What is left, and what gets harder
The State of Martech report has a clean answer. It calls it the middle layer. The space between data and activation, where decisioning, orchestration, constraints, governance, and feedback live. Pega frames it as next-best-action across customer journeys. SAS frames it as the integration of decisioning into operational systems. Hightouch frames it as the marketing context layer. The names differ. The substance is the same. The middle layer is where context, rules, permissions, and judgment live. It is the part of software that AI does not eat from underneath, because it is the part that AI itself depends on to be useful.
That is the half of the story Dries pointed at without naming directly. Agencies are rebundling around the middle layer. Strategy, governance, orchestration, outcomes. Software is rebundling around the middle layer too, with the same logic, on a slightly delayed schedule.
What the software rebundling actually looks like
It looks like the collapse of separate categories into integrated platforms. A CMS that is also a CDP that is also a marketing automation tool that is also an analytics layer that is also an agent runtime. Not because anyone wants a giant monolith, but because the only way to serve an AI agent the context it needs is to put content, customer data, workflows, permissions, and decisioning behind one coherent layer. Five separate tools cannot do this without a context bridge, and the context bridge is, in practice, a sixth tool that nobody wants to buy.
We are already seeing the early moves. Adobe is bundling its DXP, its analytics, its CDP, and its Firefly agents into one Experience Cloud command surface. Salesforce is doing the same with Data Cloud and Agentforce. Acquia, the largest commercial vendor in the Drupal world, launched Acquia Source in 2025 explicitly as an agentic content governance layer on top of its existing stack, and added three more agents in December. Sitecore is moving in the same direction. Optimizely is too. The Forrester DXP Wave for Q4 2025 frames the entire category as moving toward “agentic DXP,” and the leaders are scored on how well they orchestrate agents across the bundle.
This is the rebundling, and it is happening fast.
Two paths
Here is the part the essay did not fully spell out, and the part I think matters most.
The rebundling will happen one of two ways.
Path one: closed proprietary suites do the rebundling. Adobe, Salesforce, Sitecore, Optimizely. They have the channel, the data, the budget, and the installed base. They will collapse what was 30 line items in a CMO’s stack into 3, and they will offer governance, agent orchestration, and AI conformity as part of the deal. The customer will love the first 18 months. The customer will hate the next 10 years, because the lock-in that used to be at the level of “we are stuck on this CMS” becomes “we are stuck on this entire experience cloud, our data is in it, our agents run on it, our decisioning logic lives in it, and there is no realistic migration path.”
Path two: open platforms do the rebundling. The same collapse happens, but the substrate is open source. The CMS, the marketing automation layer, the CDP, the analytics, the agent runtime, the MCP server. The customer keeps the keys. Data ownership is real. Model choice is real. Integration choice is real. Migration is hard but possible, because the alternatives exist and the protocols are open.
For most of the last decade, the open path was the harder sell. Open source DXPs were technically credible but commercially disadvantaged. The integrators were smaller. The buy cycles were longer. The proprietary suites had better top-down sales motion. The procurement language defaulted to incumbents.
Three things are flipping that.
First, AI is collapsing the cost advantage of large vendors. The work that used to require a 5,000-engineer R&D budget can now be done by a focused community contribution effort with 100 well-coordinated engineers, because most of the heavy lifting is in the foundation model layer that everyone has access to. The Drupal AI Initiative crossed 1.5 million dollars in committed cash and in-kind contributions in five months, with 31 partner organisations and 26 pledged full-time equivalents. That is a small absolute number compared to Adobe’s R&D budget. It is also more than enough to ship a credible agentic substrate, because the substrate does not need to be invented from scratch anymore.
Second, the European procurement framework has moved. The EU Cloud Sovereignty Framework, the Cloud and AI Development Act, SecNumCloud 3.2 in France, BSI C5 Type 2 in Germany, the European Digital Sovereignty Declaration of December 2025. These are not opinions. They are tender qualifications. For a growing share of European public sector and regulated industry workloads, closed US proprietary suites are now disqualified before the demo. Open European platforms are not a preference. They are a requirement.
Third, customers have learned what lock-in costs. The CMOs and CIOs who bought monolithic proprietary suites between 2015 and 2022 are the same CMOs and CIOs who are now writing down the migration costs, the integration costs, and the strategic optionality they lost. They will not sign the next decade away as quickly as they signed the last one.
“But doesn’t open rebundling create its own lock-in?”
This is the fair question, and the one I want to answer directly, because the answer is the whole point.
When an open DXP rebundles, the lock-in changes shape. With a closed suite, you are locked into the software, the runtime, the data model, and the vendor relationship as one indivisible thing. With an open DXP, you are locked into nothing that you cannot take with you. The software is open source. The data model is yours. The integrations are open protocols. The agents run on standards like MCP that any vendor’s tools can speak.
What remains is the runtime. Software still needs a well-built PaaS to run at scale, with security, performance, monitoring, deployment discipline, and uptime. That is real work. We do it at Dropsolid AI because we built the open DXP and we know how to operate it. The Dropsolid AI Open DXP runs best on the Dropsolid platform for the same reason any open source software runs best on the platform of its main contributor. We tune for it, we operate it daily, we ship the integrations first.
But this is where the openness story has to become real, and where I want to be clear with the market.
The dream, as I wrote in Freedom to leave, applied to the whole stack, is that the Dropsolid AI Open DXP becomes portable to any serious PaaS that can run Drupal at scale. I want to see Pantheon offer it. I want to see Platform.sh offer it. I want to see Acquia offer it. That would prove the openness is not marketing. That would make the customer’s freedom to leave real, not aspirational. And it would force every vendor in the category, including us, to compete on the strength of the software and the operations, not on lock-in.
This is not naive. We win in that world because we built the substrate, we contribute a lot of code & expertise to it, we know it better than anyone, and our customers stay with us because we deliver outcomes, not because they cannot leave. That is the only kind of vendor relationship that ages well. Our intention is to open up the code of the open DXP.
The closed rebundling path makes the customer’s exit cost grow every year. The open rebundling path makes the customer’s exit cost shrink every year. Over a 10-year horizon, those two curves diverge by an enormous amount. And the customers who paid the price last decade are not going to pay it again.
Where this ends up
We are not heading for a Wild West fragmented software market where AI dissolves every category and customers buy 50 specialised tools. That is the 2023 hypothesis, and it is wrong. We are heading for a market where software rebundles aggressively. The only question is whose rebundle wins.
The closed rebundle has incumbency, capital, and channel.
The open rebundle has freedom, sovereignty, and the procurement law of an entire continent.
I think both will exist. I think the closed rebundle will dominate the US enterprise market and large parts of the global brand market. I think the open rebundle will dominate the European public sector, the regulated industries, the mission-driven brand cohort, and a meaningful share of mid-market everywhere that values not being locked in. Those are different territories. Both are large.
The work for everyone building software, or building agencies, or buying from one, over the next 24 months, is to pick a side and execute on it. Half-measures will be eaten by both rebundles. A point tool with no platform story will be commoditised by the foundation model layer. A platform with no governance and sovereignty story will be outflanked by the closed suites or by the open alternatives, depending on the geography.
The agencies that survive will rebundle around outcomes, as Dries said.
The software vendors that survive will rebundle around context, governance, decisioning, orchestration, and sovereignty.
The companies that get to do both, on an open substrate, with the customer holding the keys, are the ones I am betting on. And the next 24 months are when that bet gets settled.
Note: Dominique De Cooman is founder of Dropsolid, a member of the Drupal Association Board, and a leader in the Drupal AI Initiative. Dropsolid AI is a key contributor to Drupal (AI) and Mautic.
