look like if users could simply talk to it and have it do what they want. For buyers and operators, shift evaluation toward tools that deliver outcomes through simplicity and natural interaction, not complexity and menus. And for the broader SaaS ecosystem, this moment represents a pivotal shift: software that just does what you ask.