Key Takeaways
- Bryan Cheung, CEO of Liferay, offers that CMS quit being a publishing endpoint years ago. With 60% of organizations running three or more teams through one content tool and 63% of content managers publishing beyond the website, governance has to live where content actually ships, not in scattered email threads and spreadsheets.
- AI sits in 86% of content stacks, yet only 14% of teams trust it to publish on its own. Heavy AI users juggle more tools, not fewer, proving AI is feeding tool sprawl instead of fixing it.
Your CMS was built to publish. That job ended years ago. Today it’s coordinating approvals, metadata, brand standards, and content bound for a dozen destinations at once, and most teams are improvising to keep up.
Liferay’s report, The Modern CMS Has Outgrown Its Original Job, puts hard numbers to the strain as 60% of organizations now run three or more teams through the same primary content tool and 63% of content managers publish well beyond the website. The CMS stopped being a publishing endpoint long ago. The way you operate hasn’t caught up.
Then there’s artificial intelligence (AI), which muddies the water rather than clearing it. AI now sits in 86% of content stacks, yet just 14% of teams fully trust it to publish on its own. Here’s the twist: instead of consolidating your work, AI is feeding tool sprawl. Heavy AI users juggle more tools, not fewer, and feel the friction sharper. More capabilities, more logins, more gaps between where content gets made and where it goes live.
To unpack what it means for your content operations, we sat down with Bryan Cheung, CEO of Liferay. In this Q&A, Cheung digs into how to evolve the CMS from a publishing endpoint into a governed content hub where approvals and standards actually hold as well as how to close the AI trust gap across teams that range from AI-enthusiastic to AI-wary.
Demand Gen Report (DGR): Bryan, thanks for taking time to talk with us today. What separates organizations that are successfully operating their CMS as an enterprise content hub from those still treating it as just a publishing endpoint?
Bryan Cheung: Thanks for having me. The biggest differentiator is whether the content platform is where governance actually happens, or just where content lands at the end of someone else’s workflow. Our survey found that 60% of organizations have three or more teams sharing the same primary content tool, and 63% of content managers are publishing beyond the website.
The organizations pulling ahead have restructured around that reality: approvals, metadata, brand standards, and channel publishing all run through a single system. The ones still stuck in publishing-endpoint mode are managing all of that coordination informally upstream, which works until you are running content across multiple channels, languages, and internal teams simultaneously.
DGR: The report shows that 78% of respondents switch between multiple tools to complete a single content task — which tool transitions appear most harmful to speed, governance, or content quality?
Cheung: The most damaging transition is the one between drafting and publishing, because that is where metadata gets skipped, formatting breaks, and any AI-assisted work done upstream fails to carry over into the governed environment.
The second category that causes serious harm is anything involving approvals and compliance review: when those workflows live outside the publishing system, in email threads or project management tools, it becomes very difficult to maintain a clean audit trail. Governance breaks down not because anyone is being careless, but because the system is not built to enforce it.
How to Bridge the Trust Gap
DGR: You found that AI is present in 86% of content stacks, yet only 14% fully trust it to publish autonomously. What specific trust-building mechanisms would most effectively close that gap?
Cheung: The trust gap is less about autonomous publishing and more about where AI fits reliably into a human-led workflow. The highest-value applications right now are the ones that reduce manual work without removing human judgment from the equation: AI-assisted drafting, automatic metadata tagging, content localization, and flagging compliance issues before review.
The trust-building mechanism that matters most is giving teams clear visibility into what AI is doing and where it is falling short, so that confidence is built on actual track record rather than assumption. As that track record develops, organizations can thoughtfully expand AI’s role in the workflow, with human review remaining the standard for anything that goes live.
DGR: Did the survey reveal any meaningful differences in AI trust or workflow complexity by team type — for example marketing versus IT, CX, operations, or sales?
Cheung: The teams most embedded in the publishing stack are also the most AI-enthusiastic and the most fragmented. 58% of IT and web administration respondents use AI extensively, the highest rate of any major role group, and they are among the most likely teams to switch between multiple tools to finish a single content task. Operations roles like finance and HR sit in the middle of the curve, with steady adoption but more caution about autonomous publishing.
Teams closer to the customer look very different. Only 14% of customer experience respondents use AI extensively, none of them completely trust AI to publish autonomously, and 20% say they do not trust it at all. Sales is in a similar place, with just 2% completely trusting AI for autonomous publishing. What this suggests is that AI trust is not a uniform organizational sentiment. It is roughly proportional to how often a team has to live with AI-assisted content in production. Teams that own the platform and publish constantly are pulling AI deeper into their workflows and feeling the seams sharper. Teams that own customer relationships and revenue conversations are keeping AI at arm’s length. For organizations rolling out AI broadly, the trust gap is real, and it is biggest with the people closest to your customers.
