Now Reading
Artisanal Vs Modular Approaches While Building An AI Function

Artisanal Vs Modular Approaches While Building An AI Function


Building an AI and advanced analytics function in any organization is a challenging part as it requires extensive involvement in terms of building, managing and growing the team into functionality that suits the company’s requirements. Hiring an advanced analytics team can be entirely different from hiring a developer team, which requires specific skills and domain expertise to be scanned by a candidate.

John K. Thompson who is the global head, advanced analytics & AI at CSL Behring, a leading biopharmaceutical company takes through an interesting journey of how to build an AI function that suits company requirements. With more than 30 years of experience in the field, he has worked with the likes of IBM and Dell and has published a book with another book which will be published in June 2020. 

Register for FREE Workshop on Data Engineering>>

He takes us through artisanal and modular approaches of building an analytics team and how to select a starting point for it. While many companies like to take on one of these approaches, there are several companies that are resorting to a hybrid approach as well. He takes us through how a high-performance analytics team can be managed in the best way. 

Artisanal Vs Modular Approaches

Two most popular approaches to building an advanced analytics team are artisanal and modular. Whichever approaches the companies decide to go with, it is highly important that the advanced analytics team consults the subject matter experts before proceeding. “Being a biopharmaceutical company, we never advance to next developments without consulting subject matter experts,” says Thompson. 

The artisanal approach involves hiring experts with a wide understanding of the subject. A single hire is responsible for everything in a team from selecting a model to putting it into production and yielding results. Hires under this approach are expected to pose skills such as good communication skills, understanding business and data, BI tools, data handling and are essentially multifaceted. 

“We at CSL Behring hire graduates, MBAs and PhD under this approach as Business Analysts and Data Scientists. We usually like to begin with hiring interns so that we can analyse all the skills in the individual,” says Thompson.

It is difficult in a traditional hiring approach to know if a candidate possesses all the skills required for the job and whether they can do all the expected work such as building models, interacting with the operational team, engaging with professionals, among other things. Therefore, hiring an intern works best in many ways. 

Thompson shares that this is the best way of hiring for most companies if they are looking for hiring a limited number of professionals. It works well as it can get quick results. 

See Also

A modular approach, on the other hand, is good when it comes to hiring in greater numbers and for specific tasks and functions. Under this approach, a candidate is expected to carry one part of the entire functioning in a defined manner. For instance, a candidate is expected to either be good at creating algorithms or having a business conversation but not necessarily both. This works well if the company has huge analytics capabilities. 

The best approach however as Thompson shares is the blended approach where the best of both artisanal and modular hiring is picked. “To build a real-world analytical process we have a blended approach where we keep moving across various functions based on tasks in hand,” shares Thompson. 

Challenges While Building Analytics Team

To build an advanced analytics function, there are several challenges such as:

  • Friction between the teams
  • Teaching and training the team
  • Achieving results in the given deadline
  • Working and coordinating with various teams in an organization
  • Changes in AI models such as in the COVID scenario where the current AI models started getting slow and had to be refreshed to make it work in the current scenario with the given data
  • Training challenges where the team needs to understand how the given AI function works

Wrapping Up

Analytics teams are dynamic and need to grow and evolve with time. To retain the talent it is important to engage with them. Thompson also stressed on the fact that synchronising the needs of the production cycle with the modelling team’s ability to deliver is a key to success.

What Do You Think?

Subscribe to our Newsletter

Get the latest updates and relevant offers by sharing your email.
Join our Telegram Group. Be part of an engaging community

Copyright Analytics India Magazine Pvt Ltd

Scroll To Top