Hiring AEM talent? Request profiles →

Java 25 and the Deprecated-API Deadline on AEM as a Cloud Service: What Adobe’s Notes Say and Where They Disagree

Two dated items in Adobe’s current AEM as a Cloud Service release notes can’t be postponed: the Java 25 runtime timeline (optional enablement from November 2026, lower environments from February to May 2027, all runtimes in June 2027) and the enforcement of deprecated Java API removals, where the cutoff for critical updates has already passed. This post summarizes the dates and the 14 packages Adobe lists, shows how Adobe says to find usages with the AEM Analyser Maven Plugin, and points out where Adobe’s own pages disagree, so you know what to confirm for your environments.

Java 25 and the Deprecated-API Deadline on AEM as a Cloud Service: What Adobe’s Notes Say and Where They Disagree Read More »

Moving AEM Code to the Cloud with AI: The Cloud Migration Skill, the Cloud Migration MCP, and the Best Practices Analyzer

Moving to AEM as a Cloud Service means reworking Java code, OSGi configurations, Dispatcher setup, and legacy UI, not just content. Adobe documents an agent-based toolset for this: the Best Practices Analyzer’s findings, a hosted Cloud Migration MCP that pulls them from Cloud Acceleration Manager, and a Migration Skill that writes a read-only runbook first, then works one pattern per session in batches of five with a review pause after each. This post covers how the pieces fit, what patterns are named, and what Adobe’s pages leave out: maturity labels, accuracy guarantees, rollback guidance, and MCP data handling.

Moving AEM Code to the Cloud with AI: The Cloud Migration Skill, the Cloud Migration MCP, and the Best Practices Analyzer Read More »

AI-Assisted AEM Development on Your Laptop: AGENTS.md, Adobe’s Agent Skills, and the Quickstart and Dispatcher MCP Servers

Adobe documents local tooling that lets an AI coding assistant see your AEM project’s real state instead of guessing: a project context file, a set of agent skills, an AEM Quickstart MCP server for logs and OSGi diagnostics, and a Dispatcher MCP server for validating configuration against a local Docker Dispatcher. This post summarizes what each piece does, its documented prerequisites and limits (Java stack only), and Adobe’s security notes. It also flags what the page doesn’t say, including any GA or Beta label.

AI-Assisted AEM Development on Your Laptop: AGENTS.md, Adobe’s Agent Skills, and the Quickstart and Dispatcher MCP Servers Read More »

The AEM MCP Server: What AI Assistants Can Do in Your AEM, What They Can’t, and What to Lock Down First

Adobe documents an AEM MCP Server that lets AI assistants and coding agents work with AEM as a Cloud Service through one endpoint: pages, content fragments, assets, brand-rule checks, and Cloud Manager pipelines. This post summarizes what the documentation says it exposes, which capabilities carry prerequisites or licensing, how permissions work (tools run with the signed-in user’s own access), and Adobe’s safety guidance. It also covers what to decide before connecting a client, and what we couldn’t verify, including the product’s maturity label, which Adobe’s page doesn’t state.

The AEM MCP Server: What AI Assistants Can Do in Your AEM, What They Can’t, and What to Lock Down First Read More »

Protecting an AEM Edge Delivery Services Site: Site Tokens, CDN Authentication, and the OIDC Beta

Edge Delivery Services sites are public by default, which is wrong for a staging site, a partner portal, or an intranet. This post walks through the access-control options Adobe documents: token-based site authentication, CDN-level Basic authentication (GA, but not for live production domains), and OIDC sign-in (currently Beta). It also covers the limits Adobe states, like authentication only (no authorization) and whole-site-only protection, plus the documented fix when the Universal Editor starts returning 401.

Protecting an AEM Edge Delivery Services Site: Site Tokens, CDN Authentication, and the OIDC Beta Read More »

AEM + AI Data Privacy: Catching PII and Consent Violations Before a Page Goes Live

Most AEM privacy incidents aren’t sophisticated breaches they’re a testimonial copy-pasted straight from a support ticket with a real customer’s name and email still in it, published because nobody built a check assuming it needed one. This post shows how pre-publish PII scanning (both pattern-based and contextual), component-level consent tagging gated against live CMP state, and automated privacy-policy alignment checks catch these gaps before publish with every flag routed to a human privacy owner, never resolved automatically.

AEM + AI Data Privacy: Catching PII and Consent Violations Before a Page Goes Live Read More »

AEM + AI Security: Catching Vulnerabilities Before Release, Not After an Incident

Most AEM security incidents aren’t sophisticated zero-days they’re known CVEs that sat in a dependency tree for months, or an injection pattern in custom code that a rushed review missed. This post shows how continuous dependency/CVE scanning on every build, AI-reviewed pull requests for AEM-specific risk patterns (advisory only, human-confirmed), and production anomaly detection close the gap between quarterly audits, without ever auto-remediating a live instance without a security engineer’s sign-off.

AEM + AI Security: Catching Vulnerabilities Before Release, Not After an Incident Read More »

AEM + AI Commerce: Scaling Product Content and Recommendations Across Tens of Thousands of SKUs

A catalog with 40,000 SKUs and a copywriting team of three means most products ship with a manufacturer’s spec sheet as the entire product page. This post shows how Claude API drafts product copy strictly from supplied attributes (no invented specs), Adobe Sensei delivers session-aware recommendations instead of static rules, and AEM keeps AI-generated content and live Adobe Commerce pricing/inventory on separate refresh paths that converge cleanly at render time closing the long-tail content gap without adding headcount.

AEM + AI Commerce: Scaling Product Content and Recommendations Across Tens of Thousands of SKUs Read More »

AEM + AI Migration: How to Move Thousands of Legacy Pages Into AEM Without a Multi-Year Project

Most enterprise CMS migrations don’t fail on anything conceptually hard they blow their timeline on the sheer volume of repetitive, page-by-page mapping decisions a team makes by hand. This post shows how AI-based structural fingerprinting classifies legacy content by its actual markup pattern, Claude extracts fields into the target Content Fragment Model (returning null instead of guessing on anything uncertain), and 100%-coverage automated QA catches broken links and missing content before cutover not after launch, from user reports.

AEM + AI Migration: How to Move Thousands of Legacy Pages Into AEM Without a Multi-Year Project Read More »

AEM + AI Headless: Designing GraphQL Content Models That Serve Apps, Agents, and Everything Else From One Source

GraphQL on AEM was supposed to end the one-schema-per-channel problem but a sloppy Content Fragment Model just moves the pain downstream. This post covers how to design self-describing, machine-readable schemas, run them through an AI readability review before they ship, and use persisted GraphQL queries as one versioned contract that serves web, mobile, kiosks, and AI agents alike because an agent calling your API unsupervised won’t ask what an ambiguous field means, it’ll just guess.

AEM + AI Headless: Designing GraphQL Content Models That Serve Apps, Agents, and Everything Else From One Source Read More »