Eric Siegel highlights the chronic under-deployment of ML projects, with only 22% of data scientists saying their revolutionary initiatives usually deploy, and a lack of stakeholder visibility and detailed planning as key issues, in his industry survey and book “The AI Playbook.”

Maximizing the Effectiveness of Machine Learning Deployment

Eric Siegel’s survey and accompanying book “The AI Playbook,” underlines a severe issue in the data science industry: a major lack of deployment of machine learning (ML) projects. In fact, only about 22% of surveyed data scientists reported that their revolutionary AI initiatives are typically deployed.

Key Hurdles in Deployability

  • Lack of Stakeholder Visibility: The information revealed by the AI models often stay within the domain of data science, failing to reach stakeholders who need the insights for decision-making.
  • Inadequate Detailed Planning: ML projects often lack sufficient planning, which is necessary for successful execution and deployment.

Long-term Implications and Future Developments

Addressing this under-deployment issue is critical for an organization’s AI maturity, which could have implications for competitive positioning in the long run. As businesses become more dependent on AI and machine learning to drive decisions and strategies, those capable of deploying AI initiatives efficiently will emerge as industry leaders.

In the future, we may witness companies investing more in establishing robust deployment strategies for their ML projects. This would not just involve incorporating technical aspects like data integrity and model accuracy but also business aspects like clearer stakeholder communication and comprehensive planning.

Actionable Advice

  1. Improve Stakeholder Visibility: Ensure your ML project findings are presented using easily understandable insights and recommendations, making it simpler for stakeholders to comprehend and act upon. Regular stakeholder meetings, briefings, and accessible dashboards are some ways to do this.
  2. Invest in Detailed Planning: Create a comprehensive project plan, including necessary resources, responsibilities, timelines, contingency measures, and expected outcomes, thereby improving the chances of successful deployment.
  3. Integrate Deployment Strategy Early On: Ideally, the deployment strategy should not be an afterthought; instead, it should be integrated with the overall project design from the inception stage. By considering the deployment phase early on and throughout the project life cycle, the overall efficiency and effectiveness of the ML project can be significantly improved.

“The failure of ML projects is less about the technology itself and more about how it’s deployed. Therefore, shifting focus from developing to deploying is a game-changer.”

Read the original article