inLABs – research projects

CDT

Research on expert knowledge continuity

CDT (Cognitive Digital Twin) is a concept developed at InLabs for organizing knowledge related to the tasks and everyday context of a specific person's work. We check how such knowledge, along with its justification and current information, can be collected, organized and used responsibly. The goal is to support continuity of work, not to recreate a person's identity or to take over their decisions and responsibilities.

Technology is not intended to imitate humans or present uncertain information as certain. First of all, we want to honestly separate what can be described and conveyed from knowledge that still requires experience, conversation and practice.

Why it arises

Important knowledge in an organization is usually dispersed among procedures, decisions, instructions, other documents, conversations and everyday practice. When roles, team composition, or priorities change, it is easy to lose track of not only what was done, but also the reasons for choosing a particular method, the conditions considered, and the precautions taken.

Merely collecting the files does not solve the problem. Information that is unrelated to a specific task, person, source and date of creation may be difficult to find, outdated or used inappropriately. In turn, too far-reaching generalization of an expert's experience may give a false sense of certainty, especially in unusual situations requiring many years of practice.

Therefore, we explore how to maintain the continuity of knowledge without reducing it to a list of facts. This includes, among others: consent to the use of information, access rules, ownership of knowledge, the possibility of correcting and deleting it, timeliness and responsible use. The system should clearly demonstrate its limitations, rather than replacing human judgment with apparent certainty.

Potential application scenarios

The examples below show possible research directions. They do not describe ready-made implementations, available functions or services.

Onboarding a successor into demanding tasks

Research scenario. CDT can help organize the information needed by a person taking responsibility for complex tasks: key sources, conditions, previous decisions and exceptions. Such material is intended to facilitate preparation for an interview with an expert and subsequent independent work, and not to automatically replace it.

Restore the decision context

Research Scenario. When the team returns to the case after some time, they may need information about what was known when the decision was made, what restrictions were in place at the time, and what needs to be checked again. CDT does not make decisions for the team, but helps prepare reliable material for evaluation.

Maintain knowledge when changing teams or roles

Research Scenario. Research can help identify the knowledge that is really needed to maintain work continuity and separate it from material that is outdated or impossible to apply to the new situation. This allows the delegation of tasks to be better prepared while maintaining the individual and organization's control over how the information is used.

Human cooperation with AI tools

Research scenario. An AI tool could only receive context that consent and permissions allow for its use. This would allow them to better understand the task without gaining access to all the resources or the right to act independently. This is a direction requiring further research, and not a description of the current integration with a specific system.

Role in the InLabs ecosystem

CDT concerns knowledge related to a specific person's tasks, memory and work context. In the broader InLabs portfolio, it can be considered alongside TWIN:DESK, as a work environment, and SAVANT-AI, which examines the knowledge of the entire organization. However, these projects have separate scopes. This description does not mean that they share an operating environment, data flow or out-of-the-box integration.

The limits of responsibility are clear. CDT does not manage organizational change, does not control access to models and does not assess whether a given person has the authority to approve decisions. This last issue is covered by InLabs' separate work on HCG. CDT is also not a tool for observing employees. It may provide needed context, but it does not make decisions and does not relieve a person of responsibility.

Research approach

We assume that useful knowledge must remain related to the task, context, source, date of creation and level of confidence. The system should help recognize not only facts and procedures, but also the conditions of their application, justifications and signals indicating that information needs to be reconfirmed.

We explore three basic possibilities: acquiring and organizing knowledge covered by consent, linking it to a specific task, and presenting it to a person who still makes decisions independently. The result is not intended to be a digital copy of the human being, but a clearly marked support tool that supports the understanding, testing and transfer of knowledge.

Not all knowledge can be easily described. Some of the experience can be recorded as facts and procedures, some of it can be extracted in conversation, and some of it is only learned through observation and practice in a specific situation. We do not promise to fully reproduce this last category. A fairer solution is to indicate when contact with an expert or practical preparation is needed.

