The DXP Catalyst Update - Sept 25, 2026

OPENING

Most of what shipped in the last two weeks was operational rather than headline material, and it clustered around a question buyers will have to answer once agents are doing real work: can you see what an agent did, what it cost, and whether it worked? Sitecore now routes AI content recommendations into Page Builder as new content versions that move through existing approval workflows. Contentful began streaming Content Management API logs into customers' own observability tools. Adobe started rolling out reporting in Customer Journey Analytics that ties conversational agent interactions to conversion. Optimizely shipped automatic model-tier assignment inside its agent platform, which moves decisions about cost and model choice to the vendor side.

The other thread was vocabulary. Acquia relaunched its brand and no longer describes itself as a digital experience platform, joining Optimizely and Contentstack in stepping away from the label.

INDUSTRY PULSE

Salesforce completed its acquisition of Contentful on September 1. The price, undisclosed at announcement, appears in Salesforce's most recent quarterly filing as approximately $1.5 billion in cash, net of the Contentful shares Salesforce already owned. Salesforce's stated rationale is adding a native content layer to Headless 360, with Agentforce as a primary consumer of that content. Contentful's roadmap is now Salesforce's to set, and organizations evaluating Contentful, including those that run no other Salesforce products, should weigh how Agentforce and Headless 360 priorities could shape it over time. Contentful's product updates from the first weeks under Salesforce ownership are covered in the vendor section below.

Acquia's rebrand extends the relabeling we noted in our Opticon recap. Acquia now calls itself an 'agentic content platform,' joining Contentstack ('Agentic Experience Platform') and Optimizely (an AI platform for marketing) in leading with content and agents rather than the DXP category name. Most of the vendors we track still use the DXP label, and so far the change is mostly in positioning, with the products underneath largely the ones already in market.

VENDOR ROUNDUP

Acquia

Acquia Relaunches as an "Agentic Content Platform," Powered by Drupal

Acquia launched a new brand on September 14. The logo now reads "Powered by Drupal," and the homepage describes Acquia as an agentic content platform rather than a digital experience platform. Co-founder Dries Buytaert wrote that Drupal had become less visible in how Acquia described itself over the past five years. He said the bigger reason for the rebrand is that the products have evolved faster than awareness of them.

The product story centers on Acquia Source, which brings content management, digital asset management, and web governance into one workspace, with Acquia AI coordinating agent work across those tools. One detail for existing customers: Source is positioned to bring sites running on either offering into that shared workspace. That covers Acquia Source CMS, the fully managed Drupal SaaS offering, and Acquia Cloud, for teams that want full control of their Drupal sites.

The relaunch makes Drupal the explicit basis of Acquia's governance argument. Structured content, granular permissions, workflows, and revision history are presented as the controls agents need to work safely. Buytaert also said Acquia will invest more in helping the Drupal community reach developers and marketing leaders who have not looked at Drupal in years. The Acquia Partner Summit at DrupalCon Rotterdam on September 29 is the next public checkpoint.

Adobe

Customer Journey Analytics Adds Reporting on Agent Conversations

Customer Journey Analytics (CJA) began rolling out Conversation Insights on September 9. General availability is now listed for September 30 after slipping from September 22. The feature collects prompts, responses, and agent metadata from conversational agents, whether an organization's own or Adobe Brand Concierge, and brings them into Analysis Workspace. There, teams can analyze intent and sentiment and connect those interactions to conversion and engagement outcomes.

It arrives alongside two related pieces. CJA MCP server plugins for ChatGPT and Claude, generally available since September 1, let those assistants query CJA data directly. A planned integration connecting Adobe Brand Visibility to CJA data is listed for September without a firm date. Taken together, CJA is being set up to measure agent-driven journeys in both directions: agents can query the analytics layer, and the analytics layer can report on what agents do. For organizations deploying conversational agents on their own properties, the useful test is whether those interactions tie into the same journey data used for everything else, rather than sitting in a separate reporting tool.

Contentful

First Weeks Under Salesforce: Observability and Slack Access

Contentful's changelog since the close shows a steady cadence of enterprise-oriented releases. On September 22, Enterprise Observability began streaming Content Management API logs to customers' own AWS, Google Cloud, or Azure storage in near real time, for ingestion into tools like Datadog or Splunk. That extends the near-real-time audit logs Contentful introduced in August, which capture read access as well as writes. The same week, the Contentful Optimization Agent became available in Slack as a read-only app for questions about audiences, personalizations, and experiments.

For organizations running agents against Contentful through its MCP server, management API logs are how a team reconstructs what an agent changed and when.

Contentstack

Self-Serve Sign-Up and a Free Tier

On September 16, Contentstack opened self-serve account creation. Anyone can now sign up, start on a Free subscription, and move up to Build or Growth tiers, with usage tracking, optional overage, and billing managed in the account. Contentstack described the offering as aimed at smaller teams and individual builders.

For a vendor that has sold primarily to enterprises, this lowers the cost of a hands-on evaluation considerably. Partners and in-house developers can now test content modeling and the developer experience before a sales cycle starts. Check what the self-serve tiers include relative to the AXP components, Data Cloud and Agent OS, before drawing conclusions about the enterprise offer.

