Happal

A market figure appears on a slide during a strategy meeting. It looks precise and has already found its way into the recommendation. Then someone asks a familiar question: “Where did that number come from?”

The figure may have originated in a market report, an analyst presentation or a spreadsheet prepared several months earlier. Finding the original source is only the beginning. The team still needs to know which year the estimate covers, how the market was defined, what was included and which assumptions shaped the forecast.

Anyone who works with market research will recognize the situation. AI has made the initial search for information much faster, but the usual questions about definitions, context and evidence remain once that information enters a business decision.

Getting the first answer is becoming easier

Modern AI assistants are useful at the beginning of a research project. They can explain unfamiliar terminology, summarize a complex subject and help an analyst work out which questions deserve further investigation.

This makes it easier to enter an unfamiliar market. An analyst looking at antibody-drug conjugates, for example, can quickly understand the basic science, the main parts of the value chain and the language used by companies in the field. A strategy team can explore possible market-entry questions before deciding where a more detailed study is required.

Previously, much of the first day of a project could be spent opening browser tabs, reading introductory material and building a basic vocabulary. AI can compress a good part of that work into a short conversation.

Later stages require more care. An explanation that is perfectly adequate for initial orientation may not be ready for a board presentation, an investment review or a commercial forecast. Once a figure starts influencing a decision, its origin and meaning become part of the analysis.

When someone asks where the number came from

Language models are designed to produce fluent responses. A clear and confident answer can still combine figures from different years, use a broader market definition than the user intended or repeat information whose original source is unclear.

Even an accurate number may be misleading when it is separated from its context.

Consider a forecast for a pharmaceutical manufacturing market. One study may include both captive and contract manufacturing, while another counts only outsourced services. A third may measure manufacturing revenue, whereas the user is interested in production volume. Each estimate can be reasonable within its own definition, yet the figures should not be compared without understanding what they represent.

The same issue arises with company counts, pipeline numbers and regional market shares. A list of developers may include companies with early research programs, while another may count only those with clinical-stage assets. Neither method is necessarily wrong. The definition determines what the number can be used for.

For informal exploration, a plausible summary may be enough. In work that will be reviewed by other people, the analyst needs to know which source was used and whether it supports the statement being made. That is often the difference between a figure that belongs in working notes and one that can be placed in front of a decision-making team.

Bringing the source closer to the answer

One practical approach is to connect AI to a defined collection of specialist research.

From the user’s perspective, the process can be understood as Ask → Find → Answer → Verify. A person asks a normal business question. The system searches the research collection for relevant material and uses it to prepare a response. The reader can then examine the source supporting the main information.

This keeps the answer close to the material from which it was prepared. If the available research covers North America but says little about Europe, that limitation can be made clear. If two studies use different market definitions, the reader can review both rather than receiving a single figure with no explanation.

The usefulness of this approach depends on several ordinary research disciplines. The underlying material needs to be credible and current enough for the question. The search needs to find the relevant sections, and the response needs to represent them faithfully. When the available research cannot support a confident answer, the system should make that clear.
Technical explanations of these systems often focus on models and databases. For most research users, the practical test is simpler. They need to see which material informed the answer and be able to check it without beginning a separate investigation.

Take a simple market-size question

Suppose a strategy team asks: “What is the expected size of the gene therapy market by 2035?”

Using a traditional research process, an analyst would locate relevant reports, search for the forecast and check the publication date, market definition and assumptions. If several estimates were available, the analyst might compare them before choosing the one that best fits the team’s purpose.
A general AI assistant can provide an answer almost immediately. That is useful for understanding the likely range, common growth drivers and areas that need closer examination. The user may still need to locate the original forecast before using the number formally.

With an AI system connected to a research collection, the same question can lead directly to the relevant material. The response can identify the study behind the forecast, allowing the user to review the definition and surrounding analysis while considering the answer.

Much of the time saving comes from finding the appropriate evidence sooner. The normal checks around scope, assumptions and comparability still apply, but the analyst spends less time searching through long documents for the relevant passage.

Making a research library easier to question

This type of access is offered by LabKairos™(https://www.labkairos.com/?utm_source=guestpost), a market intelligence platform developed by Roots Analysis.

Users can ask questions of the Roots Analysis research library in natural language. The underlying research has been developed over many years and covers pharma, biotech, healthcare and related areas. Responses are intended to connect users with the relevant Roots Analysis research so they can examine the material behind the information provided.

For someone asking about a market forecast, the experience is more direct than receiving a figure and then searching separately for the report from which it may have come. The user can review the study associated with the answer and decide whether its scope fits the question being considered.

The same form of access can help when reviewing a competitive landscape, comparing CDMO or manufacturing capabilities, examining company pipelines or preparing an initial market-entry assessment. These tasks still require judgement. Company descriptions may need updating, market definitions may differ between studies and a forecast may need to be interpreted in light of recent events. Faster access gives the analyst more time to work through those questions.

Why the detailed report still matters

A short AI response cannot always carry everything contained in a well-prepared market research report. Detailed studies document methodology, segmentation, definitions, assumptions and the wider analysis surrounding a finding. That context often explains why a number differs from another estimate or where its limitations lie.

Most readers, however, do not need to work through every page whenever they have a specific question. They may be looking for one forecast, a group of companies or the section dealing with a particular technology.

A searchable research interface helps the reader reach that part of the report more quickly. If the question develops, the complete study remains available for deeper examination. This is particularly useful when the answer depends on several sections rather than one isolated table or paragraph.

Research reports therefore continue to provide the documented body of evidence. AI can make that evidence easier to locate and use during everyday work.

What research users should expect from AI

As AI-generated answers become routine, professional users are likely to examine the supporting evidence more closely. Fluency and speed are useful, but neither tells the reader whether a market has been defined correctly or whether a forecast is suitable for the decision at hand.

Research tools should make it straightforward to see what informed an answer, understand the limits of the available material and return to the original analysis when more detail is needed.

For anyone relying on market information, that should become a normal expectation rather than an additional verification exercise.