Professor Thomas Davenport Showcases Holistic AI’s work with Unilever

Distinguished Professor Thomas Davenport and industry thought leader Randy Bean showcased Holistic AI’s partnership with Unilever in the article, “AI Ethics at Unilever: From Policy to Process” for the MIT Sloan Management Review. The article exhibits Holistic AI’s partnership with Unilever as a leading example of the effective implementation of responsible AI at scale.

The consumer goods juggernaut, which encompasses over 400 brands from Vaseline to Ben & Jerry’s, has consistently operated some of the most cutting-edge supply chains and customer service channels, leveraging technologies like Artificial Intelligence (AI). In their article, Davenport and Bean showcase the Holistic AI-Unilever partnership as a case study for how large companies can ensure their AI systems are not only ethical but effective.

The Holistic AI x Unilever Partnership

Unilever set a goal to only put AI systems into production that had been reviewed and approved – they needed AI assurance. As described by Davenport and Bean, a new pipeline for AI deployment was developed:

1. A Unilever employee or supplied proposes a new AI solution for a given use case.

2. Before it is developed, the AI solution is reviewed by Unilever or Holistic AI depending on the complexity of the use case.

3. The proposer is informed of potential ethical and efficacy risks along with any solutions.

4. The AI application is developed with these risks in mind and Unilever or Holistic AI runs statistical tests to determine how the system performs against bias, fairness or efficacy criteria.

5. If appropriate, Holistic AI tests how the system complies against local regulations.

6. The system is either approved, updated, or rejected under Davenport and Bean’s fifth phase, ‘action.’

Nonetheless, such a process can be onerous and time consuming for a company as large as Unilever. But with Holistic AI, the company could deploy it at scale. Through Holistic AI’s platform, Unilever’s data ethics team reviews the status of AI projects – which includes any prediction or automation function as rudimentary as an Excel formula – and monitors their status. The team can review any newly submitted use cases for the AI system, whether the information is complete, and risk assessment of a system.

Ensuring compliance with Holistic AI

As highlighted by Davenport and Bean, the evolving AI regulatory landscape presents significant obligations to global firms like Unilever, stressing the need for compliance with key legislation and standard such as the European Union’s landmark AI Act, which is set to become a global standard. Companies operating in the United States will be subject to voluntary federal standards – like the National Institute for Standards and Technology’s AI Risk Management Framework (AI RMF).

Partnerships with impartial third-party firms is the “only current way to ensure that AI systems are aligned with human interests and well-being.”

Get in touch at we@holisticai.com to find out how we can help you on your journey towards AI governance.