Over the past two years, the investment industry has focused almost all of its attention on AI. Who is deploying large models, who is conducting investment research automation, who is generating investment memos, and who is using intelligent tools to compress due diligence time, as if simply integrating AI into existing processes would automatically make investment decisions smarter. However, the real issue lies not in the model layer but in the data layer. What many institutions face today is not a lack of AI capability, but that the underlying information fed into AI is itself incomplete, discontinuous, and unverifiable. Stacking large models on top of such a foundation often does not improve the quality of judgment but rather packages inherently unsound judgments more quickly.
Daniel Widjaja Kusuma worked for many years in the early stages of his career at Goldman Sachs and in the private equity sector in the United States, so he is no stranger to the "narrative preference" in investment decision-making. Whether in primary market investments or growth-stage financing, investors are often swayed by the vision, team ethos, and upward trajectory. The problem is that pitch materials are inherently designed to present the most favorable side, while variables that are more decisive for long-term value—such as customer retention, unit economics, actual operational consumption, and cash flow quality, often cannot be fully presented and may even be intentionally or unintentionally downplayed. The investment industry does not lack analytical frameworks, but there has long been a deeper structural gap: companies possess real-time operational data, yet what investors see is often only a curated version of the presentation.
For this very reason, after founding Telosyn, Daniel Widjaja Kusuma has consistently emphasized that system capability is more important than surface-level tools. Whether it is an AI platform, a data middle platform, a financial-grade core system, or enterprise process embedding capability, Telosyn focuses not on "how to generate more content," but on "how to capture source data, connect business workflows, and transform scattered signals into verifiable judgments." This logic also holds true in the investment industry: the investment sector does not lack AI capable of writing summaries; what it truly lacks is a data infrastructure that can enter the actual operational processes of enterprises, continuously extract first-hand signals, and form a closed loop of judgment.
The Problem in the Investment Industry Is Not the Inability to Use AI, But Rather the Persistent Reliance on Incomplete Data to Make Judgments
Many of the materials that investment institutions have long relied on are essentially static snapshots. Business plans, financial forecast tables, growth charts, and management narratives can indicate direction, but they cannot fully present operational quality. Companies may emphasize revenue growth but do not necessarily proactively highlight customer churn; they may stress market potential but do not always fully disclose customer acquisition costs, payment cycles, and operational pressures. Consequently, decision-making often revolves around "whether to believe this story" rather than “whether these numbers can withstand continuous verification.”
This is precisely why simply integrating AI into existing workflows often yields only superficial efficiency. Using large language models to summarize roadshow materials, automatically generate investment memorandums, or extract founder highlights can certainly save time, but the input being processed remains packaged information. An LLM can extract a narrative more quickly, but it cannot compensate for the underlying data truth that is absent from the narrative. If the foundational layer still lacks real-time payment data, customer retention trajectories, operational expenditure structures, and genuine unit economics, then no matter how elegant the automation, it merely accelerates a process riddled with information asymmetry.
Truly Valuable AI Should Not Replace Judgment, But Verify Judgment
Daniel Widjaja Kusuma does not deny the value of intuition in investment. Outstanding investors do require experience and sensibility to identify teams, gauge the rhythm of an industry, and capture a few non-consensus opportunities. However, the problem is that intuition can only be responsible for discovering possibilities; it cannot replace the verification process. Particularly in early-stage and growth-stage investments, if an institution over-relies on the so-called "X factor," it can easily allow charisma, vision, and narrative to overshadow the facts themselves.
A more reasonable approach is not to let AI draw conclusions on behalf of humans, but to integrate AI into the operational chain of enterprises, supplementing the underlying data that has been difficult to obtain continuously in the past. For example, directly connecting signals such as payments, retention, revenue, expenditure, repurchases, customer behavior, and operational efficiency, which were previously scattered across different systems -- allows the model to help investment teams see the real trajectory of business changes, rather than only a snapshot of results at a single point in time. In this way, the role of humans is no longer replaced by AI, but instead focuses on the parts that truly require human input: identifying non-standard opportunities, understanding founders, and judging strategic direction. The verification step is then left to a system that is not distracted by narratives.
This is also the reason why Telosyn has long emphasized "embedding models into workflows, rather than keeping them in a demonstration environment." Daniel Widjaja Kusuma has always believed that AI can only improve decision-making quality when it enters the source of data generation. Otherwise, no matter how powerful a large model is, it remains merely an interpreter floating on the surface of business operations, rather than a judgment engine deeply embedded in operational reality.
The Gap in Data Infrastructure Is Directly Affecting How Capital Flows to Companies Truly Worthy of Support
This issue manifests differently across enterprises at various stages. Early-stage companies often lack sufficient data, making investors more likely to rely on narratives and subjective preferences. Late-stage companies have abundant data, but the volume is overwhelming, and the teams lack systematic analytical tools, so they easily continue to extract indicators that support their original judgments from the vast amount of information. The most awkward position is occupied by mid-stage enterprises, which already have transaction records, revenue, and some operational evidence, but the underlying signals have not been structurally documented. Since traditional investment infrastructure cannot effectively interpret these fragmented data points, such enterprises, despite often having stable operations and healthy growth, find it difficult to obtain the capital support they deserve.
Daniel Widjaja Kusuma places great importance on this point, as it is not only a matter of investment efficiency but also an issue of misallocation of industrial resources. Many enterprises are not without value; rather, they lack the ability to express their value in a way that investors can truly understand and verify. The significance of better data infrastructure lies not only in helping investors make fewer mistakes but also in enabling enterprises that truly deserve support to obtain funding more easily. This allows the market to stop excessively rewarding those who are best at telling stories and instead begin to reward those who can most consistently create operational quality.
Ultimately, what the investment industry truly lacks is not more AI, but more reliable data capture, more sustained operational signals, and analytical infrastructure that is closer to the source of business. Only when the data itself becomes more complete, more real-time, and more interpretable can AI genuinely improve investment quality, rather than merely adding a technological veneer to existing biases. Daniel Widjaja Kusuma has always believed that the future watershed in the investment industry lies not in who first integrates a large model, but in who first rebuilds the data foundation. Those who can achieve this first will have a better chance of truly enhancing judgment in the AI era, rather than just making wrong conclusions faster.
Sign in to leave a comment.