Assumptive Understanding Model

Information and understanding are not the same thing. A system may retrieve many correct facts about a company without establishing which facts matter, how they relate, what they imply together, or what remains unknown. Knowing that a company is hiring implementation engineers, releasing enterprise features, and publishing new enterprise customer cases gives us information. Understanding requires determining whether those observations are related and what, if anything, they reasonably support about the company's current state.

This distinction matters because the questions ASUME is designed to help answer usually extend beyond retrieval. A user may want to know what a company appears to need, which capabilities it is building, what it is prioritising, where a constraint may exist, or why its behaviour appears to be changing. These questions cannot always be answered by locating a sentence in a source. They require relevant information to be considered together and interpreted without hiding the assumptions or uncertainty involved.

AUM exists to support that transition. At the simplest level, company information enters ASUME and contributes to a maintained understanding of the company. The important output of this process is not a collection of pages or isolated facts, but a structured state that later questions and objectives can use.

AUM separates different parts of this process because they do not have the same status. Evidence is information that bears on something ASUME is trying to establish or evaluate about a company. Assumptions make visible where the available evidence does not directly establish something required by the reasoning. Inference is how ASUME reasons from what is available toward a statement that was not directly observed. Together, these contribute to the company's maintained understanding.

This distinction prevents a common problem in AI-generated analysis: allowing the boundary between what a company actually stated and what a model concluded to disappear. If a company has opened twelve cybersecurity vacancies, that hiring activity may be directly observable. A statement that the company is expanding its cybersecurity capability is different. It may be well supported, but it remains an inference rather than something the company necessarily said itself.

AUM keeps those differences visible. This makes it possible to inspect what supports a conclusion, identify where an assumption was required, and understand why two conclusions may deserve different levels of confidence even when both sound equally fluent when expressed in natural language.

ASUME can observe information about a company, but it cannot directly observe the company's complete internal state. Available information may be incomplete, outdated, contradictory, limited to one part of the organisation, or compatible with more than one explanation. AUM is designed around this limitation rather than assuming that enough information will always remove it.

As a result, the company's understanding does not have to force every question into a definitive answer. Some conclusions may be strongly supported, while others may depend on assumptions or remain qualified. Different sources may point in different directions, and there may be cases where the available information simply does not justify a meaningful conclusion. These differences remain part of the understanding instead of being flattened into uniformly confident outputs.

This is particularly important when ASUME reasons about properties such as needs, priorities, constraints, capabilities, and direction. These may leave observable signals without being explicitly stated. AUM allows ASUME to reason about them while preserving the fact that an inferred property is not the same thing as a directly observed fact.

ASUME does not need to reconstruct a company from the beginning every time a new question is asked. AUM allows relevant evidence, assumptions, conclusions, context, and changes to remain connected so that later reasoning can begin from an existing company understanding rather than from an empty state. This is what ASUME refers to as maintained understanding.

Maintained does not mean fixed. Companies change, and the information available about them changes as well. New hiring may begin or stop, products may be launched or discontinued, priorities may shift, documentation may change, and new internal information may provide context that was previously unavailable. When the basis of an existing conclusion changes, AUM can reconsider the relevant part of the understanding rather than simply adding a new answer beside the old one.

This gives ASUME continuity without requiring permanence. The understanding can persist across questions and users while remaining open to revision whenever new evidence materially changes what should be represented about the company.

AUM separates understanding a company from deciding what someone wants to know or do about that company. This allows the same underlying company understanding to support different objectives without requiring ASUME to reconstruct the company separately for each one.

A Revenue objective may focus on commercial opportunities, buying conditions, or fit. A Partnerships objective may care about complementary capabilities or strategic alignment. Another objective may concern research, risk, negotiation, investment, or a question defined directly by the user. The objective changes which parts of the understanding matter for the task; it does not change the company being understood.

The relationship can therefore be thought of as company information → AUM → company understanding → objective → output. An answer, recommendation, match, ranking, explanation, or other result is an application of the maintained understanding to a particular objective. It is not the understanding itself.

This separation is important to the broader direction of ASUME. Products and workflows can change, and new objectives can be introduced, while the underlying company understanding remains reusable across them.

AUM does not assume that every question has enough evidence for a strong answer. Sometimes the available information is incomplete. Sometimes several explanations fit the same observations. Sometimes an important assumption cannot yet be resolved, or a newer source contradicts an older one.

In those situations, the appropriate result is not necessarily to choose the most plausible explanation and express it with confidence. The understanding can instead remain qualified, preserve competing possibilities, retain an explicit assumption, or show that an evidence gap remains. This is the underlying principle behind ASUME's commitment to never hide the gaps.

Keeping uncertainty visible also makes revision possible. If ASUME knows which part of a conclusion depends on weak evidence or an unresolved assumption, later information can change that part of the understanding without treating the entire previous state as equally uncertain.

AUM is designed so that users do not have to accept a conclusion solely because ASUME produced it. Relevant conclusions can remain connected to the evidence, assumptions, context, and uncertainty that affect them, allowing the basis of an output to be inspected and challenged.

For a user, the useful questions are practical ones: what supports this conclusion, where did that information come from, was this stated by the company or inferred by ASUME, which assumptions matter, what remains uncertain, and what changed when the understanding was updated? The purpose of preserving this structure is to make those questions answerable.

This is the foundation for ASUME's approach to traceability. Traceability does not mean exposing hidden chain-of-thought, internal prompts, model weights, proprietary scoring methods, or every computational operation performed by the system. It means preserving enough of the basis of a conclusion that a user can understand where it came from and why it may change.

AUM provides the common structure behind the different ways people and agents can use ASUME. Evidence can be gathered once and contribute to a company understanding that persists across later questions. Assumptions can remain visible rather than being silently converted into facts. Inferences can contribute to the understanding without becoming indistinguishable from direct observations. New information can revise what was previously understood.

The individual pages in this section explain those parts in more detail: Evidence explains what supports a statement and how gaps are handled; Assumptions explains where reasoning depends on something not directly established; Inference explains the distinction between observation and conclusion; Maintained understanding explains how company state persists and changes; Traceability explains how the basis of outputs can be inspected.

AUM is the structure that connects those ideas. It allows ASUME to move from company information toward a maintained understanding that can subsequently be used for different objectives and outputs.

AUM does not claim that ASUME can observe the complete state of a company, and it does not claim that every inference will be correct. The quality of an understanding remains bounded by the information available, the relevance and quality of that information, the assumptions required, and the uncertainty inherent in reasoning about something that cannot be observed completely.

AUM also does not turn an assumption into evidence simply because it is useful, treat a plausible explanation as proof, or imply that generated output is equivalent to objective truth. Nor does the term understanding imply consciousness, subjective comprehension, or human-equivalent cognition.

For ASUME, understanding is computational and revisable. It is a structured representation of what the available information can currently support or reasonably imply about a company, maintained in a form that can be used, inspected, and updated as that basis changes.