top of page

How to setup a organization for becoming a AI first company

What it takes to do and which positions are involved.


Position AI talents

First things first: Position AI talent where they can learn about real-world problems. Designing an organizational structure that positions the best data science and ML staff close to business units makes it an AI-First company. Give these talents tools to help you deliver results quickly. 

"AI first companies put AI everywhere"


Learn in the next step how teams are put together in an AI-first company.


Centralized AI Teams
vs. Decentralized AI Teams

Centralized AI Teams

Centralized AI teams have an executive-level leader, such as an chief data officer, who manages all of the data science, analytics, and ML people in the company. That CDO collaborates with the chief technology officer (CTO) and the chief information officer (CIO) to decide which data infrastructure to use. Requests for data, reports, analytical tools, and predictive models go to this central unit, and the unit decides which requests to fulfill. Companies often choose to centralize at the start of their path to becoming an AI-First company. 

  •  Benefit

The key benefit is centralizing decisions about data infrastructure, pipelines, and projects so that data scientists and ML engineers can collaborate effectively on hard problems. 

  •  Drawbacks

The drawbacks are the reduced speed of making data available to the rest of the company, limited access to resources from the core information technology budget or team, and lack of knowledge transfer from data scientists to domain experts. 

Decentralized AI Teams

Decentralized AI teams have a business unit leader, such as a general manager, to manage the core product delivered by that unit. Data scientists and ML engineers work on projects for and with others in that business unit who have both domain expertise and consistent access to customers. 

  •  Benefit

The combination from domain expertise and access to customers can be a key benefit, given that AIs often need heuristics from domain experts, data from the real world, and feedback from customers in order to reach a sufficient degree of accuracy and thus utility. 

  •  Drawbacks

The drawbacks can include using shared IT infrastructure that doesn´t have the features required to build intelligent systems, fragmentation of data, inability to get complementary data from within the company, limited executive engagement with AI-based projects, and lack of knowledge transfer from domain experts to AI experts. 


The middle of the spectrum looks like a combination of the two, with some degree of coordination in hiring AI-First teams, placing them across the organization, setting up data infrastructure, managing data, and deciding on projects to pursue, but ultimately deferring to business unit managers on how to utilize data and ML specialists. AI-first companies distribute AI talent across their organization, have AI working behind all of their products and use AI-enabled computing infrastructure. They also have decentralized data science and ML functions, and an executive team that sets data strategy for the whole company. However, not all companies start as AI-First companies, so different companies may be at different waypoints on this journey. 



Data Scientists are independent Contributors

CDO leads a Team of Data Scientists



Organizational Types

Data Scientists

Business Units



Most of the previous user - startups, large enterprises or general public companies- are all at different points on their way to building AI-First companies, but the difference between those are moving fast and those who aren´t is the mindset: the ones who are constantly integrating the industrial with the technical are building AI-First companies. 


Embedding AI talent across an organization is a management challenge as hard as any other, but one worth solving in order to be a key figure in the AI first Century.

Building AI-First teams is the heart of an AI- First company, but there is no perfect template to follow; no one right way of doing things. There is, however, an approach and methodology by which companies can find suitable candidates, suitably manage them, and structure an organization around them. Developing a fine appreciation of the tasks involved helps to focus recruiting and may go beyond those with just a computer science background. Understanding the work they do helps in managing and supporting them, and ultimately enables embedding them across an organization.

  • AI first companies need a diverse group of people to manage different technologies.

The competencies required to manage AI technology are captured by titles such as data infrastructure engineer, data engineer, data scientist, ML engineer and ML researcher.

  •  AI first teams need people with different skill set.

Specialisation in databases, statistics, mathematics, physics, econometrics and economics all help to understand the fundamental principles of AI and run it on modern computing infrastructure. 

  •  AI first teams need AI first tool like

Notebooks, framework, cloud services, tools created for data scientists and machine learners help these teams get the job done.

  •  Engineers and data scientists need a different type of manager. 

Data scientists ask questions like researchers. They deliver results as numbers in spreadsheets, graphs in presentations, models in code, or discussion in person. They don´t get requirements from business users, so they require regular meetings. They may need a larger budget and ad hoc approval of purchases- with commensurate oversight. This is different from a software engineer, who receives clearly specified goals and deliverables from business users. 

  •  AI first companies put AI everywhere. 

AI-first companies distribute AI talent across their organization, have AIs working behind all of their products, use AI-enabled computing infrastructure, decentralize data science, and have and executive team that sets data strategy for the whole company. 

bottom of page