Technical and Socio-Business assessments of AI Maturity in pilots of Explainable AI
The EU-funded XMANAI project focuses on Explainable AI, as the ability to make machine decision-making processes understandable; Explainability has proven to be a key element in stimulating the adoption of AI in various areas because it provides transparent and understandable information about algorithmic decisions and outcomes.
The XMANAI project envisions four pilot cases of Explainable AI being developed at partner industries.
To assess the real impact of the pilots on day-by-day operations, their evaluation was not limited to the technical success of the AI solution alone but sought to understand how much, in practice, Explainability affected the processes and the socio-economic aspects of the company.
For this purpose, the 6Ps AI transformation model, developed by POLIMI, was used. The 6Ps AI transformation model is one of the tools developed by Politecnico di Milano to assess the digital and AI maturity of organizations and to suggest roadmaps and journeys for the implementation of digital transformations; it supports manufacturing companies to assess their current and expected level of Digitization and to track a Digital Transformation roadmap; from this point of view, it can be considered more a “tactical” tool than a “strategical” one, in the sense either that the company has a project to improve some aspects of its approach to digitization (in which case the roadmap to the implementation of digital solutions is implicitly defined by the project, and the 6P helps measure where progresses take place and how large they are) or the company has identified some digitization gaps and wants to fill them (in this case, 6P provides a support to focus on the potential areas of improvement, and as an “inspirational” tool, suggesting concrete ideas of actions to implement digital solutions that may add to the company’s ideas).
The 6Ps analysis takes place on six pillars (from which, the name “6Ps”): a basic assumption is that, in order to succeed in a digital transformation process, it is important to boost not only the technical dimensions but also the so-called “socio-business” dimensions. For this reason, the six pillars are Product, Process, and Platform, clustered as “technical pillars” and People, Partnership, and Performance, clustered as “socio-business” pillars.
Each pillar contains 6 questions (that is, 6 fields of analysis) and each question has 5 possible answers, corresponding to 5 sequential levels of development, from level 1-Initial to level 5-Exploited.
The first pillar, PRODUCT, has the objective of evaluating to which extent the manufacturing SME is digitally mature, in terms of the Product or Product-Service System, that it offers to the market.
The PROCESS pillar refers to the industrial manufacturing processes, and to the transition towards automated, smart, and connected processes.
PLATFORM is focused on CPS and embedded systems, Industrial IoT, Industrial Internet, Industrial analytics, and Vertical and horizontal interoperability.
The PEOPLE pillar is the first of the so-called “Socio-Business” pillar, and is given particular prominence (12 questions instead of 6 questions as in the other pillars) because in a digital transformation process, the involvement of staff is the real driver of digital transformation: staff skills, training, and involvement are the real heart of the change. In the PEOPLE section, smart features (typical of Industry 4.0 and AI) are crossed with different types of professions in the company.
The last two pillars, PARTNERSHIP and PERFORMANCE, refer to the partnerships with Digital Innovation Hubs, Universities, and IT providers and suggest possible KPIs to measure the transition: KPIs are not only Operational and Economic but also refer to the ability of the company to measure its performances over other aspects, such as Environmental, Social, Product-Service Lifecycle and performances of the entire Supply Chain.
The survey is conducted online, taking about 1 hr, and the output is given in form of a radar chart, allowing us to visually compare the current (AS-IS) and the expected (TO-BE) levels for each dimension of analysis.
As mentioned above, in the XMANAI project the 6Ps model was suitably modified to account for how aspects of Explainability were transposed, especially in the section measuring the impact at the staff level.
The application of the evaluation model went through presentation webinars followed by a few sessions to fine-tune the questions to fit the Explainability concepts (especially in the PEOPLE section of the model, which deals with the impact on staff), and ended with the compilation by the four pilots involved, who represented the current situation of the processes affected by the pilot (AS-IS) and the expected progress (Expected TO-BE).
After the pilots will have come to an end, the compilation of the 6Ps will be proposed again, and answers collected about the progress that was actually achieved (Actual TO-BE). This will be followed by a comparison of Expected and Actual TO-BE, to see how well expectations have actually been met and to identify the reasons for possible deviations from the targets.