Skip to main content
AmitVishwakarma
Community Advisor
Community Advisor
September 20, 2026

AEM Sites + AI: Are We Automating Migration—or Moving Technical Debt Faster?

  • September 20, 2026
  • 1 reply
  • 39 views

AEM Sites modernization is moving beyond manually rebuilding every page and component.

With AI-assisted modernization, teams can potentially accelerate activities such as:

  • Migrating content from legacy websites or CMS platforms
  • Mapping page structures to reusable blocks
  • Applying design-system patterns
  • Creating new Edge Delivery Services pages
  • Making style and template updates
  • Previewing changes before deployment

But migration speed alone does not define a successful implementation.

A site can be migrated quickly and still create long-term challenges if:

  • The content model does not match the authoring workflow
  • Page structures are converted into non-reusable blocks
  • SEO metadata, URLs or redirects are missed
  • Accessibility and responsive behavior are not properly validated
  • Integrations, personalization, search or commerce logic are overlooked
  • The final site looks visually correct but is difficult to maintain

The current AEM Experience Modernization Agent approach is interesting because it combines natural-language interaction, live preview and GitHub-based review workflows, while developers retain control over what is finally deployed.

This raises some practical questions for the community:  

  1. Would you trust an AI agent to analyze a legacy site and recommend a target content model?
  2. Which part of an AEM Sites migration can AI genuinely accelerate?
  3. Where should human review remain mandatory—content modeling, block design, SEO, accessibility, integrations or final deployment?
  4. How would you measure the success of an AI-assisted migration?
    • Migration timeline
    • Author productivity
    • Core Web Vitals
    • Accessibility score
    • SEO performance
    • Defect rate
    • Long-term maintainability
  5. Would you use the agent directly, or would you prefer an architect/partner-led implementation with AI assistance?
  6. For an existing AEM project, would you consider moving to Edge Delivery Services if AI could reduce the migration effort?

Please share your own views, project experience and lessons learned. It would be helpful to know whether your project was a greenfield implementation, a legacy AEM migration or a migration from another CMS.

I am especially interested in practical feedback about what AI actually automated successfully—and what still required significant manual effort.

1 reply

KishorThummar
Level 2
September 22, 2026

​@AmitVishwakarma, This is an interesting discussion. I think the biggest risk is not necessarily AI moving technical debt faster, but AI allowing teams to scale an incorrect target model very quickly.

Adobe's current Experience Modernization Agent workflow is interesting in this regard. It can catalog the existing site, analyze page structure and authoring needs, map content to reusable blocks, generate migration infrastructure, and provide preview/critique workflows. It also follows Content-Driven Development, where the authoring model is considered before implementation.

For me, that makes the content model and block strategy the most important human checkpoint. Visual similarity alone doesn't necessarily mean a successful migration - a page can look almost identical while introducing poor authoring patterns, duplicated blocks, weak reuse, or missed business/integration requirements.

I would therefore think about the process as:

Discover → Model → Migrate → Validate → Scale

with human architectural review concentrated around content modeling, reusable block design, SEO/URL strategy, integrations, accessibility, and business rules, while AI handles the repetitive extraction, mapping, transformation, and iteration.

I also like that Adobe's current console keeps developers in control through GitHub review and approval before changes are deployed.

One question I'd be interested in hearing from teams actually using it:

What validation gate are you using before scaling an AI-generated migration from a few pilot pages to the full site?