iMerit on the AI in Business Podcast: Key Takeaways

November 24, 2021

When it comes to the realities of AI adoption and deployment, it’s often the service providers  – executing the hands-on work within the enterprise – who have an in-depth understanding of the challenges that may arise.

On the AI in Business Podcast, “No One AI Tool Can Solve All Problems”, hosted by Emerj Artificial Intelligence Research CEO Daniel Faggella, iMerit Chief Revenue Officer Jeff Mills walks through from his perspective at the juncture of data value and workflows, what it looks like to define a business problem, to determine the right tools, and to bring the project to life.

The podcast is also available on, or listen to it on iTunes, Stitcher, Google Play or Soundcloud.

Here are 5 key takeaways from the session:

  • Areas of priority will depend on a company’s level of AI deployment and maturity. Companies that are more mature in terms of their AI deployment, their focus is mostly streamlining and validating their workflows. While companies just getting started with AI projects the workflow would be much more hands-on.

  • There is no one AI tool that can solve all problems. One needs to understand which tool is best for a specific use case. Leading AI companies, besides working with their own tools internally and working with their partner tools may also need various other tools that are top of the game in the sector they are working in.

  • Challenges that AI companies need to be ready for and be able to overcome. Companies need to understand how to manage workflows between the prototype stage and the deployment of systems. Few key areas would be – how to manage the dataset already created under one use case to potentially move that use case over to a different model, edge case management and human-in-the-loop.

  • Human involvement comes into play at different parts of the workflow. Depending on whether the project is primarily using a data-driven or self-learning approach, human involvement is required at different stages of the workflow. When a data-driven approach is being used, human intervention is very important at the beginning of the process, the least necessary in the middle, and slightly important at the end to verify that the quality meets project expectations.

  • Companies need to understand whether they are looking to build something proprietary or just simplify workflow. If it is a core capability that will distinguish the company in the market and is a part of real strategy, teams need to think about how they can potentially own it.