DGR: Did the data show any signal that tool sprawl is becoming a bigger problem because of AI, rather than being solved by AI?
Cheung: It seems that heavy AI users are more fragmented, not less. 31% of teams that use AI extensively switch between tools “very often” to complete a single content task, compared with 17% of limited AI users and just 10% of non-AI or piloting users. The pattern shows up on stress too: 27% of heavy AI users name “too many tools or platforms” as a top day-to-day stressor, against 17% of non-AI or piloting users.
The survey can’t tell us exactly whether it’s AI causing that fragmentation or whether organizations with already-complex content stacks are simply the ones adopting AI most aggressively. But the practical conclusion is similar either way. For most organizations, AI is being added on top of the existing stack rather than replacing anything. A drafting assistant lives in one system, localization in another, compliance review in a third, and publishing in the CMS. Each new AI capability tends to arrive in its own tool, with its own login, its own data context, and its own gap with the system of record. Until AI capabilities are integrated into the core publishing environment rather than bolted on around it, “more AI” will continue to mean “more tools to manage” rather than fewer.
How to Navigate Complex Buying Committees
DGR: For B2B organizations with complex buying committees and long sales cycles, how should CMS and content platform strategies evolve to better support account-based marketing, personalization, and post-click experiences?
Cheung: The core problem is that most content platforms are optimized for publishing, not for what happens after the click. For B2B organizations managing complex buying committees, that disconnect is costly: a CIO evaluating a platform for security needs a fundamentally different experience than a marketing director evaluating it for usability, and if the CMS cannot support that kind of contextual differentiation at scale, teams default to building separate landing pages manually, which is slow and nearly impossible to govern consistently.
What needs to evolve is the relationship between the CMS and the data layer around it, with tighter integration with CRM, marketing automation, and intent data so personalization can be applied systematically rather than one-off.
DGR: What does “integration” actually need to mean in practice for modern content operations— shared metadata, workflow orchestration, identity permissions, analytics continuity, or all of the above?
Cheung: If I had to identify where integration failures cause the most harm most often, I would point to metadata and analytics continuity. Metadata is where content strategy breaks down in practice: when content moves across tools before reaching the publishing system, metadata either does not transfer or gets filled in inconsistently, creating downstream problems for search, personalization, and governance that compound over time.
Analytics continuity is the other critical gap, because most teams are measuring performance inside individual tools without a clear picture of how content performs across the full journey. Workflow orchestration and identity permissions matter too, but they tend to be more tractable. The metadata and analytics problems are harder because they require common standards across tools, not just technical connections between them.
DGR: What should enterprise teams measure if they want to evaluate whether their content operations are improving — beyond output volume?
Cheung: Output volume is one of the least useful measures of content operational health, and it is still the one most teams default to. The signals that actually matter are workflow cycle time (how long does it take to move from draft to publication, and where is the friction), governance adherence (what percentage of published content has complete metadata and a clean audit trail), and team-reported stress around coordination and tool complexity. That last one sounds soft but it is not: the top stressors in our survey, cross-team coordination, too many manual processes, too many tools, are leading indicators of where content operations are about to break down.
Keys to CMS Architecture Producing Content that Delivers
DGR: The report frames the CMS as the destination, not the workspace. Do you believe the future is a better CMS authoring experience, or a more connected ecosystem where drafting still happens elsewhere?
Cheung: Probably both, but on different timelines. The honest read of our data is that the CMS as a drafting environment has not succeeded yet: only 7% of content managers draft in their primary content platform, and that number barely moves even among heavy AI users.
The near-term opportunity is not replacing the drafting environment but closing the gap between where content is created and where it is governed, so that metadata, brand standards, and workflow context carry over rather than getting lost in the handoff. Longer term, as AI becomes more deeply embedded in the authoring process, the case for doing more of that work inside a single governed environment gets stronger, but that shift will be earned, not assumed.
DGR: For organizations currently rethinking their MarTech stack, what are the clearest warning signs that their CMS architecture is no longer fit for how content is actually produced and governed?
Cheung: There are a few I would treat as serious signals. The first is when your team has built informal workarounds everyone knows about but no one has sanctioned: shared drives standing in for asset libraries, spreadsheets tracking approvals, Slack threads serving as the content calendar. The second is when adding a new channel or team requires a development project rather than a configuration change. The third is when governance depends entirely on individual discipline rather than system enforcement. If your metadata standards and brand guidelines live in a style guide but are not enforced anywhere in the publishing workflow, you have content guidelines, not content governance, and at scale that distinction matters enormously.