Test methods and validation

The development of CDT requires checking both the quality of the accumulated knowledge and the principles of its use. We plan controlled research cases, expert reviews and voluntary studies involving people with different experience, conducted with appropriate data protection. We will compare ways of describing context, communicating uncertainty and preparing materials for the person taking over the tasks.

The assessment will include, among others: usefulness in a specific task, consistency with sources, timeliness, clarity of limitations and quality of information provided to a human for further evaluation. In parallel, we are examining the rules regarding consent, access, attribution of the author of knowledge, and its correction and deletion. Detailed procedures, configurations, data and results are not part of this public description.

CDT is still in the concept development stage. Current work focuses on clarifying research questions, the limits of the responsible use of knowledge and ways of checking the solution in real work conditions. We also analyze how to reconcile the utility of context with a person's right to decide about their own knowledge.

We are not announcing a product, pilot, early access program or commercialization date here. The scope and results of further work will depend on the collected evidence, risk assessment and decisions of the knowledge owners and the R&D team.

Let's talk about collaboration or research partnership

As part of the InLabs research program, allclouds.pl invites organizations, companies and research partners who want to have a real impact on the development of the presented concepts to cooperate. We offer flexible forms of cooperation, various levels of partnership and benefits related to participation in the creation of new technologies that will not be available to other customers on the same terms after the solution is introduced to the market.

You can contact us regarding research and development cooperation via the e-mail address InLabs@allclouds.pl.

HCG

When human participation in AI decisions is of real importance

HCG (Human Competence Gate) is a way to check whether, before approving an AI-supported decision, the right person understands the situation, is able to evaluate the recommendation and consciously takes responsibility for the decision made.

In many processes, the organization uses AI recommendations, but further action is still decided by humans. The methodological foundations of HCG have already been developed, and the corresponding control stage is described in the CDF implementation methodology. At InLabs, we are currently investigating how to apply such checking to specific processes and how to evaluate its effectiveness.

Why acceptance alone is not always enough

The presence of a human in the process does not determine the quality of supervision. When recommendations come frequently and their content is complex, approval can become a reflex. The decision maker may not have a complete picture of the rationale, exceptions or limits of application of the recommendation. In this case, the system has a formal approval trail, but the organization still does not know whether the decision was actually considered.

This is a particularly important problem where the decision affects the customer, employee, operational security, resources or process continuity. Speed ​​is valuable, but it cannot lead to a situation where a person signs a result without being able to understand what it means in a given case.

The second challenge is the difference between general knowledge and preparation for a specific decision. A competent person does not always have the current context of the case, and someone who knows the details of the case may need support in assessing the limitations of AI recommendations. Therefore, the organization needs an approach that does not reduce responsibility to the user's presence at the screen.

HCG answers the question of how to make a human decision have real meaning: before approving an action, it allows you to make sure that the person responsible understands the specific matter and has the competences needed to evaluate it.

It is important to maintain proportion: the scope of attention and support should correspond to the importance of a specific decision, and not create the same obstacle for each action.

Where HCG can make a difference

The following scenarios are possible directions of application and not a description of implementations or available features.

Decisions supported in purchasing processes

In an organization, AI recommendation can help organize the information needed to evaluate variants. The control point being examined would help the approver confirm that he or she understands the key assumptions, constraints, and consequences of the decision before taking action.

Changes with significant operational impact

When changing the configuration, priorities or method of implementing the process, the AI recommendation may require conscious confirmation by the appropriate person. Such a check could direct her attention to the context of the decision and indicate that additional explanation or expert involvement is needed.

Working with recommendations in complex cases

In matters requiring domain knowledge, AI can organize and summarize available information, but does not take responsibility for assessing it. HCG helps determine whether the decision maker recognizes significant limitations and knows when not to rely solely on AI recommendations.

Introducing AI into new roles and processes

In the initial period of working with AI, teams learn how to interpret its results and where automatic support ends. We explore whether the first decisions of a certain type require special attention, and how to design such reflection points without slowing down every action.

From a methodological perspective, approval of an action may require confirmation that the responsible person has the appropriate competences and understands the matter. However, HCG does not decide which decision is correct; it remains on the side of the person responsible for the process.

