AI Will Rewrite the Rules of Software
The impact of artificial intelligence on the software industry brings to mind an analogy from 20 years ago when software-as-a-service came along to challenge the on-premises applications that dominated enterprise computing. The big players in enterprise resource planning, CRM, and human resources management initially scoffed at the concept. Who would want to rent software instead of own it and store sensitive data on external, third-party servers rather than secure servers? Why would customers give up the power to deeply customize their solutions to match unique business workflows?
It turned out the benefits of multi-tenant platforms, continuous releases, subscription economics and self-service onboarding were well worth the platform shift. Customers were happy to exchange the year-long installation and testing grind for the instant productivity software-as-a-service applications delivered. IT organizations were delighted to make patches and bug fixes someone else's problem.
SaaS rapidly vanquished some of the dominant on-premises vendors. Those who survived underwent years of pain re-architecting their applications and licensing models. Meanwhile, cloud-native upstarts like Salesforce, ServiceNow, and Snowflake thrived.
Legacy application vendors have spent the past two years adding chatbots and copilots to paper over their inflexible architectures and outdated data models. But such superficial changes are the high-tech equivalent of putting lipstick on a pig. AI-native applications are built on intelligent automation, shared context, unified data models, and autonomous workflows. Answering chat questions is only a small part of the package.
Meanwhile, SaaS sprawl has created an unwieldy mess in many organizations. Software has become easy to buy, so companies have bought a lot of it. One company where I previously led sales, marketing, and customer success assembled a collection of 22 go-to-market products costing about $3 million a year in subscription fees. Sellers used roughly a dozen applications in a typical day, often entering the same information multiple times. Each tool addressed a legitimate need, but together they created redundant processes, duplicate data, overlapping workflows, and the need for constant training. Costly salespeople became data entry clerks and only spent 28 percent of their time with customers and prospects. An entire industry emerged to connect incompatible applications to each other and created a shadow services economy where software spend is trumped by an average of 3:1 ratio of services spend to cobble together that brittle aging infrastructure.
Data fragmentation and context switching kill productivity. That's where AI-native applications shine. They can capture unstructured information from emails, calls, meetings, messages, documents, web research, signals, and AI enrichment. A genuinely AI-native platform uses a unified architecture for structured and unstructured data, a context layer that understands the relationships among them, and an orchestration layer that can turn insight into action. Role-based access controls, data policies, and auditability are built in, not tacked on.
This creates a huge opening for startups. While incumbents possess formidable advantages in distribution, customer relationships, and capital, they labor under fragmented product portfolios, per-module pricing, per-seat pricing, complex legacy integrations, and hard-coded processes. A startup designing from AI-first principles has no such burden and moves at the pace of innovation in an AI world.
Go-to-market software built on a unified data and context layer allows account research, contact enrichment, outreach, coaching, forecasting, auto-updates, prep, follow-up, and agentic workflows to operate with humans and agents harmoniously, doubling productivity. In our experience, customers migrating from legacy GTM stacks can eliminate about 60 percent of their existing workflows because the AI architecture performs that work natively. Salespeople recover hours each week to spend selling instead of laboring across dozens of outdated products. Administrators get their weekends back while building custom agents to assist with specific business workflows.
AI-native applications ultimately make legacy CRMs simply a structured data source within a broader revenue operating system. Humans don't need to spend hours updating fields, generating quotes, preparing deal reviews, or reconciling forecasts because software can derive those outputs from actual conversations and signals where real selling interactions live.
Purchasing decisions will be simpler, too. SaaS transformed software buying by letting customers evaluate and adopt products without speaking to a salesperson. Agents will soon evaluate products, complete onboarding, connect systems, and execute routine transactions autonomously with other agents. High-stakes enterprise purchases will still demand a human touch, but lower-friction decisions will increasingly happen automatically.
Legacy SaaS companies face difficult choices. They can layer a little AI on top of their decade-old stack and report progress. But that incremental approach might also create agentic thrash, with multiple agents stepping on one another because none was built into the core codebase. Those that re-architect around AI-native constructs will have a long-term advantage, but the next couple of years could be full of pain.
The winners in the SaaS revolution weren't the companies that treated the internet as another delivery channel. They understood that a new architecture required a new business model and a different relationship with customers. The same principle applies to AI.
An AI-native platform would make permissions, provenance, and data boundaries part of the architecture. The system would know where information originated, what the customer had authorized, and whether the proposed use was consistent with that permission. In an AI-era, customers shouldn't expect to pay for access to their data or give it away for general training; they should be guided into how their data and workload patterns can be augmented and automated with AI for the benefit of unlocking positive business outcomes. That's the value that the unlock of application layer platforms built on an AI-native architecture.
Much has been made of the SaaSpocalypse that hit software stocks in the early days of generative AI. In the long run, that sell-off might look like only the first stage of a larger transformation. AI-native software will make users wonder why they ever tolerated systems that forced them to conform to the vendor's workflow instead of adapting to the way the business actually works and pay for seat-based access as opposed to value-added outcomes.
Jason Eubanks is CEO of Aurasell.