The Software Industry After Agents: From “Buying Apps” to Owning Workflows

The Software Industry After Agents: From “Buying Apps” to Owning Workflows

Software is about to go through a transition that looks less like “cloud eating the world” and more like electricity commoditizing the machine. Not because demand for software disappears—but because the unit of value shifts.

A recent (intentionally extreme) scenario piece by CitriniResearch imagines what happens if agentic AI keeps compounding: coding becomes cheap, feature velocity explodes, differentiation collapses, and the modern SaaS stack turns into a knife fight on price—while the broader economy struggles to absorb displacement. It’s a scenario, not a forecast. But it’s useful because it forces the uncomfortable question:

If building software becomes dramatically easier, what exactly are we paying for when we pay for software?

Let’s talk about what’s likely to happen to the software industry—even if the macro doomsday parts never show up.

---

1) SaaS won’t die. But “renting interfaces” will.

The old world: a company buys software because building it is slow, expensive, risky, and requires specialized teams.

The new world: a competent team with strong product taste can stand up “good enough” internal software quickly using agentic coding tools—often in weeks, not quarters. CitriniResearch frames this as the moment procurement stops treating renewals like inevitabilities and starts treating them like negotiations with a credible threat: “we can build this.”

That doesn’t mean every company becomes a software company. It means every company becomes software-capable enough to create leverage.

Implication: SaaS vendors face chronic pricing pressure, higher churn risk, and a collapse of “long-tail” defensibility. Renewals become competitive bids against internal agent-powered builds.

---

2) Feature advantage collapses; distribution and data become the moat

When shipping gets cheap, everyone ships. Incumbents used to outpace challengers via bigger teams. Now challengers can reach parity fast, and incumbents can’t rely on slow-moving roadmaps as protection.

So what holds?

• Proprietary data loops (not just “we have data,” but data that improves outcomes)
• Embedded distribution (default in the workflow, not a tab you can replace)
• Trust + compliance + reliability (auditability, governance, uptime, incident response)
• Switching costs that are real (process + integration + risk), not artificial UI familiarity

The core shift is brutal: features become table stakes, not differentiation.

---

3) Seat-based pricing gets punished (especially in systems of record)

When AI drives headcount down, seat-based revenues mechanically shrink. Even if the vendor is “mission critical,” fewer humans means fewer seats.

This shows up first where pricing is tied to seats, per-user tiers, and other “humans in the loop” assumptions.

Expect pricing models to move toward:

• outcome-based pricing (time saved, tickets resolved, dollars recovered)
• usage-based pricing (API calls, workflow runs, automations executed)
• risk-based pricing (compliance coverage, guarantees, SLAs)

If your product’s value is “help humans do X,” and there are fewer humans doing X, you’re selling into a shrinking denominator.

---

4) “Build vs buy” becomes “compose vs own”

Enterprises won’t just switch from buying SaaS to building everything. They’ll adopt a third posture:

• Compose: assemble workflows from components (LLMs + tools + internal systems) quickly
• Own the workflow: keep the orchestration layer in-house
• Buy commodities: pay vendors for infrastructure, compliance, and boring reliability

In that world, vendors who sell interfaces get squeezed. Vendors who sell capabilities that plug into orchestration win.

---

5) The “intermediation layer” in software gets wiped out

Agents dissolve businesses built on human friction—search costs, switching costs, inertia—because they don’t get tired, don’t get distracted, and don’t accept “good enough.”

In software, that maps to products whose main value is:

• gluing things together because integration is annoying
• dashboards that exist because humans need to check dashboards
• “workflow” tools that mostly manage human follow-ups

Agents don’t need many of these as products. They need them as APIs, policies, and execution surfaces.

---

6) Winners, losers, and the new map of software

Likely winners

• Platforms with deep data + trust (identity, security, payments infrastructure, compliance)
• Infrastructure that reduces operational burden (observability, governance, evals, audit trails)
• “Agent-ready” systems (clear APIs, deterministic actions, strong permissioning)
• Vendors who embrace customer-owned orchestration (you don’t own the workflow; you power it)

Likely losers (or forced to reinvent)

• Long-tail SaaS with thin differentiation
• Tools priced per seat with no outcome leverage
• Products whose value is primarily “UI convenience”
• Middle-layer workflow tools that are out-automated by agentic workflows

---

7) What SagentLab thinks you should do (if you build software)

If you’re a software company, your roadmap should shift from “more features” to “more leverage.”

A practical checklist:

1. Build for agentic usage

   • Strong APIs, webhooks, tool calling interfaces
   • Permissions, audit logs, deterministic operations
   • Clear semantics (agents need unambiguous primitives)

2. Price on value, not headcount

   • Align pricing to outcomes or usage that scales with benefit, not org size

3. Assume competitors can copy your UI

   • Invest in data loops, reliability, compliance posture, and distribution

4. Make your product composable

   • Customers will increasingly own orchestration; don’t fight it—be the best component

5. Treat “trust” as a product

   • In an agent world, governance and safety aren’t features; they’re the purchase decision

---

The bottom line

The software industry isn’t heading toward “less software.” It’s heading toward software that’s cheaper to create, faster to change, harder to monetize with old models, and increasingly owned (or at least orchestrated) by the buyer.

SaaS won’t vanish. But the SaaS era—where companies buy dozens of apps primarily because building is too hard—is fading. Agents turn software from a collection of products into a continuously reconfigurable system.

The companies that win will stop selling “apps” and start selling capabilities that survive in a world where the interface is optional and the workflow belongs to the customer.

Inspired by a scenario analysis from CitriniResearch (“The 2028 Global Intelligence Crisis”).