Role in the InLabs ecosystem

HCG plays a cross-cutting role: it concerns the moment when a human has to evaluate or approve an action supported by AI. It is not a separate application, training platform or HR system. Its scope includes the relationship between the recommendation, the context of the decision and the responsibility of the person authorized to act.

CDF and HCG complement each other but can be used independently. In CDF, issues of human competence are taken into account at the stage when the organization checks its readiness to consciously approve an AI-supported decision. A separate area of ​​InLabs' work is CDT, which concerns the continuity of knowledge about human work. The AICF, dedicated to operational compliance supervision, is also being developed independently. This description does not imply out-of-the-box integration, a common operating environment, or data flow between these solutions.

HCG does not replace organizational policies, risk assessment, access controls, expert knowledge or human decision-making; nor does it abolish the responsibility of the implementing organization. It is not used to profile employees individually. Its use does not require other elements of the ecosystem.

Principle of approach: from recommendation to informed decision

The starting point of the work is a simple hypothesis: formal acceptance has limited value if the person approving it does not understand what the decision is about and what its limits are. HCG examines how elements requiring conscious assessment can be presented before a decision is made, so that the person remains an actual participant in the process.

The approach can be described by three elements. The first is the context of the decision: what it concerns and what information is important. The second is understanding: whether the approver can address the assumptions, limitations and possible impacts. The third is responsible action: approval, consultation or resignation. This model refers to the requirements for human competence and supervision described in the EU AI Act and the ISO/IEC 42001 standard, but does not constitute a declaration of compliance or legal advice.

It is not about replacing human judgment with automatic evaluation. The goal is a way of organizing decisions that helps a person carefully consider a specific matter. HCG is therefore different from general training or a universal checklist: it takes into account the situation, the responsibility assigned to the role and the limitations of a given recommendation.

The approach must remain understandable to the person using it. The research will therefore focus not only on the moment of decision itself, but also on whether the presented context helps to notice uncertainty instead of seemingly removing it.

Research and validation

Further research on HCG may include analysis of various decision-making situations, expert reviews, and comparison of several ways of supporting a person before confirming action. We will evaluate, among others: clarity of the presented context, validity of additional checking, usefulness for the person making the decision and the possibility of clearly explaining the human role in the entire process.

The limits of this approach also need to be defined. Additional verification should improve the quality of the decision and not become an empty formality or excessive burden on the process. Therefore, research must take into account different roles, risk levels and actual operating conditions of the organization.

It will be important to assess whether the method of support leaves the decision-maker with a clear basis for his or her own judgment, rather than suggesting automatic certainty. Depending on the scenario, both qualitative analyzes and comparisons of process behavior before and after the introduction of a control point may be needed.

We do not present experimental results or efficacy claims here; We also do not refer to customer implementations. Assessment criteria mature with research.

The methodical part of HCG is worked out; its development and measurement remain the subject of research. This site does not announce product, pilot program, availability or commercial schedule.

Future work may focus on distinguishing decisions where a human point of control has real justification, ways of presenting the limitations of recommendations, and maintaining a clear line between AI support and human responsibility. Each future scenario requires a separate assessment of context, risk and organizational policies and a responsible decision holder.

Let's talk about collaboration or research partnership

As part of the InLabs research program, allclouds.pl invites organizations, companies and research partners who want to have a real impact on the development of the presented concepts to cooperate. We offer flexible forms of cooperation, various levels of partnership and benefits related to participation in the creation of new technologies that will not be available to other customers on the same terms after the solution is introduced to the market.

You can contact us regarding research and development cooperation via the e-mail address InLabs@allclouds.pl.

AICF

Towards operational compliance control for generative AI

AICF (AI Compliance Framework) is an InLabs research and development project for which we are applying for funding under the European Funds for a Modern Economy programme. We want to study how organizations can continuously supervise the compliance of generative AI use with regulations, standards and their own rules. We do not want to reduce compliance to a one-off checklist. We plan to examine how to describe organizational requirements clearly, relate them to a specific task and preserve the evidence later needed by the person responsible for assessment. Work will begin only after funding is awarded; the project cannot start without a positive assessment of the application.

