AI can support business sustainability by helping teams identify waste, use resource data and check the evidence behind environmental claims. The leadership task is to connect that information to a decision, then measure what changes.
A water alert, a farm-data answer and a draft sustainability report do different jobs. They also need different measures of success.
My perspective comes from connecting projects recognised through the AI Awards with conversations on the AI Ireland Podcast. The projects reveal specific applications. The conversations add detail about the work around them. I use both to develop practical questions for leadership teams.
For this topic, the question I would put on the boardroom table is simple: what will someone do differently because this information is available?
Where can AI help with business sustainability?
I find it useful to separate three jobs before discussing a product:
- Spot a resource problem. Identify a possible leak, wasted food or an unusual consumption pattern so someone can investigate.
- Make information usable. Help a person explore records, understand a pattern or prepare a decision.
- Check the reporting. Find missing support for an environmental statement or help organise evidence for review.
A team may make progress in one area without demonstrating progress in the others. A report prepared more quickly is a reporting improvement. A reduction in water use needs its own evidence. Making that distinction early gives the project a clearer brief.
Three Irish AI Awards examples and their leadership lessons
Herdwatch, Musgrave Retail Partners Ireland and Uisce Éireann are among the finalists for Best Application of AI in Sustainability in the AI Awards 2026 shortlist. The short descriptions below reflect the published entries. They are not independent verification of environmental outcomes. The leadership lessons are my interpretation.
Uisce Éireann: connect an alert to an operational response
The finalist description covers AI for leak detection, infrastructure inspection and operational support, alongside workforce learning and governance.
The lesson I draw is about the handover. An alert creates work for somebody. A useful project brief should explain who receives it, what they check and how an action is recorded.
For a business considering a resource-monitoring tool, I would map that sequence before selecting the technology. Otherwise, the organisation risks collecting more signals than its team can investigate.
What I would measure: how often an alert is useful, whether it leads to action and whether the relevant resource measure changes afterwards. Those are proposed measures, not reported Uisce Éireann results.
Herdwatch: make the organisation’s own records easier to use
Herd Intelligence lets farmers explore records such as fertiliser use, soil information, livestock treatments and weights through conversation.
This connects directly to Proprietary Data in my 5Ps framework. A familiar interface becomes more useful when it can answer a question using the organisation’s relevant records.
I would begin by asking users which decision is difficult today. Then check whether the records needed for that decision are available, current and understood. An easier way to ask a question cannot resolve information that was never recorded.
What I would measure: whether users can find the relevant evidence, recognise gaps and make a better-informed decision. Any claim about environmental improvement needs a separate outcome measure.
Musgrave: challenge the evidence behind the claim
Musgrave’s Green Claims Agent reviews text and visual material against approved guidance, flagging potentially misleading statements and missing evidence for human review.
That raises a useful question for any marketing team: if we publish this statement, can we show what supports it?
My recommendation is to connect each material claim to a source, a defined scope and a reviewer. For example, a statement about one packaging component should not quietly become a statement about the whole product.
What I would measure: the quality of the review and the unsupported statements it identifies. A claim-review tool still needs accountable people to make the publication decision.
What the podcast conversations add
Positive Carbon: fit measurement into the working day
In episode 226 with Positive Carbon founder Mark Kirwan, published in January 2026, the discussion explored food-waste measurement in hospitality. The published summary describes cameras and scales helping kitchens identify what is discarded and when, with an emphasis on fitting the process into existing work.
My takeaway is to examine the effort required to keep a measurement system working. A tool can produce useful information in a demonstration yet struggle if every busy shift requires extra manual recording.
For a hotel or catering team, I would ask who reviews the findings, how they influence purchasing or preparation, and whether those changes are checked. The relevant outcome is the waste avoided through a changed decision.
EcoEasi: make it possible to check the input
In episode 173 with EcoEasi’s Muiris O’Dwyer and Aidan Connolly, published in November 2024, we explored supply-chain ESG data. The summary describes bringing information together, pre-populating reports and allowing businesses to confirm its accuracy.
