Helping organisations make better use of sustainability evidence.

Brace for Impact is built around a difficult problem. Organisations need to work with large amounts of sustainability evidence, often drawn from different sources and with varying levels of completeness, quality and structure. That creates friction. Information may be available, but that does not mean it is easy to interpret or act on. AI could help here, but it also introduced risk. In a product dealing with evidence and reporting, outputs need to be grounded in the underlying data. They need to help users move forward, not create more uncertainty. The challenge also involved product judgement, trust and integration.

We approached the work as a product design and systems problem first. Working closely with the ImpactIO team, we focused on how AI could support real product workflows. That included preparing and structuring underlying data so it could support meaningful interpretation, testing how AI could help users identify patterns, gaps and next steps, designing outputs that were clear, relevant and tied back to source information, integrating AI into the product experience, and prototyping quickly to test which approaches were useful and which were not. This helped narrow the work to the parts of the product where AI could add practical value.

Before, users could access sustainability evidence, but interpreting it often required more manual effort and more judgement outside the platform. The direction of the work was to bring more of that support into the product itself, so users could work with evidence more easily, understand what mattered, and make progress with greater clarity.

The project showed that AI can play a useful role in products built around complex, structured evidence, when it is applied carefully. It helped establish a clearer direction for how AI could be embedded within Brace for Impact in a way that supports interpretation and decision-making, without losing sight of trust and traceability. More broadly, it reinforced an important point. In products like this, the value of AI does not come from novelty. It comes from how well it fits the underlying data, the user's workflow, and the decisions they are trying to make.

Many AI projects focus too early on the model. In practice, the harder and more valuable work is often elsewhere: shaping the data properly, understanding where users need help, and designing outputs that people can actually use. That was the focus here. For organisations building products in complex, data-heavy domains, this is what makes AI commercially useful. Not because it sounds innovative, but because it can make a product more usable, more trusted and more valuable over time.