Explainable AI in all its states (or almost!): a few use cases

Posted on 2024-02-06 by Anabasis

Translations: fr

Categories: Enterprise

2024 marks a turning point for Anabasis: after successful production launches with key customers, we needed to take a step back, beyond the sectoral or business applications of our Karnyx explainable AI technology.

Indeed we are not limited to a single sector. Certainly, defense, aeronautics, health and HR, which were the first to trust us, continue and will continue to call on us on issues of heterogeneous data to align, sensitive information to protect, strategic decisions requiring to be justified at 100%. At a time when Anabasis is experiencing dynamic growth in the functionalities of its AI suite, Karnyx, where the workforce is growing (+50% in one year), we are ready to continue our penetration of other sectors. We have already done this in the agri-food industry, bibliographic research and mergers and acquisitions. But there are others: legal and accounting professions, banking & insurance, pharmacy, standardization bodies, process or manufacturing industries, government, etc.

Rather than covering sectoral or business use cases, in this post we have decided to present four applications which we believe are eminently transposable to other sectors. Let’s start with our key differentiator: the explainability of the results produced by our AI.

The observation: decision-makers are confronted with an abundance of data and computerized solutions. However:

  • Data warehouses shield the sources: heterogeneous data, not updated at the same time.

  • Usual AI is a black box. However, strategic decisions or very sensitive contexts (health, finance, industry, defense) require fully justified decisions.

Karnyx, Anabasis' explainable AI, fits into this context and provides the missing link of explainability.

Let's start with life without Karnyx.

Decision-makers collect the raw material, the data, analyze it (by hand or with the help of a usual AI) and report back in the form of an HMI or a dashboard. It looks roughly like this:

Without Karnyx

This the exact point where problems start to occur:

  • The quality adjustment after extraction is incomplete and “tricky” because it is not controlled by the profession,

  • Manual or algorithmic analysis (AI) applies to heterogeneous data,

  • The presentation of the results is based on intuition (manual analysis) or statistics (algorithmic AI) or, when based on generative AI, can produce hallucinations.

Basically it's the infernal triptych of heterogeneity of basic data, intuition/hallucination in the analysis and in the end, blind trust in a black box:

Without Karnyx

As a result, many decision-makers look at the results and ultimately decide based on entirely different criteria, called intuition, experience, common sense, etc. Basically, decision-makers find themselves reduced to doing what they did before IT even existed, constructing a posteriori reasoning and a set of hypotheses to justify the decisions they are about to make.

What about a world with Karnyx' explainable AI, then?

If there is only one thing to remember: Karnyx does not replace everything, but simply fits into existing IS, providing data quality and explainability driven by a semantic model:

With Karnyx

What does this actually change for the decision-maker?

With Karnyx

  • Quality improvement is driven by the semantic model (business),

  • The analysis is entirely justified by the semantic and logical model,

  • The presentation highlights the reasoning followed and allows the results to be 100% justified.

Before talking about application cases, we can state, without much risk of being mistaken, that our explainable AI approaches and technologies are applicable in all sectors, as long as we need:

  • to make critical decisions,

  • a precise and reasoned justification of decisions.

Of course, the following application cases are anchored in sectors since they correspond to projects actually carried out by Anabasis. However, we hope that our reader, if he does not belong to the sector in which the project was carried out, will be able to abstract himself from the sectoral context presented and see the application for his sector, possibly with the help of the indications in the paragraphs “applications to other sectors”.

First application: making justified decisions in the health sector.

Justified decisions

Initial situation :

  • The client wants to be able to make 100% justified decisions regarding health: vaccination, foresight, etc.

  • Health data is in heterogeneous information systems.

  • This data is critical for the medical monitoring of personnel but cannot be used with confidence.

With Anabasis and Karnyx Explainable AI:

  • The solution was delivered in less than 6 months and used by several hundred health practitioners and epidemiologists.

  • It now makes it possible to make decisions in a rapid, reliable and reasoned manner that could never have been made in this way.

Application to other sectors: wherever you have heterogeneous data to quality and critical decisions to make and justify – commercial strategy, R&D performance, environmental and social performance, operations performance, legal or accounting relevance.

Second application: Justify and guarantee the protection of sensitive information.

Protected Information

Initial situation :

  • An industrial company in a sensitive field must work with external actors but must absolutely preserve industrial and strategic secrets.

  • The protection process is centralized and manual: it takes several days to validate certain information sharing that is necessary for operations.

  • This delays operations in the field unless it puts some at risk.

With Anabasis and Karnyx Explainable AI:

  • 100% justified protection of sensitive information.

  • Secure automation of processing thanks to data marking by sensitivity.

  • Increased operational efficiency thanks to new access to all authorized information.

Increased operational efficiency thanks to new access to all authorized information. Application to other sectors or professions: wherever:

  1. you have sensitive information (trade secret, institutional secret) and associated documents to protect from organizations that are part of the extended enterprise but do not need to know but:

  2. You want this sensitive information not to slow down your extended mode operations.

**Third application: Work confidently on aircraft maintenance with explainable information.


Initial situation

  • The availability of aircraft is reduced by the absence of a common reference system between maintenance stakeholders (state, industrial).

  • The customer wants all these maintenance players to share the same understanding. He decides on the migration to a single information system.

With Anabasis and Karnyx Explainable AI

  • 80% of data was migrated in a few weeks, automatically and 100% justified

  • Aircraft availability.

Application to other sectors or professions: wherever your master data management needs to be pushed to an additional level of performance in order to be able to better collaborate with other stakeholders in critical processes: production, maintenance, etc. in a logic of digital continuity.

Fourth application: unified and reliable HR management.


Initial situation

  • Personnel are managed in different information systems, using their own repositories.

  • A ministry wants to bring together these different information systems and use a single repository

With Anabasis and its explainable AI suite

  • Secure management of data recovery.

  • Possibility of making HR policy decisions (job and skills management) assignment and rewards justified on reliable data.

  • For each data consulted Karnyx AI explains which original data were used and which transformations were applied.

  • A single, readable repository facilitates recruitment, retention, internal and external bridges.

Application to other sectors or professions: wherever you need to manage the consequences of a merger, a merger – technical, production, marketing, commercial information. The goal is to make the most of these mergers or mergers in terms of synergies, using pivotal and unifying semantic models developed in Karnyx.


The fields of application of explainable AI are vast. Its uses are even more so: control by the business of the data to be used in the case of a migration or convergence towards a target IS, decision support in support of strategic decisions, etc.

The common denominator of all these variations is ultimately to master and promote the sensitive knowledge of your organization for decisions justified by explainable AI.

To continue this exchange with the Anabasis team: contact@anabasis-assets.com.

Richard Roll

Founder Anabasis richardroll@anabasis-assets.com

Explainable artificial intelligence and Semantic Web tools - Part 2

In the previous post, we focused on RDF-style knowledge graphs, the basis of Anabasis technologies. Now we a

Retour sur la participation d'Anabasis à SemWeb.Pro 2023

Parce qu'Anabasis Assets a construit sa suite logicielle sur les normes et standards du web sémantique (entre autres), il était naturel de participer à la Jou

Automatic ontology learning or how all texts hide schemas

Text-based automatic ontology learning supports knowledge engineers in analysing business knowledge from company documents. This is a promising area of innovation at Anabasis, which has initiat