Why this research matters

Generative AI increasingly supports business processes in which data, decisions and responsibilities change. A policy or assessment prepared before deployment is still needed, but it does not always explain what to do when the task, model, data source, tool or organizational conditions change. AICF is intended to focus precisely on this difference between a written requirement and the actual way AI is used.

Requirements are not static in practice

Organizations must take into account regulations, standards, contracts and their own rules. Each of these sources may change at a different pace and may have a different meaning for specific teams, processes and data. We want to study how to preserve the readability of these requirements when the way AI is used changes. We do not assume that one rule or one result will be sufficient in every situation.

Evidence is often fragmentary

Information confirming how AI operates is often scattered across systems and people. As a result, it is difficult to establish what was actually checked, which issues were considered important and what actions should be taken next. We plan to examine how to connect the context of use, compliance requirements and the record of decisions in a way that can later be verified. The goal will be better accountability, not replacing experts.

Control mechanisms must be understandable to people

Control mechanisms must help responsible people understand the situation and decide what to do next. A solution may be technically advanced, but if it cannot be understood, challenged or supervised, it will not solve the underlying problem. That is why transparency, the ability to reconstruct the decision path and human responsibility are as important to us as control itself.

Potential use cases

The following examples show possible research directions. They do not describe customer deployments, product commitments or current availability of the solution.

Internal work supported by AI

In a controlled workflow inside an organization, AICF could be used to study how requirements are taken into account in an AI-supported task. We also want to check whether such an assessment remains available for later verification by the responsible person. A possible result would be a clearer basis for discussion when the use case changes or raises doubts.

Regulated business processes

In processes shaped by more than one source of obligation, the research is intended to check whether a unified view of control supports qualified verification. The aim will not be automatic certification of a process or a binding legal assessment, but better visibility of relevant issues and limitations.

Change in policy or operational context

When a policy, operating principle or material condition changes, the research scenario involves indicating the places where the change may require renewed attention. We want to examine how an organization could turn such a signal into a documented need for verification instead of leaving it implicit.

Transition from project to operations

Teams building AI-enabled services may need to connect design requirements with their later operational use. AICF research in this scenario is intended to cover only the operational phase; pre-deployment assessment, approval, deployment and risk acceptance remain with the appropriate people, methodologies and organizational processes.

Role in the InLabs ecosystem

AICF is a separate research and development project with its own scope. We plan to study the operational side of compliance control, not the way requirements themselves are defined or the systems through which people use AI. The description does not include supervision of other work in the InLabs portfolio.

CDF structures the controlled AI deployment process, and PROXY:AI supports controlled access between users, applications, agents and models. SAVANT-AI is a separate research and development project focused on knowledge and reasoning at organizational scale. AICF may be considered alongside these items, but this page does not declare implemented integration, a shared runtime environment or production data flow. AICF does not create organizational requirements or policies, does not assess pre-deployment maturity and does not take responsibility for the decision.

From requirements to verifiable evidence

At a public level, the research can be described as a simple relationship:

organizational requirements -> operational context of AI use -> evidence for human verification

We want to check whether requirements can be written in a way that makes it easier to assess AI-supported activities in a specific context. We also plan to examine whether significant changes can be signaled without creating a false sense of certainty, and whether the evidence gathered helps the responsible person understand and challenge the basis of a decision. AICF is intended to concern ongoing compliance control; it will not be a universal security tool, a legal opinion or a system that automatically approves decisions. This description does not constitute legal advice.

Detailed mechanisms, data models and decision procedures are deliberately outside this public description. What matters is the direction of the planned research: control that is more continuous, interpretable and connected with real responsibility in the organization.

Research methods and validation

AICF is a research and development project whose assumptions will require evidence, not declarations, after it is launched. The work is intended to include limited test cases, controlled comparisons with appropriately selected reference approaches and qualified expert assessment. The assessment will concern the understandability, consistency, traceability and robustness of the approach in a defined context. It will also include situations in which requirements overlap and the available information is incomplete.

Results will require cautious interpretation. A positive result in one scenario will not establish universal compliance, legal sufficiency or usefulness for another organization. A partial or negative result will be a valid research outcome: it may show where human verification, a clearer requirement or a different control approach is needed. Detailed research materials and unpublished results will not be part of this page.

