March 2025: Sitecore called itself AI-first, and the field spent the month writing retry loops

  • JD + AI
  • Sitecore
March 2025: Sitecore called itself AI-first, and the field spent the month writing retry loops

At SUGCON Europe 2025 Sitecore's message was that it had stopped being a composable DXP and become an AI-first platform, end to end. In the same four weeks, five people with no connection to each other published posts about rate limits, missing analytics exports and hand-filtered path arrays, which is what running an XM Cloud site actually consisted of.

That distance between the keynote and the ticket queue is the story of March 2025. It is not a complaint about the roadmap, it is a question about what you should have been budgeting for that quarter.

The record for March 2025 is partner-heavy and one link short

This is a backfilled month, reached through feed archives rather than harvested as it happened: 45 Sitecore items in the window, 13 cited here. Four of those thirteen are by Nick Allen at Think Fresh Digital, which describes itself as a specialist Sitecore delivery partner, and three of the four are conference write-ups. Almost everything below about SUGCON is therefore one delivery partner's account of a vendor event rather than an independent record of it.

A fourteenth item is named here and deliberately not linked. Shrikant Sahu published a walkthrough of custom definitions in Sitecore Content Hub on 25 March; our harvest captured only the blog root, and we do not hand a reader a link that fails to resolve to the article.

On stage: AI-first, end to end

AI-first by design: SUGCON keynote insights on the future of Sitecore and digital experience — Nick Allen, Think Fresh Digital (a Sitecore delivery partner) Allen's summary of CEO Dave O'Flanagan's opening leaves no room: "Sitecore is no longer simply a composable DXP, it's now an AI-first platform, end to end." The concrete developer artefact was the Marketplace SDK, a CLI-driven way to build field types and interface panels across XM Cloud, Content Hub and the portal, which Allen reports was open-sourced.

Inside Sitecore Stream: what it is, how it works, and what's coming next — Nick Allen, Think Fresh Digital The day-two detail, in which Stream is not a product but a set of AI services layered across the stack, with brand kits carrying tone and style. Allen flags the session as "forward-looking, subject to change" and is equally clear on the limits at the time: no API to manage Stream or brand kits, brand awareness withheld from the free tier, generation working at field level without knowing what else is on the page. Every version, tier and roadmap claim here comes from a partner's account of a vendor session, so check it against Sitecore's own documentation before it goes near a proposal.

Future of conversational UIs: key takeaways from SUGCON EU — Nick Allen, Think Fresh Digital Rob Coyle's session, arguing that a CMS will have to publish content for machine consumption and that ai.txt and llms.txt are the coming equivalent of robots.txt. The post then asks ChatGPT, Claude and Gemini to predict the next five years and prints all three answers in full, which we would have cut: the models are paraphrasing the article they were handed, and Allen half-admits it.

Off stage: the month was mostly about rate limits

Frontline insights: Building with Sitecore XM Cloud — Nick Allen, Think Fresh Digital The most operationally useful of the four, reporting Derk Hudepol's session against Allen's own delivery experience. Three things to act on: JSS does not follow semantic versioning strictly, so a caret in package.json will hand you breaking changes on a minor bump; Edge publishing is batched with no UI to watch the queue; and revalidating an ISR site after a global component changes means passing every path, which earns 429s from Edge and from whatever third-party services your pages call at build time. The workaround is Edge webhooks into Vercel API routes with retry and queueing behind them, which is a lot of machinery to keep a footer current.

Handling Rate Limits in Sitecore Send API: Retrying 429 Error in Add Multiple Subscribers — Chirag Goel, Coders Lifeline The same wall from a different direction. Goel puts the Add Multiple Subscribers endpoint at two requests per ten seconds, and reports the detail that makes it expensive: the API answers HTTP 200 with the 429 buried in the response body, so a client that trusts the status code records a successful import that quietly dropped subscribers. His C# retry loop reads the body code instead, which is both the right fix and an indictment of the contract.

How to Exclude a Page from Static Site Generation (SSG) in Sitecore Headless with Next.js — Debtanu Biswas, Debtanu's Blogs – Tech Tales Four hundred words, and the answer is to filter the paths array in [[...path]].tsx against a hardcoded string. It works, and it also buries an editorial decision in application code where no author can see it and no reviewer will remember it.

