From Content to Context: One Year Later

The Market Verdict

The user behaviour shift I described in Frankfurt didn't slow down. It accelerated. By late 2025, 58% of Google searches were ending without a click. In Google's AI Mode, that figure has already reached 93%. Organic click-through rates dropped 61% on queries where AI Overviews appeared. HubSpot's CEO Yamini Rangan called it a "traffic apocalypse" at INBOUND 2025, and whatever you think of the framing, the data is hard to argue with. 

The transactional shift moved just as fast. AI agents drove an estimated $262 billion in global retail revenue over the 2025 holiday season, roughly one in five purchases. The TikTok-as-stylist behaviour I pointed to in Frankfurt wasn't a curiosity; it was a leading indicator. 

And the underlying market signal is striking: in Brinker and Riemersma's State of Martech 2026 report, CMS & Web Experience Management is the single fastest-growing subcategory this year, up 21.4%. The reason they give is that this category provides "structured, machine-readable context that AI systems can use." I didn't underestimate the direction. I did underestimate the speed. 

The Infrastructure Gap

Here's the view my crystal ball was still a bit fuzzy on twelve months ago. Marketing teams saw the shift and most understood what needed to change, and then ran straight into their infrastructure. 

Despite 80% of marketing leaders feeling pressure to adopt AI, Supermetrics reported that only 6% of organisations have fully embedded it into their workflows. Gartner found 45% of martech leaders say existing AI agents are failing to deliver the business performance they were promised. But the statistic that really jumped out at me is from Jasper.ai: 62% of CMOs report they can prove AI ROI at their organisations, but only 12% of individual contributors agree. That's a measurement-and-reality gap. The people closest to the work, the ones running campaigns, building journeys, trying to activate years of customer data, understand what the infrastructure can and cannot do. 

It isn't a budget or skills problem; it's an access problem. PwC's 2025 CMO Pulse Survey found that CMOs named unclear ownership and limited access to data as their top barrier to strategy execution. The data is there. They just can't reach it. Anyone who's had to copy and paste spreadsheet data from one place to another just to get an insightful overview will know exactly what I'm talking about. 

You cannot orchestrate context across systems that don't share data. You cannot run a learning loop when signals are siloed and campaign logic sits somewhere else entirely. As our CEO Dominik Pintér wrote recently in The Hidden Cost of Legacy Platforms, the most damaging cost of a fragmented stack isn't system failure. It's the way organizations quietly adapt their ambitions to what the infrastructure permits. 

Why the Industry Has the Governance Framing Wrong 

Here's where I want to push back on something the category has nearly unanimously agreed on. Almost every major DXP shipped governance features in the last twelve months: audit trails, approval workflows, brand compliance checks, AI content guardrails. The collective bet has been that governance is the answer to AI risk, and the race is on to see who can add the most checkpoints. 

Forrester is already predicting the outcome: a third of brands will harm customer experiences in 2026 by deploying AI self-service prematurely. More governance features won't prevent this. The framing is the stumbling block. 

Treating governance as a control layer still presents the choice as the age-old trade-off between speed and safety. The assumption underneath is that oversight needs to be uniform. Every AI action reviewed, every output approved. This treats risk as a constant when it isn't. Context is the thing that determines where risk actually lives. 

Governance without context is just bureaucracy. 

The question that matters isn't "was this approved?" It's "did the agent or platform understand enough to know when it needed to ask?" 

Borrowing a Page From Cybersecurity

My years working in cybersecurity just won't let me leave this alone. Security architects have been building risk-adaptive systems for more than a decade. Zero Trust assumes nothing is safe by default and re-evaluates at every interaction. ABAC makes decisions at runtime based on who's acting, what they're doing, and the conditions around it. Gartner's CARTA framework makes it explicit: trust is dynamic, so it should be scored continuously and applied proportionally. 

You see this in production every day. You don't get challenged for MFA every time you sign in; you get challenged when context changes, like a new device, an unusual location, or a high-stakes action. The system reads the signals and applies friction only when warranted. It's arguably safer, because the user's attention is preserved for the moments that genuinely matter. 

The CMS and DXP industry hasn't really imported this thinking yet. The Cloud Security Alliance has formalized an Agentic Trust Framework for AI agents. Gartner predicts half of organizations will adopt zero-trust data governance by 2028. None of it has crossed into the DXP conversation, and it needs to. 

A context-orchestrating DXP already has every signal required to do this well: who's acting, what they're trying to achieve, what the customer's history says, what brand and compliance constraints require, how confident the model is in this specific domain. The same context that makes the experience adaptive can make the governance adaptive. Same architecture, same data, just a different decision being scored at runtime. 

What Comes Next 

Time to lay out some predictions for me to revisit twelve months from now. Heading into year two of context orchestration, four things look clear. 

First, the agentic shift is decided. What isn't decided is whether marketing teams will have the platforms and context infrastructure in place when their AI-native ecosystems go live, or whether they'll deploy them blind into fragmented stacks and be surprised when results don't match the promise. The organizations best positioned won't be the fastest. They'll be the ones building foundations now. 

Second, data activation is the next defining frontier for agencies. The brief will shift from designing what good content delivery looks like to designing for data activation. I'd like to see the vendor conversation move on from "what does this CMS do?" to "how does this platform understand what our customers are doing, and what they need next?" 

Third, the marketer's role keeps changing, and the platform has to change with it. Marketers are increasingly behaving as citizen developers, connecting automations across their stack, building their own assistants, and extending the agents they already have. Context engineering, that my colleague Sean Wright wrote about late last year, is here to stay. 

Over the next twelve months we'll standardize around where it actually lives. I'd argue it lives inside the DXP itself: the new "context brain" that captures intent inside the product and receives signals from outside it, then hands work back to specialist tools without losing the thread. 

Fourth, governance will adopt risk-adaptive thinking from cybersecurity, and the platforms that move first will own the conversation. Risk-adaptive governance changes the ROI story the way adaptive MFA changed enterprise security a decade ago. Vendors who can claim genuine risk-adaptive governance, not configurable approval flows with a new label, will reset what "enterprise-ready AI" means in our category. 

From where I stand today, the narrative I see getting louder is this: humans decide, the platform executes. Marketers define the outcome (campaign, audience, journey, channel, and what good looks like) and the platform handles the orchestration, with the marketer back in the role of reviewer, strategist and guardrail-setter rather than the hands-on mechanic. 

At Kentico, our continued investments into evolving AIRA and KentiCopilot are how we're building toward this future. The single pane of glass vision we've been pursuing isn't the destination anymore. We're already there, and it's simply the starting line.