Reliable AI depends on more than the model.

AI conversations often focus on models, algorithms and performance. But as AI moves into real-world applications, reliability increasingly depends on what surrounds the model: the data, people, processes and standards behind it.

Across August, three developments highlighted different parts of this challenge, from responsible business practices to clinical AI and the realities of delivering reliable data for clients.

Together, they point to the same lesson: trust in AI is built long before a model reaches deployment.

‍

Responsible Business Starts With How We Operate

Why the practices behind technology matter

As AI becomes increasingly embedded in business-critical workflows, responsible practices are becoming part of the foundation around the technology. How organisations manage people, processes and standards can influence how responsibly AI is developed and deployed.

On 12 August, IngeData joined the United Nations Global Compact, becoming part of a global community of more than 23,000 companies across more than 160 countries committed to principles covering human rights, labour, environment and anti-corruption, as well as the Sustainable Development Goals.

For IngeData, the commitment builds on an existing focus on continuous improvement, including ISO 9001 certification, an EcoVadis Sustainability Rating and support for teams across its international operations. It also aligns with the UN Global Compact's 2026–2030 strategy, which focuses on helping companies turn commitments into measurable action.

The broader lesson is that responsible AI is not only about how a model performs. It also depends on the standards and practices surrounding the technology.

Read more about why IngeData joined the UN Global Compact and what this commitment means for our organisation.

Read our full commitment

‍

Healthcare AI Is Only as Reliable as Its Ground Truth

As AI moves closer to clinical practice, the data behind it matters more.

At Singapore's National Day Rally on 23 August, Prime Minister Lawrence Wong highlighted breast cancer screening as an example of how AI can support radiologists. He noted that dense tissue and tumours can appear similar on mammograms, making them difficult to distinguish, and that AI can act as another pair of highly trained eyes for radiologists.

That raises a less visible but important question: what does it take to train those eyes?

Reliable clinical AI depends on the quality of the ground-truth data behind the model. In collaboration with University Hospital Basel, IngeData supported open-source breast cancer segmentation research using more than 600 manually annotated breast carcinoma cases, with over 45 hours of specialist radiologist review and consistent annotation protocols across different breast densities and imaging conditions.

The same principle applies to lung cancer screening. Thousands of pulmonary nodules were annotated by radiologists to support an AI-driven screening solution that achieved 97.7% precision in detecting lung cancer across all stages, with 96.8% recall for early-stage (Stage I) lung cancer detection.

The takeaway extends beyond any individual model: as AI moves further into clinical practice, the quality and clinical rigour of the data used to establish ground truth become part of what makes the technology trustworthy.

IngeData will bring this expertise to Medical Fair Asia 2026 in Singapore, 9–11 September, where the Healthcare team will be available to discuss medical AI data and annotation workflows.

​Let's talk about healthcare data

‍

Reliability Is Also Measured in Delivery

What reliable data delivery looks like in practice

For organisations working with AI data, reliability extends beyond the quality of the final output. It also depends on how requirements are understood, how work is managed and how consistently teams deliver throughout an engagement.

This is reflected in feedback from Dan Hickman, IT Director at Flogas Britain, following his experience working with IngeData.

“Ingedata provided a professional and reliable data labelling service throughout the engagement. The team quickly understood our requirements and delivered the work to a consistently high standard, with strong attention to detail and accuracy.

Communication was clear and responsive, and any questions or clarifications were handled efficiently. The final output was well structured, easy to work with, and met our expectations. Overall, we were very pleased with the quality of the service and the support provided by the Ingedata team.”

For clients, these details matter. Reliable data work is not only about technical accuracy. It is also about having a delivery process that is consistent, responsive and easy to work with.

See how we deliver reliability across industries

‍

Looking Beyond the Model

Reliable AI does not begin when a model is deployed. It is built through the quality of the data, the expertise involved, the processes that govern the work and the standards that support responsible delivery.

Across responsible business, healthcare AI and client delivery, the same principle holds: trust in AI is built long before the final model is used.

If these challenges reflect something your organisation is working through, we welcome a focused conversation with you.

Start the conversation