AI and Sustainability: Turning Technological Innovation into a Performance Lever

The integration of artificial intelligence is entering a defining phase for companies. At Tennaxia Connect, Rémy Balangué, Chief Product Officer at Tennaxia, and Grégoire Carpentier, Chief Technology Officer at Tennaxia, shared their view on the challenges of AI applied to sustainability. Moderated by journalist Sébastien Borgnat, the session compared technological promises with operational and environmental realities. Between optimising data processing, controlling digital carbon footprint, and governance requirements, the talk showed that AI must not be a mere announcement effect. When properly integrated, it becomes a clear, useful framework connected to organisations' ambitions for sustainable performance and strategic management.

Why AI and sustainability are converging now

The convergence of artificial intelligence and sustainability trajectories comes at a time when the volume of data to manage – whether regulatory texts, charters or non-financial reports – is becoming critical for companies. To gauge the situation on the ground, Tennaxia carried out a major survey among its clients. The results highlight three particularly telling figures:

• 60% of respondents already use AI systems as part of their work, but via personal accounts on consumer platforms. AI has thus made its way into day-to-day work before even being formally structured by IT departments.
• 70% of them cite time savings as the primary expected benefit of AI in their daily tasks.
• 20% say they do not use it at all, mainly due to a lack of trust or because the cost-benefit ratio is not yet considered satisfactory.

This shift confirms that the main expectation is no longer discovering the technology, but the ability to govern it. Companies are asking for solutions capable of delivering a genuine productivity gain while guaranteeing a trustworthy environment.

A pragmatic approach that favours genuine usefulness over gadgets

Integrating AI into CSR and HSE initiatives should not aim to overturn everything purely for the sake of innovation. The objective, as Tennaxia points out, is to focus on real business needs in order to avoid “gadget” tools. Large language models (LLMs) are, by nature, text processors designed to predict output text from input text. They are therefore naturally well suited to managing sustainability data, which relies heavily on textual and tabular formats.

The question is not to reinvent existing systems, but to make them more effective where the administrative burden is heaviest: information research, rewording for different audiences, or consolidating scattered sources. AI becomes effective when it informs decisions without adding complexity to business processes.

The three flagship features launched on the ESG side

To put this logic into practice, Tennaxia has rolled out three concrete features this year on its ESG platform:

• Bulk translation of questionnaires: automatically translating a reference questionnaire into more than 20 languages, to make it easier to collect data from subsidiaries worldwide without leaving the application.
• Creating dashboards in natural language: the user states their request directly to automatically generate the desired dashboard, with the relevant indicators, data and chart formats.
• Assistance with drafting the sustainability report: a module capable of consolidating and reusing information from previous reports or policy documents to pre-fill data for the current reporting cycle.

Internal experimentation and the transformation of HSE roles

On the HSE side, the approach is based first on internal experimentation before large-scale rollout. Tennaxia is currently testing, with its own consultants, AI tools applied to audits: speech-to-text for entering reports, note-taking via photos, and automatic generation of summary tables or presentation materials.

This approach makes it possible to validate the tool's functional relevance and standardise practices upstream. By turning drafting time into proofreading time, AI acts as an accelerator. More importantly, it changes the nature of the tasks, opening the way to new decisions.

Concrete example: When opening a factory in a new country, analysing an unfamiliar and voluminous set of regulations – a task often so tedious it can delay a project – can now be handled and accelerated by automated reading and summarising capabilities, thereby unlocking decision-making.

The environmental footprint of AI: the territory of frugality

One of the key messages of the talk concerns the environmental responsibility linked to the use of the technology itself. AI is a major energy consumer, an unavoidable issue for sustainability players. Globally, data centre consumption stands at more than 400 TWh. AI currently accounts for 15% of this consumption, but according to the International Energy Agency (IEA), this share could reach at least 30% by 2030.

In light of this, governance must incorporate digital frugality criteria, drawing in particular on the work of associations such as Green IT or the think tank The Shift Project. Tennaxia applies this vigilance through two levers:

• An internal usage policy backed by continuous scientific monitoring to track key publications.
• Common-sense heuristics during development, choosing the most suitable and most frugal model for a given task, rather than systematically relying on the heaviest and most energy-intensive infrastructure.

Trust, sovereignty and keeping humans at the centre

The reliability of AI remains a major point of vigilance, as no model is 100% reliable. The answer to this limitation lies in strict governance based on the “human-in-the-loop” principle. On the platform, AI operates exclusively on an “opt-in” basis: it is never activated by default and requires an explicit action by the company. Furthermore, the tool is limited to suggesting text or pre-filled content; it cannot overwrite or replace data without explicit human validation that is fully traceable and time-stamped.

In terms of sovereignty and security, data is stored and processed within the European Union, and specifically in France for the highest confidentiality levels (requiring data residency and inference to take place in French data centres). Client data is never used to train providers' models.

Finally, this approach fits within the requirements of ISO 27001 and SOC 2, guaranteeing the system's auditability and resilience: the software is designed to operate without AI, thereby avoiding any loss of functionality for the company in the event of a technical disruption or a surge in the technology's costs. This architecture is also aligned with the requirements of the European AI Act, whose provisions apply progressively through to 2027.

From technology to sustainable management

Artificial intelligence does not ask companies to produce technology for its own sake, but to make better use of it in order to make better decisions. In an increasingly demanding context of sustainability and compliance, it provides a useful framework for moving from passive data collection to strategic management. By automating repetitive administrative tasks, it restores value to human expertise on the ground.

For organisations, the challenge now is to turn technological enthusiasm into an opportunity for transformation: making AI a well-governed, frugal and sovereign foundation to drive CSR performance that is robust, coherent and fully manageable.