When McKinsey introduced the term post-SaaS this year, they captured a quiet revolution already underway in software. The SaaS era centered on humans using cloud apps. The next phase centers on AI agents using software on our behalf — interacting with APIs, automating decisions, and redefining value creation.
From Users to Agents
McKinsey predicts that the number of human software users will plateau, while the number of autonomous AI users will grow exponentially. These agentic systems will log in, query, analyze, and transact just as employees do today — only faster and at scale.
At Pynest, we already see this shift in how enterprise clients deploy our automation frameworks. A single AI agent can now replace multiple routine workflows — from CI/CD triggers to customer onboarding sequences — previously requiring human coordination.
“We’re moving from software that serves people to software that collaborates with software. The interface is no longer the dashboard — it’s the protocol.”
— Roman Rylko, CTO at Pynest
Three Architectures of Post-SaaS
McKinsey’s framework identifies three archetypes shaping this transition:
- Agents as Users – augmenting human workers by operating within existing SaaS environments.
- Agent-Centric Architecture – where a single front-end agent orchestrates tasks across multiple systems.
- Agents as Experts – trained on domain-specific data to execute specialized reasoning, such as legal or medical interpretation.
These models are not hypothetical. Startups like Anysphere (Cursor) and Gamma are already proving that lean teams can reach massive ARR by building AI-native architectures from day one. In traditional organizations, this same architecture is emerging within DevOps, support, and analytics teams.
“The agent layer is becoming the new middleware,” says Fei-Fei Li, Co-Director of Stanford’s Human-Centered AI Institute. “It connects tools, decisions, and outcomes — not through code integration, but through reasoning.”
The End of Per-Seat Pricing
Perhaps the most disruptive consequence of the post-SaaS era is economic. The familiar per-user or per-seat subscription model no longer maps to a world where agents — not humans — perform most actions.
Instead, usage-based and outcome-based models are emerging.
In practical terms, this means billing by data processed, tasks completed, or measurable business results rather than headcount.
At Pynest, for example, some of our clients have transitioned to event-driven billing — they pay per automated workflow execution rather than per logged-in engineer. This approach aligns software value directly with operational output.
As Gartner noted in its CIO Agenda 2025, more than 45% of B2B software providers expect to adopt mixed monetization models within three years. For customers, that means unpredictable invoices may become predictable again — but only if usage transparency and cost explainability mature alongside it.
Why the Shift Is Accelerating Now
Three forces make this transformation unavoidable:
- Economics of automation – As gen-AI productivity gains plateau, enterprises turn to agentic AI for deeper process automation and reduced headcount.
- Data centralization – Modern data architectures (like Lakehouse and VectorDB) allow agents to reason over unified datasets instead of siloed applications.
- Trust infrastructure – With auditability frameworks such as AI provenance chains and model governance APIs, companies can now deploy agents securely within regulated environments.
The result is what McKinsey calls “AI-centric software”: systems built around continuous learning loops, explainable decision paths, and measurable business outcomes.
Implications for Customers
For end users, this shift brings both empowerment and opacity.
AI agents will deliver outcomes faster — but understanding how those outcomes were generated will require visibility into model reasoning, not just software logs.
Customers should expect:
- Dynamic pricing models that reflect computation or decision complexity.
- Autonomous orchestration — where apps negotiate data exchange via APIs without user input.
- Service-level guarantees for AI reasoning — explaining decisions becomes part of the SLA.
- Continuous delivery of intelligence — updates will target agents’ reasoning capabilities rather than app features.
“AI won’t replace people, but people using AI will replace people not using AI.”
— Andrew Ng
In the post-SaaS world, that logic extends to software itself: tools that don’t interoperate with agents will be replaced by those that do.
Risks: Vendor Lock-In and Data Entanglement
With great autonomy comes great dependency.
When AI agents learn from proprietary customer data, the line between vendor service and business intelligence blurs. Enterprises risk data entanglement — where leaving a platform means losing the trained behavior of their agents.
“The new lock-in won’t be your data — it will be your agent’s memory. Whoever controls how that memory evolves controls the customer relationship.”
— Roman Rylko, Pynest
To mitigate this, industry leaders advocate for open agent standards similar to what REST and OAuth achieved for web interoperability. Initiatives such as OpenAI’s A2A protocol and Anthropic’s Claude Context Sharing hint at this direction.
Navigating New Licensing Models
CIOs and procurement teams will need to rethink software evaluation frameworks:
- Audit for explainability – Vendors should provide traceable logs of agent decisions.
- Evaluate total cost of outcomes – Measure not the license cost, but the efficiency gain.
- Demand interoperability clauses – Ensure agents can export knowledge or connect to external orchestrators.
- Plan for adaptive budgeting – With variable usage, financial planning becomes continuous rather than annual.
According to a 2025 survey by IDC, 58% of enterprise buyers say pricing complexity is their top barrier to adopting AI-centric platforms. Vendors that simplify billing transparency — showing exactly which actions drive cost — will earn long-term trust.
Case in Point: AI Orchestration at Pynest
At Pynest, we piloted agentic delivery within our internal DevOps systems.
Previously, human engineers triaged incidents, coordinated deployments, and tracked metrics. Our AI orchestration layer now automatically classifies incidents, deploys corrective scripts, and updates dashboards — all without direct human input.
Over six months, this automation:
- Reduced average incident resolution time by 37%.
- Cut repetitive manual interventions by over 60%.
- Enabled engineers to focus on architecture rather than firefighting.
For clients, we’ve extended similar principles to customer support automation, where AI agents integrate directly into ticketing and CRM systems to resolve cases autonomously.
This agent-to-agent interaction represents the post-SaaS delivery model in action — not a dashboard for a human operator, but a self-improving workflow mesh between intelligent systems.
What the Next Five Years Look Like
McKinsey projects that AI-centric software could account for $600–800 billion in global value creation by 2030.
We’re entering a period where the differentiation between “software product” and “AI service” fades.
Expect three major trends:
- Consolidation around AI platforms – just as Salesforce unified CRM, a few ecosystems will dominate agent orchestration.
- Data-driven defensibility – access to proprietary datasets will outweigh UI/UX as a differentiator.
- AI-native ecosystems – marketplaces where agents trade capabilities and data securely.
In this environment, success won’t come from the breadth of features but from the depth of reasoning — how well a product’s agents can understand business context and act autonomously.
The Strategic Imperative
Becoming post-SaaS isn’t about adding AI features — it’s about rethinking the entire business architecture around autonomy and intelligence.
Software companies that embrace this shift will need to evolve across four dimensions:
- Product: Embed reasoning agents at the core, not as add-ons.
- Pricing: Align monetization with measurable outcomes.
- Operations: Automate internal workflows with the same intelligence offered to clients.
- Governance: Establish ethical and transparent oversight for agentic behavior.
The post-SaaS landscape will reward adaptability over scale.
“In the last decade, SaaS democratized software. In the next one, AI will democratize capability. The companies that survive will be those that learn to sell not subscriptions — but intelligence itself.”
— Roman Rylko, Pynest
Top comments (1)
Hi Roman,
How is it going?
I am Roman too. I am a software engineer.
I am very interested your post.
I want work with you.
Kind regards.
Roman.