Skip to content

How Caribou uses AI

Caribou’s mission is to help make the digital world more inclusive and sustainable. We hold ourselves to that same ambition in how we adopt AI.

Best practice around responsible AI use is still evolving, and governance across our sector is still developing. We aim to help shape and demonstrate responsible practice through thoughtful adoption, rigorous governance, and openly sharing what we learn, including where we encounter challenges. Our company values remain our foundation as we navigate this evolving way of working with care and rigor. 

General approach

At Caribou, we use AI to thoughtfully support human expertise and we believe human judgement is irreplaceable. Used carefully and selectively, AI helps our teams expand expert examination of evidence, challenge our ideas, and work more efficiently to deliver greater value to clients and partners. We recognize AI’s limitations and make deliberate choices about where it adds value and when other approaches are more appropriate. Our aim is to use AI to strengthen our work, while upholding the rigorous standards of evidence and maintaining the relationships and human judgment on which it depends.

The following principles put our company values into practice, supported by carefully selected tools, practical training, and clear internal policies.

1) Human expertise throughout

Guided by our value of Purposeful Inquiry

Our people lead the work and exercise judgment at every stage, using AI to support their expertise. They decide where AI can add value, guide its use, and critically assess what it produces. AI-assisted outputs are reviewed, validated, and refined before they become part of a deliverable, with the level of scrutiny matched to the purpose and consequences of the work. Contextual knowledge, professional expertise, and maintaining integrity in respectful interactions with clients, partners, and communities remain central to our approach and decision-making. Final judgment remains with our people; we do not delegate decisions about findings, recommendations, and deliverables to AI. We take full responsibility for the quality and integrity of our work, regardless of how it was produced.

2) Quality grounded in evidence

Guided by our value of Committed to Better

Expert validation is a standard part of AI-assisted work. We never present unverified AI-generated analysis as validated evidence; our findings remain grounded in source documents, transcripts, datasets, and other research material. We verify AI-assisted factual claims, quotations, and references against their sources and apply expert scrutiny to analysis and interpretation. Where AI materially contributes to research or analysis, we document its role, the method, and the validation performed so the work can be examined and explained.

3) Attention to bias, context, and inclusion

Guided by our value of Thoughtful Connections

We recognize that AI can reproduce and amplify bias, overlook less represented perspectives, and distort the languages and contexts that matter to our work. We ask whose knowledge is represented, whose voice is missing, and whom an AI-assisted process might exclude. We draw on subject-matter expertise, local knowledge, and different forms of knowledge to identify and address bias and gaps in perspective. Where needed, we seek additional perspectives, revise our approach, and engage affected people or partners so our work and decisions are informed by the contexts in which we operate.

4) Protecting information

Guided by our value of Everyday Integrity

We protect the information entrusted to us by clients, partners, and the people whose experiences inform our work. Tool approval does not authorize every use: we assess what information may be entered and under what safeguards, respecting client requirements and our commitments to privacy and consent.  We limit the collection, retention, and processing of personal and sensitive information to what is necessary for the work. We use approved, organization-managed AI tools with contractual, security, and data protection controls, including terms and settings that exclude project information from model training. 

5) Transparency and accountability

Guided by our value of Everyday Integrity

We agree expectations for AI use with clients at the start of an engagement and honor any of their requirements that go beyond our own standards. We explain where AI materially contributes to our work, including relevant limitations and trade-offs, known risks and areas of uncertainty. If AI contributes to an error or misrepresentation, we take responsibility, inform those affected, and take proportionate steps to put it right. AI is a meaningful part of how we deliver, and we are open with clients and partners about its contribution to our work.

6) Thoughtful and proportionate use

Guided by our value of Committed to Better

We choose AI tools and approaches to suit the task, considering their benefits, risks, costs to clients, and environmental impact. We equip our teams to use AI responsibly through practical training and clear internal guidance. We regularly review our practices as the technology and our understanding evolve, applying what we learn to improve how we work. We bring a metered mindset to AI use, designing workflows to limit unnecessary computation, repeated processing, and excessive outputs. This helps us to minimize resource consumption and environmental impact while maintaining the quality of our work.

Queries

If you have any queries about Caribou’s AI approach, please get in touch with our AI Steering Group (ai-steering-group@caribou.global)