How to build an effective AI team

August 10, 20203 Min Read

A cross functional team beats specialists. The team size does matter and bigger is not better.

In this article we want to investigate the above hypothesis. This might help you in your journey to extract business value out of your next AI solution.

Why bigger teams do not guarantee success

The two pizza rule, coined by Amazon CEO Jeff Bezos’ is well known in the industry:

If a team can’t be fed by two pizzas, that team is too big.

Yet many companies still misunderstand the essence of the message. Team sizes have reduced but the number of teams have increased in an attempt to accelerate product delivery.

Having multiple teams in a new initiative does not guarantee business results. Team structures and team domains need to be well defined. Which, getting right, is really hard in the beginning. Amazon, Netflix, AirBnB... can pull it off because of their deep understanding of their business and products.

But even they started small when the uncertainties were high. AWS, the on-demand cloud computing subsidiary of Amazon was not created with 50 engineering teams it grew into a large engineering powerhouse.

Another benefit of small teams is communication

It should come as no surprise that the size of a group effects communication - it is easier to keep in touch with a small number of people. Science also agrees. Harvard Psychology professor J. Richard Hackman introduced the concept of link management. In his findings he notes “as a team gets bigger, the number of links that need to be managed among members goes up at an accelerating rate“. It is managing the links between members that is difficult to do and leads to delays in delivery.

Yet another benefit is short-term wins. Small teams can create visible success much faster. Individual effort increases and is more noticeable. That translates to more engagement and work satisfaction which boosts productivity.

Should companies look for data science unicorns?

A few years back. Companies looked to PhDs to kick start their foray into AI solutions. There is nothing wrong with having PhDs, we at neurocode have one 😂

However, in our experience when it comes to delivering software, and let's be honest AI solutions = software

A PhD with no background in bringing software into production is going to face difficulties.

What about "AI architects" ?

A person that can fill the role of a data engineer, data scientists and software engineers?

That position is incredibly difficult to fill. If you are an AI architect please send us a message 👊 😃

Our experience with AI teams is that having 3 to 4 motivated and capable engineers is usually all it takes.

If the initial AI solution is a success the team can start to expand to 6 to 8 people. Certainly no more than 10.

Forming more teams should only be done when the initial team's code base exceeds its cognitive capacity. That means, if you wake a team member up at night and ask him where the data sources for the recommendation engine are and he doesn't know. You should start adding teams.

Skilled data science are clearly still in demand but a shift is taken place. We at neurocode believe that companies want a team capable of converting jupyter notebooks and AI jargon to business value.

If you are like minded and want to see business value above all else. Contact us.

The business side

Although data scientists and software engineers are the predominant factor for successful AI teams. We believe that having people from the business side to be essential in a successful AI solution.

This is the reason why we like to work with business people when we build AI solutions. Business people understand the problems and the product domain better and can provide invaluable insights. These insights can speed up the initial AI solution.


  • A well composed AI team can deliver business value
  • Team size matters and more teams and more team members does not guarantee success
  • Having a data scientists only department working on AI solutions can work if its filled with unicorns
  • Taking a blended approach will become the norm

Talk to us if you are serious about bringing your AI solution to production.

Spread the

Thirsty for more?