The leadership lesson I take is that someone must be able to check where a reported answer came from. A supplier should be able to distinguish a recorded fact, an estimate and an unanswered question.
These are lessons from the published episode summaries. They do not establish a particular supplier’s current capabilities or guarantee results for another organisation.
Measure the result your project is meant to change
Before starting, I would agree a small measurement sheet with the operational owner. Keep resource outcomes, service quality and financial effects separate.
| Proposed project | Useful outcome measure | Context to record |
|---|---|---|
| Food-waste monitoring | Kilograms discarded per meal served, alongside total waste | Meal numbers, menu changes, service quality and the intervention made |
| Water-use investigation | Measured consumption before and after a confirmed action | Activity levels, operating hours, seasonality and meter coverage |
| Environmental-claim review | Unsupported claims identified, missed issues found by reviewers and correction time | Which guidance applied, evidence available and who approved the final wording |
These are suggested evaluation measures, not findings from the finalist projects.
A lower total bill might reflect a price change. Lower consumption might reflect a quieter week. An AI project needs a comparison that accounts for those factors before the team attributes an improvement to it.
My guide to measuring AI value develops the wider business case. For sustainability, I would also keep the underlying units visible: money, kilograms, water volume and emissions estimates answer different questions.
Include the footprint of the AI system
AI itself uses resources. The International Energy Agency’s Energy and AI report examines both the electricity needed for AI and its potential applications in the energy sector.
My recommendation is to compare an AI option with a simpler way of doing the job. Ask what additional equipment and computing are required, what the supplier can substantiate and how usage will grow.
Where the evidence is incomplete, keep that uncertainty visible. A useful operational pilot is not automatically evidence of a net environmental benefit.
A 30-day starting plan for a leadership team
This is a suggested planning exercise. Thirty days may help establish a baseline and test a workflow; some environmental outcomes require much longer observation.
- Week 1: choose a decision. Identify one resource problem, its owner and the people who can act. Use the People and Process parts of the 5Ps to map the current work.
- Week 2: examine the evidence. Review the available Proprietary Data. Record missing periods, unreliable inputs and the comparison you can reasonably make.
- Week 3: test the smallest useful change. Assess the Platforms needed. Compare the proposed AI workflow with the current process or a simpler alternative. Retain human review where a consequential decision is involved.
- Week 4: decide what deserves further testing. Review errors, effort and initial results. Under Products & Services, ask whether the change improves what customers receive. Continue, revise or stop based on the evidence available.
For a wider starting assessment, use my 5Ps AI readiness checklist.
Common questions about AI for business sustainability
Does better ESG reporting mean a business is more sustainable?
Better reporting can improve visibility and accountability. It does not, on its own, demonstrate lower resource use or emissions. Track the reporting improvement and the environmental outcome separately.
Can a general-purpose chatbot calculate our carbon footprint?
I would use it cautiously to help organise questions or explain a documented method. For a business footprint, use appropriate activity data, relevant emissions factors and a suitable calculation methodology, with qualified review. The GHG Protocol’s calculation guidance is a useful starting resource. Fluent text is not evidence that a calculation is complete.
Where should an SME start?
Choose a recurring problem with an owner who can change the process. Check what can already be measured. If a straightforward process change solves it, that may be the right starting point. Add AI when there is a clear role for it and a practical way to evaluate its contribution.
Bring real AI examples into your leadership discussion
The perspective I bring to an event is the connection between specific projects, practitioner conversations and the decisions leaders face. An Awards entry gives us an application to examine. A podcast conversation adds context. My role is to turn that material into a discussion your audience can use.
For a leadership conference, retail event, hospitality gathering or industry forum, an AI keynote can help the audience identify useful opportunities and ask better questions about the evidence.
Share This Story, Choose Your Platform!
Planning an event or leadership offsite?
Mark Kelly delivers evidence-based AI keynotes and hands-on AI leadership workshops for boards, executive teams and conferences across Ireland and Europe.