Sitecore Search: Best Practices, Benefits, and Limitations — Chirag Goel, Coders Lifeline An implementation checklist, and the value is in its closing section, where Goel names cost, learning curve and the customisation ceiling as trade-offs. That is more than most write-ups of a first-party product will say.

Measurement was the gap nobody on the keynote stage mentioned

Assessing Analytics in Sitecore XM Cloud — Cade Whitbourn, cadewhitbourn.com Whitbourn, writing as a digital strategist, calls the built-in analytics "a mixed bag of convenience and limitation" and then itemises it: visitors, sessions, bounce rate, time on site, fixed ranges from one hour to one month, no custom range. His verdict is "extremely limited analytical capabilities", with no export and no way to bring external data in. If a client signed for XM Cloud expecting reporting, this is the post that explains the meeting you then had to have.

Leverage Sitecore Analytics to Create Stunning Reports — Boris Brodsky, borisbrodsky.com The same problem from the data side, with CDP and Personalize feeding Microsoft Fabric and the reports built downstream. The line to keep is his parenthesis that "you currently need Sitecore CDP to access your analytics data", which prices the answer. The post closes with an offer to walk you through it; we have no independent record of his commercial interest, so read it as a practitioner's architecture proposal rather than a neutral comparison.

Someone rebuilt the official demo in Astro to make a point

Three posts in one week take the Sitecore Verticals demo, rebuild it on Astro instead of Next.js, and publish the result. Our harvest recorded no author for any of the three; the Demo Portal page carries Anton Tishchenko's name and date in the body, so that record of ours is wrong and the series reads as one team's work on exdst.com. That team builds and promotes the Astro JSS SDK the demo depends on, so the numbers are their own measurement of their own product against the alternative.

Sitecore Astro Verticals Demo Website: The Process — exdst.com; our record shows no byline on the page The honest engineering diary of the three. Stage one, layout in Astro with React components untouched, took one developer under a week with no prior Astro experience and proved nothing, because the output was the same. Stage two moved roughly 98% of components to .astro, tested by screenshot comparison against the Next.js build with Backstop.js. The claimed result is +15 Lighthouse on matched hosting and matched ISR settings, which is their figure, not an independent one.

Sitecore Astro Verticals Demo Website: Running on Sitecore Demo Portal — Anton Tishchenko, exdst.com (our record had no byline; the page carries his) The deployment walkthrough, and the reason it exists is the interesting part: the Demo Portal hardcoded Next.js as the framework until VERCEL_PROJECT_FRAMEWORK was exposed as an environment variable.

Sitecore Astro Verticals Demo Website: Running Locally — exdst.com; our record shows no byline on the page Short and mechanical. The line worth noting is that the editing chromes Astro JSS emits are the same ones the Next.js SDK emits, which is the only thing that makes the swap arguable at all.

The community built its own XM Cloud tooling

Sitecore Hackathon 2025: Team DED - My Thoughts and Review My Contributions — Dylan Young, dylanyoung.dev Young's team rebuilt a content export tool for XM Cloud as a standalone Next.js app against the Delivery and Authoring APIs, because the old XM/XP package-install route no longer exists. That is the real content of the post: a class of admin tooling stopped being installable, and the gap is being filled by hackathon projects on Vercel. He is also blunt about the AI assistance, noting that Copilot Edits "frequently overwrote my changes, causing me to lose code".

What we would take from March 2025

Judge a vendor month by the ratio between what was promised on stage and what the field had to work around the same week. In March 2025 that was three conference write-ups against six posts whose entire purpose was getting past something the platform did to you.

The through-line is rate limiting, and it is not a coincidence. Edge revalidation, the Send API and third-party calls at build time produced the same constraint from three unrelated directions in one month, and that constraint is what decides whether an XM Cloud site is operable rather than merely live. If you were planning that quarter, the thing to fund was queueing and retry around publishing, not a generative copy assistant. Reading a month from the archive has one advantage over reading it live: check the Stream roadmap Allen wrote down, and the Forms permissions gap he flagged, against what actually shipped before you quote either today.

Sources

sitecore · xm-cloud · sitecore-xp · ai · how-to · sitecore-personalize · content-hub · sitecore-headless