Why an Investment Group Runs Its Own AI Lab

AI in asset management is the use of machine learning and, increasingly, agentic systems to support investment research, portfolio construction, risk management and operations across an investment firm. That is the working definition the field has settled on, and AI in investment management covers a familiar scaffold: research, portfolio optimisation, risk, operational efficiency and client personalisation. Each is a place where models read more, test more and reconcile more than a team could by hand.

One disambiguation matters first. Investment asset management is not enterprise asset management. The first manages capital, securities and portfolios; the second manages physical equipment, where AI handles predictive maintenance and sensor analysis on plant and infrastructure. The two share three words and nothing else, and much of the material that ranks for the phrase describes the industrial meaning rather than the management of investment portfolios.

What the standard account does well is describe the use cases. Where it stops short is the harder question underneath them: almost every treatment asserts that AI matters and lists what it could do, and almost none address the distance between having a model and operating one. That distance is the subject of the rest of this page.

Appetite is everywhere. Grant Thornton's 2025 survey of roughly 500 firms globally found AI adoption broad in intent across the industry (Grant Thornton / ThoughtLab, 2025); McKinsey has estimated that AI and agentic adoption could improve the asset-management cost base by approximately 25% to 40% over time (McKinsey, 2025), a figure that describes potential rather than a measured result; and J.P. Morgan Asset Management has estimated that AI could add roughly 1.4% to 2.7% per year to developed-market productivity over the coming decade (J.P. Morgan Asset Management, 2025), a forward-looking, economy-wide estimate rather than a realised asset-management result. Read together, those numbers tell one story: the ambition is near-universal, the realised gain so far is a fraction of the potential, and the distance between the two is exactly the gap between having AI and operating it.

That gap is structural, not a matter of effort. Licensing a model gives a firm a feature; getting that model to run as a system is a separate piece of work entirely. A capable model off the shelf does very little inside an investment firm on its own. To run, it needs clean, connected proprietary data; a place inside the firm's actual research and decision workflow rather than alongside it; human oversight and model-risk governance with the authority to overrule the machine; and the integration engineering that turns a demonstration into a system the firm relies on daily. Assemble those and AI operates. Leave any out and the firm owns a capability it cannot run. This is the same thesis the cluster develops elsewhere, from agentic AI as an operating system to the wider view of AI in investing.

Why "use a vendor's model" is not the same as running an investment operating system

A firm that uses a vendor's model has reached for a tool. A firm that operates an investment operating system has rebuilt part of how it works around that tool, with the data, workflow, oversight and integration to match. Industry research has repeatedly found that only a small minority of organisations have scaled an AI agent into genuine production in any function; the ambition is shared, the operating capability is not. That rebuild is measured in years of engineering, not a procurement decision.

IMS Group's answer to the gap was to build the operating layer in-house through IMS Labs rather than license it, and what that lab runs is already public. Cortex is a 41-agent underwriting engine coordinated by an Agent Mesh, built as the intelligence layer for private credit; its operating result, stated plainly on the firm's own sites, is that two-week diligence cycles compress to under two hours. Three platforms run in production: Cortex for underwriting intelligence, Onyx, the patent-pending truth layer that authenticates tokenised assets with digital twins for real-world assets, and Nimbus for agentic commerce.

Control over that operating layer is the reason a private markets group builds rather than buys. The native work of private capital, sourcing off-market deals, underwriting them, and monitoring illiquid holdings that do not reprice on a public tape, benefits from owned infrastructure rather than a generic tool. A private transaction carries weeks of document review where a public trade carries a screen and a price, exactly the kind of work an agentic system compresses, and the kind a firm wants to own end to end. This private-markets lens is largely absent from a published field written for and about public-markets managers; the firm develops it in where AI actually changes underwriting. Owning the build is what lets a firm speak to that work as something it runs rather than something it plans, and the proof sits one level down, with IMS Labs and its external site, imslabs.ai.

"AI washing" is the marketing of AI capability a firm does not genuinely have, or the material overstatement of what it does, and it is common enough that the term now appears in the industry's own list of challenges and in the attention of financial regulators. The useful response is not cynicism but a short, reusable test an allocator can apply before taking any AI claim at face value.

Three questions do most of the work. Is there something in production, running today in a named workflow with an observable effect, rather than an exploration or a pilot? Is there human oversight and governance, with a person who reviews, validates and can overrule the model, and the model-risk controls to support it? And can the firm describe the operating layer itself, the data, workflow and integration, rather than only the ambition around it? A firm that operates AI can answer all three; a firm that washes can answer none. This sits alongside the governance stance taken by bodies such as the CFA Institute, whose work on AI in asset management stresses transparency, oversight and the management of model risk as the conditions for responsible use (CFA Institute, 2025). The test names no competitor and compares no fees; the questions are enough, and a firm that runs its own lab is on the right side of them by evidence.

Operating AI is a build, not a badge. That is why IMS Group runs its own lab rather than licensing the capability, and why the firm can describe what runs in production rather than what it hopes to do. The proof, not the claim, is the point: explore IMS Labs for what the operating layer actually does.

This article is for information only and does not constitute investment advice or an offer of any IMS Group product or service. Forward-looking statements are projections, not guarantees. Figures are sourced and dated where cited; AI-adoption and productivity figures are point-in-time and should be re-verified on the publication date.

This article is provided by IMS Group for information purposes only. It does not constitute investment advice, an offer, or a solicitation. Figures are point-in-time and projections are estimates.