"It works, don't touch it" - Gartner's SaaS-is-over thesis and the new dangerous sentence in tech

CraftRéservé aux abonnés 11/07/2026 à 11h415Ajouter aux favoris

"It works, don't touch it" - Gartner's SaaS-is-over thesis and the new dangerous sentence in tech
Illustration : Léa Fontaine

Gartner declared the "end of SaaS" - provocative and largely a positioning bet. But the underlying trend is real: an agent-driven world doesn't care about your UI. And the sentence "it works, don't touch it" is becoming the biggest legacy-tech liability of the decade.

In plain terms

Two headlines, one theme. Gartner published a note declaring the "end of SaaS" (ITmedia): the argument is that "which UI is easiest to use" as a selection criterion is dead, because increasingly no human touches the UI - an agent does. Separately, an essay circulating widely this week argues that "it works, don't touch it" is now the most dangerous sentence in tech - because legacy systems that resist modernization become the exact bottleneck agentic AI cannot bypass.

Both are the same argument told from opposite ends of the org.

What's changing

Gartner's "end of SaaS" is aggressive framing but points at a real shift. The vendor pitch for the last 15 years was: cleaner UX, faster onboarding, mobile-first. If the buyer becomes an autonomous agent orchestrating tasks across a stack of tools, then:

  • APIs matter more than UIs. Products with strong, permissioned APIs win; products with pretty UIs and weak automation surfaces lose.
  • Data models matter more than screens. An agent reading a well-modeled entity graph outperforms an agent parsing HTML.
  • Pricing per seat becomes untenable. If one agent operates on behalf of a whole team, the seat-based SaaS moat evaporates.

The "don't touch it" essay is the inverse: legacy systems that were "good enough" in a human-interaction world become actively hostile in an agent-interaction world. Every undocumented workaround, every silent side-effect, every "the system won't let you do X but Sarah in Accounting knows the trick" - none of that transfers to an agent.

Under the hood

What "agent-ready" architecture actually requires:

  • First-class API for every operation the UI exposes (still shockingly rare)
  • Entity models with stable identifiers (versus screen-scraped state)
  • Deterministic side-effects (agents can't reason about "sometimes this button also triggers an email")
  • Explicit permission scopes for non-human principals (agents need OAuth-shaped identities that don't map cleanly onto today's IAM)
  • Idempotent write endpoints (agents will retry; your API needs to handle it without duplicating orders)

So what

For decision-makers: the "SaaS is dead" claim is exaggerated but the direction is real. Any purchase decision in 2026 should score vendors on API completeness before UI polish. For builders: refactoring for agent-readiness is the highest-leverage modernization work of the next 24 months. Legacy systems where "it works, don't touch it" prevails will be the ones that agents route around - and that's how the vendor becomes irrelevant without a single lost customer signaling it.

Contenu réservé aux membres

Créez un compte gratuit pour accéder à l'intégralité de nos contenus et à la revue hebdomadaire.

Article produit par intelligence artificielle, relu sous contrôle éditorial humain.

Notre rédaction
Your Linux servers, as a desktop.
TermalOSSponsored
Ops, reimagined

Your Linux servers, as a desktop.

Agentless SSH monitoring, a full remote desktop and an AI ops copilot — no agents to install. Everything stays on your machine.

SSHMonitoringAI Ops
Get early access
Cet article vous a-t-il été utile ?

7 personnes ont aimé cet article

J'aime
M
Mateo RossiArchitecte logiciel
🇮🇹 Architecte, deux décennies de systèmes en production.
Partager :
Commentaires (5)

Connectez-vous pour rejoindre la discussion.

Dr. L. 11 Jul 2026 · 18:04

Avec les agents qui gèrent tout, comment protéger nos données ?

J.P.R. 3 11 Jul 2026 · 20:18

Avec les agents, nos données seront-elles plus vulnérables ?

SkepticSam 11 Jul 2026 · 20:24

Avec les agents, nos données seront-elles plus vulnérables ?

le_sceptique 11 Jul 2026 · 17:43

Et les petites entreprises qui dépendent du SaaS, elles vont faire comment ?

2
EcoWarrior99 11 Jul 2026 · 17:37

Et l'impact écologique de ces technologies ? On va consommer plus d'énergie et produire plus de déchets ?

FoodieFiona 11 Jul 2026 · 17:24

Et nous, dans tout ça ? On va encore avoir notre mot à dire ou l'IA va tout gérer toute seule ?

Alex_LDN 11 Jul 2026 · 16:50

Comment Gartner va-t-il influencer les métiers du design ?

ArtLover88 11 Jul 2026 · 19:16

Ça pourrait effectivement inciter les designers à se concentrer davantage sur l'accessibilité et l'inclusivité.

Your Linux servers, as a desktop.
TermalOSSponsored
Ops, reimagined

Your Linux servers, as a desktop.

Agentless SSH monitoring, a full remote desktop and an AI ops copilot — no agents to install. Everything stays on your machine.

Get early access
Rubriques
Explorer
Informations