With a technical background and a keen eye for commercial impact, Adam bridges the gap between technology and business. In this article, Adam talks about the possibilities with V17: what exactly can V17 do for editors that V13 can't? This one's about AI.
If your organisation is running Umbraco V13, your editors can still publish pages, manage content, and build layouts. The CMS works. Nobody's complaining that loudly. And that's exactly why this conversation keeps getting pushed down the priority list.
But something significant has changed in the Umbraco ecosystem over the last twelve months, and it's not a feature release or a performance patch. It's the arrival of a proper AI layer, built into the backoffice, that only works on V17 and above.
And if you're a Head of Digital weighing up whether this upgrade is worth the investment, the honest answer is: the CMS capabilities are broadly similar. The AI capabilities are not even close.
What your editors are doing now that V17 handles differently
Let's be clear about something. V13 is a great CMS. Your editors can build pages, manage media, set up workflows with the right customisation, and publish content. Most of the core editing experience carries across versions.
The difference is what happens around the editing.
Right now, if your team needs SEO metadata written for a new page, someone writes it manually. If an image needs alt text, someone types it in. If content needs checking against brand guidelines, someone reads it and makes a judgment call. If you want to know which pages haven't been updated in twelve months, a developer needs to hardcode a report or, more likely, the editor scrolls through the content tree trying to remember.
None of that is broken. All of it is slow. And none of it scales when you've got hundreds or thousands of pages to manage.
V17 doesn't just make the backoffice faster or cleaner (though it does both, running on .NET 10 as a Long Term Support release with three years of guaranteed updates). It opens the door to Umbraco.AI, a modular AI layer that puts genuinely useful AI assistance directly into your editors' hands.
That layer does not exist on V13. It cannot be backported. The upgrade is the only path to it.
What Backoffice AI assistance actually does
Editors get AI suggestions built from the content already on the page (not generic outputs from a disconnected tool).
Switching between tools wastes time, breaks focus, and doesn't scale.
Umbraco.AI is a provider-agnostic AI integration layer built by Umbraco HQ, designed to let editors use AI directly inside the backoffice without needing a developer to set it up.
In practice, that means your editors can do things like generate alt text for images, create SEO titles and meta descriptions, summarise content, and tag pages for consistency, all from within the property they're already editing.
With one click, the AI produces a suggestion based on the page's actual content. The editor reviews it, adjusts if needed, and moves on.
These aren't generic outputs from a disconnected tool. Umbraco.AI Prompt actions reference the property values and page context directly, so the suggestions are grounded in the content your editor has already written.
On top of that, Umbraco.AI Agent and Copilot bring a conversational AI interface into the backoffice. Editors get a chat panel that understands where they are in the application and what they're working on. It can suggest changes, help clean up content, assist with structure, compare pages, and analyse content, all while keeping the human in the loop. Nobody publishes anything the editor hasn't reviewed and approved.
Umbraco.AI is also provider-agnostic. Your organisation chooses which AI provider to use, whether that's OpenAI, Anthropic, Google Gemini, or others, and swaps whenever you want.
The governance layer lets you define which models are used for which tasks, where AI is allowed to operate, and what rules it follows. For organisations in regulated sectors, this is the kind of control that makes AI adoption possible rather than theoretical.
None of this is available on V13. Every bit of it requires V17.
Content workflow automation: what's now possible that wasn't before
Updating content across hundreds of pages one by one is nobody's idea of a good use of time. Soon, editors will be able to describe a content task in plain language and AI will handle it across the entire site without needing to login!
Umbraco's Developer MCP Server already ships with 330+ tools spanning the full Management API. It's a very powerful way to build and maintain Umbraco projects, but it's aimed at technical teams.
What's more interesting for editorial teams is what's coming next.
Soon, editors will have the same precision and control as Umbraco developers; they will be able to manage site-wide content simply by conversing with an AI.
