For decades, the economics of investment research have rested on a fragile but functional architecture: (1) Attribution, (2) Entitlement, and (3) Feedback. Research is produced by the sell side, distributed through controlled channels, and consumed by the buy side. Critically, the producer has always retained some visibility into who consumed it, how it travelled, and why it mattered commercially. The system was imperfect: MiFID II exposed how imperfect, but its underlying structure held. Value could be traced, contribution could be recognized, the producer and consumer remained, however tenuously, in relationship with one another.

That relationship is now being severed.

AI does not read research the way humans do. It ingests, fragments, recombines, and produces something new - an answer. In that transformation, the connection between producer and consumer does not only weaken, it disappears. The buy side no longer reads the report. They ask a question, the model responds, and the answer that emerges may draw on the work of dozens of analysts across a few institutions, none of whom are visible, attributed, or compensated in proportion to their contribution.

The answer becomes the product. And so, the question that follows is both simple and profound: Who owns it?

The implications of that question reach further than MiFID II ever did. That regulatory shift challenged how research was paid for. It was disruptive and consequential, but it left the fundamental structure of the research relationship intact. A report was still written. An analyst was still credited. A client still read, considered, and decided. The producer and consumer, however reorganized commercially, remained connected.

AI challenges something more foundational.

It challenges whether research is recognized at all. When a model synthesizes a view from multiple sources, attribution disappears into the synthesis. When content is ingested into a private model, the entitlements that govern document access do not extend to govern query access. When summaries are generated internally, the feedback loop that connects producer to consumer, the loop that tells a research department what is resonating, what is valuable, what justifies continued investment, collapses entirely.

This is not a pricing problem, but a structural one. If research is consumed but not seen, used but not credited, valued but not traceable, its economic foundation is quietly removed.

The direction of travel is clear. AI-driven consumption is not a feature being layered onto existing investment workflows, rather it is becoming the primary interface through which research reaches decision-makers. Research will increasingly not be opened, read, and interpreted at the pace of human attention. It will be queried, synthesized, and surfaced in moments. The systems that sit between content and answer will determine what survives of the model the industry has built.

What gives the sell side genuine leverage in this moment is something that is easy to underestimate: AI systems require a constant, high-quality supply of domain-specific content to remain relevant and accurate. Differentiated research, both current and historical, is foundational to how these systems perform. The buy side depends on what the sell side produces. That dependency creates a real opportunity to shape the terms of this transition, but the window to do so is narrow. The infrastructure decisions being made today inside buy-side technology teams will, over time, become the defaults that are very difficult to revisit.

A New Infrastructure Layer

What is required is not another distribution channel, but a new infrastructure layer, one designed from the outset for machine interaction rather than human reading, and that preserves the principles the research ecosystem depends on. Such a layer must ensure that content is retrieved in a structured, machine-readable way, rather than extracted from documents that were never designed for extraction. It must ensure that access is governed by entitlements defined by the content owner, enforced at the point of query rather than the point of document delivery. Attribution must be embedded at the moment of creation and carried through to the output, so that derived answers remain traceable to their source. Provenance must be preserved as an auditable chain and become an enforceable record. And the feedback loop between consumption and producer must be rebuilt at the infrastructure level, so that research departments can once again see how their work travels, where it lands, and what value it carries in this new mode of consumption.

Without these elements, the industry does not transition to a new model on its own terms. It is absorbed into one defined by others.

The infrastructure to make this possible does not need to be invented. The foundations already exist in the platforms that sit at the intersection of research creation and distribution, in the standards bodies capable of formalizing attribution frameworks, and in the commercial relationships that can be extended rather than reinvented. What is needed is the will to define the model before it is defined by default.

BlueMatrix has spent years building the infrastructure through which sell-side research is created, structured, and delivered to the buy side. Sitting at the intersection of production and distribution, this position is the natural place to build a governance layer. One that carries existing entitlement, attribution, and measurement capabilities into the AI consumption layer while preserving the relationships and commercial frameworks the sell side has established with its buyside clients. This is not work BlueMatrix can or should define alone. The standards governing how research enters AI systems must be shaped collaboratively, by those who understand the research business most deeply, and we are committed to building in that spirit.

The next phase of this market will not be defined solely by who produces the best research. It will be defined by who ensures that research is seen, attributed, governed, and measured even in a world where it is no longer read in the traditional sense.

The answer is already being generated, in systems and at a scale that will only continue to grow. The question of ownership, however, remains open. We believe it should be answered by the institutions that did the work, and we are committed to building the infrastructure that makes that possible.