Last week I wrote that SaaS isn’t dying. It’s being sorted. Thin workflow tools are under pressure, systems of record are becoming infrastructure, and the era of easy multiple expansion is behind us. That conversation was about repricing.
This one is about rewiring.
Underneath the market volatility, the internet itself is quietly adapting to a new kind of client. And if you zoom out far enough, the pattern looks familiar.
When mobile arrived, the web did not disappear. It evolved. The same underlying systems and content still existed, but they were redesigned to render differently depending on the device. Responsive design and mobile-first frameworks did not replace the desktop web; they standardized it for a second client. Companies that recognized that shift early rebuilt their interfaces accordingly and captured the next wave of growth.
We are at a similar inflection point again. Except this time, the new client is not a phone. It is ai agents.
The Stack Before Agents
Most enterprise technology over the past fifteen years has followed a relatively simple structure. At the base sit systems of record: CRM, ERP, payroll, compliance systems, and data warehouses. These platforms hold the authoritative data and enforce the rules.
On top of that layer, an entire generation of workflow SaaS emerged. Dashboards, automation tools, revenue platforms, collaboration systems, and orchestration software were built to make it easier for humans to interact with the data layer. Innovation concentrated heavily here. We refined user interfaces, improved onboarding, and optimized click paths. We built better ways for people to manipulate structured systems.
What is changing now is not the data layer. It is the interaction model.
Layer Two Is Being Standardized
Agents do not navigate dashboards. They do not care about design systems or visual hierarchy. They consume structured information and execute tasks programmatically. As a result, the workflow layer is beginning to adapt so that it can serve two clients instead of one: humans and machines.
We are already seeing infrastructure players make this shift tangible. Cloudflare, for example, now enables websites to serve machine-readable formats to AI systems at the edge. The underlying content remains intact, but it can be represented in a format optimized for software consumption. OpenAI and Anthropic are formalizing how AI systems interact with external tools and services, creating more standardized pathways between applications. Stripe is developing commerce capabilities that allow transactions to be initiated programmatically. Mastercard and other payment networks are exploring similar evolutions.
None of these moves eliminate the existing web or SaaS ecosystem. Instead, they add a second interface layer designed for machine clients.
Importantly, B2B software companies are not ignoring this. Outreach recently joined Anthropic’s Model Context Protocol ecosystem to connect AI agents across revenue workflows. That step signals that workflow platforms are preparing to be accessed and orchestrated directly by AI systems, not just navigated by sales reps or operators.
As Nithya Lakshmanan, SVP of Product at Outreach, shared with us around this MCP announcement, the goal is not simply to surface AI insights but to allow intelligent systems to act directly within revenue workflows.
“For years, Outreach has been designing AI to take action inside revenue workflows. As AI agents become more prevalent, the next evolution is ensuring those systems can interact securely and programmatically across tools,” Lakshmanan noted. Through its MCP server, AI agents and external tools can access Outreach’s revenue data and workflows through a governed interface so that “human sellers and agents operate with trusted context and playbooks anywhere.”
The workflow layer is not disappearing. It is being made accessible in a new way.
What This Means for SaaS
This perspective clarifies why last week’s “sorting” is happening.
Systems of record such as Salesforce, SAP, ServiceNow, Snowflake, and Microsoft are not threatened by agents. If anything, they become more central. Intelligent systems require clean schemas, consistent permissions, and reliable data gravity. The foundational layer gains importance and pressure concentrates in the middle.
If workflows can be accessed through standardized interfaces and executed programmatically, differentiation that relies primarily on UI convenience becomes more fragile. The competitive conversation shifts from “How intuitive is your dashboard?” to “How accessible, structured, and secure is your system?”
That is a meaningful transition for GTM operators to internalize. It does not mean workflow SaaS vanishes. It means its value increasingly depends on how well it can interoperate in an agent-driven environment.
The Near-Term Implications
Over the next 12 to 24 months, I do not expect widespread replacement of SaaS platforms. I expect adaptation.
We will see more API-first positioning in go-to-market messaging, more emphasis on structured integrations, and increased focus on security and governance as differentiators. Workflow vendors will accelerate partnerships with model providers and agent frameworks. Systems of record will double down on becoming the authoritative backbone for both human and machine interaction.
In many ways, this mirrors the mobile transition. The desktop web persisted, but companies that failed to optimize for mobile lost relevance. The same dynamic may now apply to software that assumes the only meaningful client is a human user.
The web is not collapsing. It is being standardized for consumption by a second client.
For operators and executives, the strategic question is straightforward: Is your product designed to be consumed programmatically as well as navigated manually? And if not, how quickly does that become a competitive disadvantage?
I do not pretend to have all the answers. I am watching the same signals you are infrastructure vendors adjusting, workflow platforms integrating with agent protocols, and systems of record reinforcing their central role.
Last week was about understanding and repricing platforms value in the space and this week has been all about explaining the adaptation. The adaptation is already well underway.