Q: How does the "AI-in-the-loop" model enhance the decision-making process in IA?
Luca Sardo: The "AI-in-the-loop" model involves using AI to automate repetitive tasks while keeping subject matter experts involved in the review and decision-making process. This approach enhances the efficiency of IA by allowing AI to handle error prone tasks, such as data extraction or synthesis, while human experts focus on evaluating, refining and verifying the information. We use AI as an assistant, that follows instructions provided by the specialist who then reviews and verifies the outcomes. It's a collaborative process aimed at enhancing human experts' ability to create detailed and accurate reports, utilizing language models for added checks and controls.
An example of the AI assistant's capabilities is during the scoping phase, where it can extract information from the knowledge base regarding projects with similar geography, typology, challenges, or best practices. By providing citations and access to sources, the assistant helps specialists understand the risks and impacts that the project may encounter and need to address.
We can streamline the document review process, verifying its compliance with WSP methodologies and regulatory guidelines and requirements. This step ensures the quality of our deliverables and supports the decision-making process.
Q: Can you give an example of an IA project where AI has been successfully used to help our teams to work more efficiently?
Luca Sardo: Our team recently answered stakeholder questions about an African project based on an ESIA prepared by several practitioners. With 30 documents and thousands of pages, finding answers was challenging. We mapped the knowledge in the document repository, allowing our AI model to generate answers and reference relevant sections. This streamlined the process and stakeholders were pleased with the results.
By integrating these innovative techniques, we significantly reduced the time and effort required to navigate the voluminous content, thereby enhancing the overall efficiency and effectiveness of our response process.
Q: How do you ensure the quality and integration of data when using AI for IA?
Luca Sardo: High-quality data is essential for accurate AI results in IA. We devised a framework that combines AI with a curated knowledge base, guaranteeing reliable and relevant information. This framework involves rigorous data collection and validation processes, ensuring that the data accessed by our AI models is both comprehensive and precise. Additionally, this approach has enhanced our workflows, reducing time and resources spent on manual data processing. Overall, the integration of AI with a well-maintained knowledge base has greatly enhanced quality and efficiency, helping to improve the overall IA process.
Q: Can you discuss any interdisciplinary collaborations that were crucial for the successful implementation of AI in your IA projects?
Luca Sardo: Interdisciplinary collaboration is key to the successful implementation of AI in IA. We worked with numerous teams, including members from our digital, ESIA , IT and legal teams to develop our responsible AI guidelines to map the IA process and identify areas where AI could be most beneficial and safely used. This collaborative approach ensured that the AI tools we developed were tailored to the specific needs of each project, compliant with regulatory requirement and integrated with our quality review processes.
Q: What are the future prospects of AI in the field of Impact Assessment?
Luca Sardo: The future of AI in IA looks promising, with continued advancements expected to further enhance the efficiency and accuracy of IA processes. As AI technology evolves, it will likely play an even more integral role in IA, driving innovation and improving project outcomes. The ongoing collaboration between AI experts and IA professionals will be crucial in realizing the full potential of AI in this field.