Executive Summary
AI has commoditised surface-level market analysis. Any investment professional with access to a capable AI research tool can produce, in minutes, a competent summary of the consensus view on almost any traded asset, sector, or macro theme. The marginal cost of a baseline market opinion is approaching zero.
This commoditisation is not a threat to investment research. It is a clarification of what investment research actually is. The value was never in the baseline opinion. It was always in the depth behind it — in the proprietary channel check, the forensic accounting analysis, the expert network call, the decades of sector expertise distilled into a single paragraph of differentiated view.
This paper argues that the AI era is not reducing the value of high-quality, independent, authored investment research. It is increasing it — by making everything that AI can replicate worthless, and exposing everything it cannot as irreplaceable.
1. What AI Has Made Worthless
The research activities that AI has commoditised share a common characteristic: they involve synthesis of broadly available information into a coherent summary. Market briefings. Sector overviews. Earnings recap documents. Macro theme summaries. Consensus compilation across publicly distributed sell-side notes.
These activities were never the heart of alpha generation. They were the scaffolding — the context-setting that allowed analysts to spend their limited time on the differentiated work that actually mattered. AI has automated the scaffolding. This is a genuine efficiency gain.
The problem is that some teams — and some research budgets — were structured around producing and consuming scaffolding. Research providers that offered breadth without depth have found their value proposition challenged.
When AI can summarise what everyone thinks in thirty seconds, the question that matters is: what does your research process tell you that AI cannot?
2. What AI Cannot Replicate
The research activities that AI cannot replicate also share a common characteristic: they require a human being with specific expertise, relationships, access, and accountability to do original work that produces differentiated insight.
Primary Research
Channel checks, expert network conversations, management meetings, customer surveys, proprietary data collection: none of these activities can be delegated to an AI tool. They require a person with the right relationships, the right questions, and the analytical judgement to interpret what they hear in the context of a specific investment thesis.
Deep Sector Expertise
There is a class of investment research that is essentially the written output of a person who has spent fifteen years understanding a single industry deeply — its structure, its cycles, its key players, its inflection points. This research does not summarise publicly available information. It contextualises it against a model of how the industry actually works that cannot be easily reconstructed from text data.
Forensic Analysis
Accounting forensics, supply chain analysis, patent citation mapping, satellite data interpretation: these are research activities that require specialist skills applied to structured data in ways that produce insights not visible in narrative research. The output is not a summary. It is a finding.
Author Accountability
A named analyst who has published a differentiated view takes professional accountability for that view. If they are wrong, there is a professional and reputational consequence. This accountability shapes both the quality and the honesty of the analysis. AI-generated synthesis carries no such accountability.
3. The Provenance Problem in AI Research
One of the least-discussed risks in AI-assisted research is the erosion of provenance. When an AI tool synthesises across multiple sources and produces a consensus view, the individual contribution of each source is obscured.
Loss of Signal
The analyst who is consistently early, consistently right, and consistently differentiated in a particular sector is a signal. Their research deserves more weight than the consensus. In a synthesised view, their contribution is diluted by every other analyst who covers the sector. AI synthesis is, in this sense, a signal-destroying machine.
Loss of Challenge
The specific analyst who consistently argues against the consensus — who has been making the bear case on a crowded long for eighteen months and documenting their reasoning carefully — is a resource. In a synthesised consensus view, they may not appear at all.
Evidence-bounded research preserves provenance. Each source retains its identity, its position, and its track record. The buy-side analyst can see not just what the consensus is, but who holds it, how strongly, and who dissents.
4. Constructing a High-Signal Research Portfolio
For investment professionals thinking about how to allocate their research attention and budget in an AI era, the strategic logic points in a consistent direction: reduce investment in research that AI can replicate, and concentrate on research that AI cannot.
Is this provider doing primary work?
Does the research reflect direct channel access, proprietary data, expert networks, or primary analytical work — or is it a synthesis of publicly available information? Primary work is irreplaceable. Synthesis is increasingly commoditised.
Does this provider have a genuine specialist edge?
Is the analyst’s sector expertise deep and long-tenured enough to produce insights not visible in the general research landscape? Breadth without depth is exactly what AI does well. Depth without breadth is what AI cannot replicate.
Does this research have an explicit point of view?
Research that hedges every conclusion and presents balanced perspectives is not differentiated analysis. It is editorial. The research that earns its place in a high-signal portfolio takes a position, documents the reasoning, and commits to a conclusion that can be tested.
Is the provenance preserved?
Can the research be accessed and interrogated at the claim level — so that specific observations can be extracted, cited, and tested against the investment thesis — or does it exist only as a synthesised narrative that cannot be decomposed?
The question is not whether a research provider is better than AI at summarising consensus views. That bar no longer matters. The question is whether they produce something AI fundamentally cannot.
5. The Marketplace Model and the Future of Research Distribution
The structural logic of AI-era research distribution points toward a marketplace model — a curated, licensed environment where specialist research providers make their differentiated content available directly to buy-side institutions, with appropriate entitlement controls and provenance preservation.
This is different from the traditional sell-side distribution model in several important respects. It is not relationship-driven in the same way — discovery is based on relevance and track record, not on the breadth of the broker’s coverage universe. It is explicitly licensed for AI use. And it preserves author identity and accountability.
Conclusion
The commoditisation of surface-level analysis is not a threat to investment research. It is an opportunity — for buy-side professionals who understand what AI has changed, and what it has not.
What it has changed: the value of aggregation and synthesis. What it has not changed: the value of primary research, deep expertise, author accountability, and differentiated point of view.
The Contours Marketplace is built around this logic. Every provider has been selected for genuine analytical differentiation — primary research capability, sector depth, and content explicitly licensed for AI-assisted investment workflows.
Contours is a product of KiteEdge Ltd. To explore the Contours Marketplace and its licensed research providers, contact us.
