“I asked both tools the same question, then tested each answer against the other. When I put the terminal’s answer into LinqAlpha, it came back with six corrections, each backed by real data. When I put LinqAlpha’s answer into the terminal, it simply accepted the corrections. So now I use the terminal for data retrieval, and will always use LinqAlpha for the reasoning.”
Kenneth Yap, Fund Manager, IQ EQ
Most AI research tools compete on how much they can retrieve. LinqAlpha is built on a different premise: for an investor, the scarce resource is not information but judgment - so the platform’s job is to reason over the data, defend its conclusions when challenged, and organise the user’s own thinking rather than bury it. Reasoning quality is an architectural choice: an orchestration layer over the models, evidence attributed to source on every claim, and outputs that follow the investor’s framework instead of a generic template.
Kenneth Yap, who is a Portfolio Manager for a global equities fund managed by IQ EQ Fund Management Singapore, put that premise to a harder test than most. As an experienced AI user with a full existing stack, he spent over a month running LinqAlpha against the tools he already paid for - adversarially, on investment research.
Key Highlights
A head-to-head test designed by the Portfolio Manager: Kenneth asked the same questions across two systems and fed each answer back into the other. LinqAlpha identified six specific, data-backed corrections to the competing answer; the other system accepted all six.
Research in his own investment framework: earnings results are distilled into ten takeaways, each scored from +10 to −10 and supported by an attributed quote or data point.
400+ of his own research files in one context: instead of maintaining separate company-by-company AI projects, Kenneth can work across his research library and portfolio from a single environment.
Earnings research on his terms: scored, sourced takeaways can be prepared for review by morning, rather than requiring a night spent following results blow by blow.
Challenge | What a Full AI Stack Still Couldn’t Do
Kenneth already had a sophisticated AI stack - access to the major general models and a terminal AI assistant. Two gaps remained - and both are structural to general-purpose AI, not fixable with more subscriptions.
The first was fragmentation. A general chat tool forces one project per company, scattering research across mailboxes, drives and chat threads with nothing holding the whole book in view. In Kenneth’s words: “everyone is a separate analyst without there being a portfolio manager.”
The second problem was direction. Generic AI pulls investors toward consuming ever more information, when the job requires the opposite - fewer, better-organised inputs and room for judgment. As he puts it:
“We actually don’t need any more data, we actually need time away from data to think about things.”
The challenge, then, was not giving a Portfolio Manager more AI-generated research. It was building an AI research environment that could help determine what mattered — and withstand scrutiny when its conclusions were challenged.
Solution | Reasoning, Frameworks and Context in One Place
1. Reasoning that defends itself
For investment research, a plausible answer is not enough. Investors need to know why a conclusion was reached and whether it holds up when challenged.
LinqAlpha’s orchestration layer coordinates research across models and sources, grounding conclusions in attributed evidence rather than relying on a single model response. Kenneth’s head-to-head test made that difference tangible: when presented with the terminal’s answer, LinqAlpha identified six corrections and backed each one with real data.
2. The investor’s framework, not a template
Kenneth did not want generic earnings commentary. His own framework asks for ten key takeaways, each scored from +10 to −10 and supported by an attributed management quote or data point.
Saved as a personal workflow in LinqAlpha, the same methodology can be applied consistently across results. During earnings season, Kenneth can review scored, sourced takeaways in the morning rather than follow every result blow by blow - leaving more time to decide what those developments actually mean for the portfolio.
3. One context for the whole book
More than 400 of Kenneth’s own research files are connected to LinqAlpha alongside the platform’s market information, creating a unified knowledge layer across his research.
Instead of maintaining isolated context for each company, Kenneth can draw on his own research and external information across the portfolio - guided by his watchlist, time horizon, and investment frameworks.
The goal is not simply to make hundreds of documents searchable. It is to ensure that each new question can build on the work and context already accumulated.
Outcome | A Different Division of Labour Between Investor and AI
After more than a month of testing, LinqAlpha became the starting point for more of Kenneth’s day-to-day research. Other tools did not disappear. Instead, their roles became clearer.
The terminal remains valuable for granular, as-reported line items and data retrieval. LinqAlpha is used where Kenneth needs to synthesise information, challenge an interpretation, apply his own framework, and reason across a broader body of research.
That division of labour was not prescribed by LinqAlpha. It emerged from a test Kenneth designed himself.
And it points to a broader question about how investors should evaluate AI research tools. The goal may not be to find the system that can surface the greatest amount of information. It may be to find the system that helps the investor spend less time consuming it.
“We actually don’t need any more data, we actually need time away from data to think about things.”
Kenneth Yap, Fund Manager, IQ EQ
For Kenneth, that is increasingly the role of AI in the investment process: not replacing judgment, but creating more room for it.
Looking Ahead | A Platform That Learns How Its User Thinks
The next step is deeper personalisation.
Kenneth’s investment frameworks can be saved and reused so recurring analysis follows the same methodology each time. His accumulated research can provide context for future questions rather than remaining scattered across individual projects and files.
Over time, that creates something more useful than another source of AI-generated content. It creates a research environment increasingly shaped around how the investor already works: what he follows, what he considers important, how he evaluates new information, and how previous work informs the next decision.