AICF is currently at the stage of applying for funding under the FENG programme; the application has been submitted and is awaiting a decision. We plan to study the principles, evidence models and validation methods needed for operational compliance control in generative AI environments, together with the limits of its application. The project cannot start without a positive assessment of the application.

Let's talk about collaboration or research partnership

As part of the InLabs research program, allclouds.pl invites organizations, companies and research partners who want to have a real impact on the development of the presented concepts to cooperate. We offer flexible forms of cooperation, various levels of partnership and benefits related to participation in the creation of new technologies that will not be available to other customers on the same terms after the solution is introduced to the market.

You can contact us regarding research and development cooperation via the e-mail address InLabs@allclouds.pl.

SAVANT-AI

Research on controlled work with organizational knowledge

SAVANT-AI (Semantic Autonomous Versatile Advanced Network Technology - Artificial Intelligence) is a research and development project of InLabs. We explore how to combine organizational knowledge, expert experience, AI agents and business systems in a single, controlled work context. It's not about replacing people or previewing a finished product. We want to check whether distributed information can be better used in complex organizations while maintaining human responsibility, access rules and the ability to return to sources.

Why this research is needed

Important organizational knowledge is created in documents, operating systems, expert conversations and everyday decisions. Each source shows only part of the situation, often describing it in a different language and from a different perspective. To answer one important question, the team must find information, check its validity, determine its relevance and agree on it across departments. This is necessary, but time-consuming and prone to loss of context.

The second problem is knowledge that depends on specific people. When experience is not visible to other teams or it is not known where to look for the right expert, it is more difficult for the organization to learn from its own decisions and changes. The risk increases when processes are complex, data is sensitive, and the consequences of an error extend beyond one department.

Just searching for documents is not enough. You still need to understand the question in the context of the task, carefully combine the information, consider access rules, and allow a human to evaluate the request. SAVANT-AI explores this space without assuming that any answer can or should arise automatically.

Potential application scenarios

The examples below are directions for research and are not descriptions of implementations, available features, or promises of results.

Industrial Operations and Critical Infrastructure

In an organization where knowledge about processes, events and exceptions is created in many departments and systems, the studied direction could help teams find an important context for jointly considering the problem. A potential result would be a more complete preparation of decision material by those responsible for the process.

Continuity of expert knowledge

When key knowledge is dispersed between specialists, procedures and event history, SAVANT-AI can be a research topic for safely pinpointing areas that require supplementation or conversation with an expert. The goal is not to automatically take over human competences, but to reduce the risk that important context remains invisible.

Regulated organizations

In sectors where authority, information provenance and the ability to control decisions are important, work could examine whether this approach supports the analytical process. The possible value depends on the specific application, data, environment and control; technology alone does not prove regulatory compliance or safety.

Collaboration between teams

For decisions that cross departmental boundaries, research can see whether a structured context makes it easier to formulate questions, compare perspectives, and document open issues. The possible outcome is a better basis for conversation and verification, rather than an autonomous resolution.

Role in the InLabs ecosystem

SAVANT-AI has a distinct research scope: it focuses on knowledge and reasoning at an organizational scale. The documentation announces relationships that go beyond mere conceptual complementarity: the working environment and place of user contact with knowledge would remain TWIN:DESK, and controlled access between users, applications, agents and models would remain PROXY:AI. The method of conducting the implementation itself would be organized by the CDF. However, these are relationships announced in documents, not confirmed by a working solution.

GENESIS-AI remains a separate direction, concerning the transition from requirements to the application package; we describe their proximity in the portfolio as distinct ranges. This site does not declare a common runtime, dataflow, or working integration. SAVANT-AI does not replace authorization mechanisms, implementation responsibility or human decisions.

General principle of approach

The starting point is simple: an organization can make better informed decisions when the right sources, user question, and expert knowledge are analyzed together, rather than remaining in separate places. The system receives only information approved for use and a question embedded in a specific task. It then helps you sort out the context, point out uncertainties, and determine what still needs to be checked. The result is to be material supporting human assessment.

