Building an Artifiicial Intelligence Explanation Capability

Though companies are building artificial intelligence (AI) systems and integrating them into business operations, executives are concerned about AI's distinctive challenges (e.g., opacity) and seeking to develop new capabilities in response. We describe a new AIX explanation capability that companies must establish before their AI initiatives can thrive. This capability has four dimensions: decision tracing, bias remediation, boundary setting and value formulation. Together, these dimensions help organizations to address the challenges of model opacity, model drift, acting mindlessly and the unproven nature of AI.(1,2,3)