ContentCon Europe Opens September 30 With AXP Certification

ContentCon runs September 30 to October 1 in Amsterdam and October 27-28 in Chicago, themed around designing experiences for human visitors and AI agents at once. New this year, Day 1 at both events includes certification workshops for Contentstack AXP and Agent OS at no additional cost to attendees. Customer sessions include two worth noting:

• Verizon will present "Page Craft," a process that lets the company launch brand-approved pages without sending each one through legal review.

• Volvo Cars will cover content modeling and localization across more than 100 markets.

Optimizely

Following up on our Opticon recap, three releases from the past two weeks extend what was announced on stage.

Model Routing Ships as Sub-Agents

On September 17, Optimizely released sub-agents in its agent platform. They assign work automatically across Mini, Fast, and Deep tiers based on task complexity, so harder requests get more reasoning, and Optimizely manages the tiers on every instance. A few days earlier, the Standard, Balanced, and Complex inference levels moved to Gemini 3.8 Flash.

This is the model routing concept previewed at Opticon, now in product. Optimizely's credit documentation states that how it allocates workloads across models does not alter credit consumption. Under current terms, that places the cost of routing decisions on Optimizely rather than the customer.

Agent Platform Reaches Further Outside the Optimizely Stack

The September 18 release added several connectors: LinkedIn Ads and OpenAI Ads connector tools, remote MCP connectors for Braze and Fullstory, and OneNote and sensitivity label tools in the Microsoft 365 connector. On September 21, Optimizely added personal files, a persistent document store that agents and Virtual Teammates draw on automatically during tasks.

The connector list tracks the integration story from Opticon, with the agent platform positioned to work across a marketing stack Optimizely does not own. The sensitivity label support is a small but useful governance signal for Microsoft-centric organizations, since it lets agents check and apply existing data classification rather than work around it.

CMS 13 Exposes Its Tools to Any MCP Client

On September 11, Optimizely added an MCP server to its CMS 13 tooling. It exposes the same CMS tools available in Opal Chat to any MCP-compatible client, including Claude Code and Cursor, covering content, content types, and display templates. Authentication is by client secret today, with Opti ID authentication planned.

For CMS 13 customers, agent access to the CMS no longer depends on working inside Optimizely's own agent interface. Organizations planning to use it beyond development teams may want to wait for Opti ID authentication, since client secret authentication does not tie actions to individual users.

Sitecore

Scrunch Recommendations Move Into Page Builder

On September 16, Sitecore added two early access capabilities powered by Scrunch. AEO opportunities in Channels surface the top improvement areas across a site, map them to the affected pages, and show projected lift where available. Optimize content with AI brings discoverability scores and recommended edits into Page Builder. Accepting a recommendation creates a new version of the content item, highlights the changed field, identifies other pages reusing that item, and sends the change through the existing workflow and approval process.

The detail worth weighing is where the recommendations land: inside the authoring tool, versioned, and subject to the same approvals as any other edit. Both features require a Scrunch subscription, which belongs in any cost comparison for SitecoreAI.

Grounded Generation From the DAM, and a More Unified Product Shell

Earlier in September, Sitecore opened early access to an integration between SitecoreAI DAM and Scrunch Knowledge Studio. Teams select approved source documents in the DAM and Scrunch turns them into a structured knowledge document. Scrunch agents flag sensitive content, factual errors, and conflicting information for review. Once a person approves the document, it becomes grounding input for AI-optimized content. Access is limited to selected organizations for now.

Separately, a September 8 SitecoreAI update describes work to bring Sitecore's applications into a shared design system with unified navigation and a single login. The grounding workflow addresses a common gap: product specs and approved messaging often live in a DAM where public AI tools never see them. The workflow is only as reliable as whoever owns the refresh cycle when that DAM content changes.

Nothing surfaced this period for Squiz, Magnolia, or Uniform.

WHAT I’M WATCHING

Optimizely's sub-agents are an early shipped version of a pattern I expect other vendors to follow, where the vendor decides which model handles a task. Under Optimizely's current credit terms, model allocation does not change credit consumption, which works in the customer's favor as long as it holds. What I am watching is whether vendors keep flat pricing as routing gets more sophisticated or move to pricing by tier. Buyers signing multi-year agreements with AI credit components should get that answer in writing.

Contentful's first weeks under Salesforce look like business as usual, with governance and observability releases and a Slack app. The signal I am waiting for is the first Agentforce integration, and whether it ships to Contentful customers generally or only to those licensed on the Salesforce side. That answer will tell independent buyers most of what they need to know about Contentful's composable positioning over the next few years.

WebMCP is moving faster outside the DXP vendors than inside them. Optimizely's implementation, Agent-Ready Experiences, remains in closed beta. In August, Shopify enabled WebMCP tools on its storefronts and Cloudflare launched a developer preview. The specification is still a W3C community group draft, and support beyond Chrome and Edge is uncertain. What I want to see from CMS vendors is authoring control over which site actions get exposed to agents, governed by the same permissions and approvals that apply to content today.

ONE QUESTION WORTH ASKING

When an agent edits a page, queries your analytics, or runs a task on a model your vendor chose, can your team reconstruct afterward what it did, what it cost, and whether it worked?