The difference from simple search is the examination of relationships between information, roles, and consequences of knowledge use. This does not mean, however, that the system automatically understands the organization, makes decisions or determines the correctness of conclusions. The scope, reliability, and usefulness of the responses remain the subject of research and depend on the quality of the sources and the context of use.

High-level model: organizational knowledge and question → controlled context to check → human judgment and decision.

Test methods and validation

Work on SAVANT-AI may include comparisons with appropriately selected baseline approaches, expert evaluation, and qualitative and quantitative testing in controlled cases. The analyzed properties may concern context relevance, usefulness for the user, the ability to find the basis for inferences, resistance to data change and the limits of the solution's application.

An important part of validation is also checking when the answer should not be treated as sufficient: when the sources are incomplete, contradictory, unavailable or require a decision by an authorized person. Security, privacy and compliance will be assessed in relation to the specific environment, data and use, rather than being declared as a universal property. We do not present experimental results, benchmarks or non-public procedures on this website.

SAVANT-AI is a research and development project based on previous conceptual work and early prototypes, with a planned research scope and evaluation criteria. The work concerns the clarification of research problems, limits of responsibility and categories of evidence needed for subsequent validation. Further directions include assessing the quality of the context, examining human-system collaboration, and analyzing limitations in different classes of applications.

This page does not announce a ready service, pilot, commercialization schedule, or availability to customers. Moving to the next phases would require separate evidence and decisions by program owners.

Let's talk about collaboration or research partnership

As part of the InLabs research program, allclouds.pl invites organizations, companies and research partners who want to have a real impact on the development of the presented concepts to cooperate. We offer flexible forms of cooperation, various levels of partnership and benefits related to participation in the creation of new technologies that will not be available to other customers on the same terms after the solution is introduced to the market.

You can contact us regarding research and development cooperation via the e-mail address InLabs@allclouds.pl.

GENESIS-AI

Towards a more controlled path from requirements to software

GENESIS-AI is an InLabs research and development project for which we are applying for funding under the European Funds for a Modern Economy programme. The project concerns the use of generative AI in software engineering. We plan to examine how to move from approved business and technical requirements to a coherent package of application elements that can be reviewed and assessed. Responsibility for engineering decisions will remain with humans.

The research is intended to cover selected types of web applications. Mobile applications, desktop applications and real-time systems remain outside the scope. We do not want to replace the product owner, architect or development team. We want to check whether better organization of work, regular verification and the ability to trace the relationship between a requirement and the result increase the credibility of AI-supported work. Implementation depends on receiving funding; the project cannot start without a positive assessment of the application. GENESIS-AI is a research and development project, not an announcement of a ready-made product or service.

Why this research is needed

Generative AI can create software fragments quickly. It is much harder to build a complete application that still corresponds to the real objective. That is why we plan to focus GENESIS-AI research on four related problems, provided the project receives funding.

Requirements are often incomplete and ambiguous

A business description often combines facts, unspoken assumptions and terms understood differently by different people. If these ambiguities enter an automated process, the resulting solution may work correctly from a technical perspective but fail to meet the organization's needs. We want to study how to detect and clarify such issues before they affect subsequent application elements.

Application artifacts can become inconsistent with one another

An application is not just code. Interfaces, data structures, tests, documentation and deployment materials should describe the same system. If they are created separately, a change in one place can create a contradiction elsewhere. We plan to examine how to control the consistency of the entire solution instead of treating its generation as one opaque step.

Speed is not evidence of quality

A fast result may be difficult to maintain, unsuitable for a given environment or insufficiently secure. Correctness, maintainability, portability, reliability and security are assessment categories, not qualities that can simply be assumed.

Responsibility cannot be handed over to a model

Objectives, constraints, acceptable risk level and acceptance of the solution remain human responsibilities. GENESIS-AI is intended to study how an AI-supported process can provide material for assessment and mark the moments that require review, without pretending that automation replaces responsible judgment.

Possible use cases

The following scenarios are research directions, not information about deployments, customers or current availability of the solution.

