Generative artificial intelligence (AI) is quickly becoming part of everyday professional work. For many organisations, the question is no longer whether to use AI, but how to use it responsibly. At INCAS, our experience is that responsible AI use is not really about the technology itself. It is about trust.
AI can support efficiency, structure and clarity. It can help with certain research, drafting and review tasks. But it cannot replace the judgement, contextual understanding and accountability that our experts bring to client work. That is why the INCAS approach is simple: we use AI to support our expertise, not to replace it.
We disclose our responsible AI usage guidelines to clients in proposals, reports and introductory conversations for new projects. We are clear about when and how tools such as ChatGPT, Scispace, and Dedoose may be used to assist our work.
This matters because clients should not have to guess whether AI has been involved in work they have commissioned. They should know, they should be able to ask questions, and they should be able to say no. If a client is not comfortable with the use of AI, or with the processing of data using AI systems, we establish that from the start, and follow their preferences.
For us, transparency is not an administrative detail. It is part of maintaining trust.
Responsible AI use also means putting effective safeguards and controls in place. This includes avoiding over-reliance on AI-generated outputs, protecting client data, and turning off functions that allow uploaded data to be used to train models more broadly. It fundamentally involves erring on the side of caution. These steps may sound technical, but they reflect a simple principle: sensitive information should be handled with care, whether the tool is familiar or new.
AI tools work best when they are used thoughtfully and with disciplined engagement. They require clear instructions, careful prompting and critical review. In that sense, we find it useful to think of AI as a junior research assistant. It can support the work, but it needs direction. It can help organise information, but it does not understand context in the way people do. It can offer a useful first draft, but it cannot carry professional responsibility. The more precise the instructions, the more an expert actively works with the AI assistant, the better the output.
At INCAS, our work is guided by kindness, respect and collaboration. Funnily, that extends to the way we engage with tools. AI is not a colleague, but it learns from how you work with it. Approaching it with clarity, patience and respect, we find often leads to better results than rushed or careless instructions. Whether that says more about AI models or about us is open to debate. Either way, the habit is a useful one. It reminds us to slow down, think clearly, and remain intentional about the work.
Every AI-supported output still requires human supervision. We've seen many examples out there of work that is characterised by lazy engagement with AI - "it sounds good, so must be good". That leads to slop. We make it a point to put in the work. To review AI outputs carefully and critically, checking whether they are accurate, relevant and appropriate, and avoiding the "next of turn" answers. We prompt engineer our AI assistants and projects to engage with our experts in ways that bring out human expertise and usefully challenge thinking. That ensures that the analysis we produce, interpretation of data, and formulated recommendations remain the responsibility of our experts.
This is especially important in conflict, evaluation, social performance and stakeholder engagement work, where context matters deeply and poor judgement can have real consequences. So while AI may help us work more efficiently, it must also be worked with carefully and in ways that preserve transparency, accountability and trust. That is our commitment to our clients and stakeholders as we continue to make ourselves an AI-enabled human consultancy.
How is your organisation approaching that balance?
#artificialintelligence #ethicalai #consulting #evaluation #dataprivacy