Umbraco HQ has confirmed a Hosted Editor MCP, designed specifically for content editors rather than developers. The vision is straightforward: open your preferred AI tool, connect to the Editor MCP, and describe your content task in plain language. Create a landing page, draft hero copy, rewrite pages to match your tone of voice, and generate meta descriptions across a section of the site. All through a conversation, no backoffice login required.
This is exciting stuff.
When it arrives, the implications for how editorial teams work day-to-day are significant. Your content editors become content directors, working at the speed of conversation rather than the pace of CMS workflows.
AgentRun: AI workflows that go beyond the page in front of you
Auditing hundreds of pages for SEO, brand compliance, or content freshness is a job that takes days. AgentRun does it in seconds and returns a structured report.
AgentRun is an Umbraco integration I built and personally contributed to the Marketplace. It's an open-source project, and Cogworks uses it alongside our client work. It's complementary to what Umbraco HQ ships, not a replacement for it.
The distinction is simple:
- Umbraco.AI helps an editor with the page they're currently working on. It's like having a smart colleague sitting next to you.
- AgentRun runs multi-step AI workflows across your entire content tree. It's the consultant you hire to take a brief, go away, do the work, and come back with a structured report.
It handles tasks such as content quality auditing across your entire site, SEO metadata auditing, content freshness reporting, accessibility scanning, brand voice compliance reviews, and content model health checks. An editor clicks a button in the backoffice and gets a structured deliverable back, not a chat response.
A practical example of how AgentRun works
Say you have a "people" section with 200 staff profiles. Every month, someone needs to check whether anyone has left, whether profile information is up to date, and whether headshots need updating. Right now, that's a manual job that takes hours.
With AgentRun, the agent reads every profile directly from the CMS, checks it against the criteria you've defined, and produces a report of what needs attention. The job that took hours takes seconds. The editor reviews the findings, makes the decisions, and updates what needs updating.
The key is that it reads content directly from your CMS. No copy-pasting into ChatGPT, no losing context between pages, no unstructured outputs. Copilot helps your editor with the page in front of them. AgentRun handles work across the entire site.

A Russell Cooke and Cogworks project
What V17 means
for your content team and commercial goals
• Editor time recovered
Every page that needs an SEO title, a meta description, alt text on images, or a brand compliance check currently costs someone time. Multiply that across hundreds of pages, and it adds up fast. AI-assisted editing doesn't eliminate the editor; it eliminates the drudge work. Your team spends their time on judgment and creativity, not on writing the fifteenth meta description of the afternoon.
• Content quality at scale
When metadata is generated from actual page content rather than rushed by an editor at 4 pm on a Friday, the quality is more consistent. When brand voice compliance is checked by an agent that reads your actual guidelines rather than relying on someone's memory, the output is more reliable. Not perfect. But consistently better than "we'll get to it when we have time."
• Compliance and governance
For regulated sectors, the AI governance layer in Umbraco.AI is significant. You control which models are used, where AI operates, and what rules it follows. Activity logs capture when and how AI is used. Nothing goes out without a human sign-off. That's auditable, explainable, and defensible.
• Reduced technical debt
V13 reaches end of life in December 2026. After that, no security patches, no bug fixes, no official support. V17, as a Long Term Support release, provides support through at least late 2028. That's runway. And if your competitors upgrade and start using AI-assisted content workflows while you're still on V13, that gap compounds over time. The upgrade cost is a known quantity. The cost of standing still is harder to measure, but real.
The Umbraco V17 upgrade is the unlock
Everything in this article requires V17. None of it works on V13.
The Umbraco V13 to V17 upgrade is not a like-for-like CMS refresh. It's the point where your CMS becomes AI-ready, where your editors get tools that genuinely save them time, and where your content operations can start scaling without scaling your headcount.
The question worth sitting with isn't "does V13 still work?" It's "what are we leaving on the table by staying on it?"
If you want to understand what the upgrade would look like for your specific setup, that's a conversation worth having. You can book an upgrade chat with me below.
Innerworks and Cogworks are proud to partner with Community TechAid who aim to enable sustainable access to technology and skills needed to ensure digital inclusion for all. Any support you can give is hugely appreciated.