Internal applications with a clearly defined scope

Research scenario. GENESIS-AI, if funded, is intended to support research into the preparation of limited web applications based on requirements previously reviewed by domain and technical experts. In that context it would be possible to assess compliance with requirements, artifact consistency and the ability to develop them further.

Verification of a solution concept

Research scenario. A reviewable application package could help a team assess whether a proposed process, information model or user interaction corresponds to the intended problem. The result would still require technical assessment and would not itself be a ready deployment.

Repeatable work of engineering teams

Exploratory scenario. Teams building similar classes of applications may be interested in research into more repeatable preparation of code, tests and documentation. The key question is whether structure-based support reduces inconsistencies without hiding assumptions or weakening engineering responsibility.

Development in a high-requirement environment

Exploratory scenario. GENESIS-AI may be studied in organizations that use their own processes to manage change, control access and approve engineering work results. The project itself does not determine the legal, regulatory or security compliance of a specific use case; those depend on the data, environment, controls and decisions of the organization.

Role in the InLabs ecosystem

GENESIS-AI is intended to focus on the transition from approved requirements to a package of application artifacts prepared for review. Other items in the InLabs portfolio have separate scopes: CDF organizes AI deployment, TWIN:DESK is a work environment, PROXY:AI concerns controlled access, and SAVANT-AI concerns organizational knowledge.

The GENESIS-AI research documentation does not establish dependencies on these elements, which is why we describe their proximity in the portfolio as separate scopes rather than confirmed technical integration, a shared runtime environment or production data flow. GENESIS-AI will not replace the CDF deployment methodology, the TWIN:DESK work environment, the PROXY:AI access boundary and policies, or the research into knowledge and reasoning conducted in SAVANT-AI. Cooperation between these elements would require separate, verified foundations.

Approach and operating principle

GENESIS-AI is intended to study the relationship between three levels:

approved requirements -> controlled software development process -> reviewable artifact package

Approved requirements are intended to be the point of reference. On that basis, connected application elements would be created in a controlled way and then checked. The completed package would go to the people responsible for technical assessment and the decision on further development or possible deployment.

This is a general description of the planned way of working, not the architecture of the solution. We want to check whether better organization of requirements and explicit checking of results can reduce the gap between the original intent and the outcome. We also plan to examine whether shared assessment criteria for code, tests and documentation lead to a more coherent result than creating these elements independently. These are hypotheses to be tested only after the project starts; experiments have not yet begun and depend on receiving funding.

Research methods and validation

The credibility of GENESIS-AI is intended to be assessed through complementary, planned classes of research: controlled comparisons with reference approaches, reviewed cases, repeatability checks across different classes of problems, and qualitative and quantitative assessment in a defined context. The analysis is expected to cover, among other things, compliance with requirements, functional and integration correctness, reliability, maintainability, portability, security, controllability and usefulness for the engineering team.

Assessment will not be based solely on demonstrations or the volume of materials. Automatic checks will require expert interpretation, and repeated attempts should distinguish a durable effect from chance. A partial or negative result will be a valid research outcome; evidence of failure will be equally important: unresolved ambiguity, divergence between artifacts, limitations of review or the need for additional human work.

The project is currently at the stage of applying for funding under the FENG programme; the application has been submitted and is awaiting a decision. The research questions, hypotheses, assessment categories and types of evidence needed for validation were described for the application. Their actual implementation, scope and research results depend on the award of funding and will begin only after the project is launched; without a positive assessment, the project will not be started.

The planned direction of further work, to be carried out after funding is awarded, includes research into fidelity to requirements, ambiguity management, consistency across the artifact lifecycle, credibility across different classes of problems and ways to conduct review without losing human responsibility.

Let's talk about collaboration or research partnership

As part of the InLabs research program, allclouds.pl invites organizations, companies and research partners who want to have a real impact on the development of the presented concepts to cooperate. We offer flexible forms of cooperation, various levels of partnership and benefits related to participation in the creation of new technologies that will not be available to other customers on the same terms after the solution is introduced to the market.

You can contact us regarding research and development cooperation via the e-mail address InLabs@allclouds.pl.