fbpx
Scroll to top

How to build a best practice AI function

Many of the world’s leading tech companies have become “AI first” and others are trying to follow them by setting up dedicated AI functions. But what does a best practice AI function look like?

 

The figure below seeks to illustrate such a function. The layers indicate different tasks performed by employees with different skillsets. The percentages indicated which tasks the employees in each layer spend their time on. To the left is an indication of the relative number of employees within each layer. Each layer interacts with the level above/below them. In addition, the top layer has some interaction with the bottom layer as they ideate on new products together.

  1. Deep Learning Scientists focus on research in machine learning and in particular deep learning, which is the technique that drives progress in AI today. They also teach the layer below them and provide 1 hours weekly sparring on up to five concrete AI projects to infuse them with the latest knowledge. This is important since the development in AI is very rapid and trans­lation from research to commercial value, i.e. revenue, is very short – often just 1-2 years. It results in smarter solutions and months/years of saved work across the more manpower intensive layers towards the bottom of the figure.
  2. Machine Learning Engineers focus on applied machine learning and deep learning. They build AI prototypes and provide project management and sparring to the lower layers. They also spend some time staying up to date on AI developments to remain qualified for the interaction with the top layer.
  3. Data Engineers work on backend, data management etc. including collecting, storing, cleaning and calling data for AI solutions.
  4. Software Developers work on frontend user interfaces, web, apps, design etc. for AI solutions.
  5. Business Developers help understand the customers need and ideate on new solutions, partly in interaction with the Deep Learning Scientists at the top.

 

To create a thriving research group at the top you need approx. 12-15 Deep Learning Scientists, including 3-4 at senior level. As indicated in the figure, each Deep Learning Scientist can activate up to approx. 125 persons in the lower layers. Hence, an AI function with 15 persons in the top layer would involve almost 2000 persons in total.

 

We believe, the companies that are able to set up such best practice AI functions will be in a good position to become the digital frontrunners of the future. But we think only a few Danish or European companies will be able and/or willing to do so on their own in the near future. How others could help them set up such AI functions and thus accelerate the uptake of AI will be addressed in a separate how to briefing.

Dig deeper

Disclaimer: This how to briefing is still being validated. It is inspired by discussions with and material from employees at DTU – the Technical University of Denmark. The initial ideas were provided by Alexander Rosenberg Johansen, drawing on practical experiences from international tech companies.

How to briefings are intended for boards and CEOs. They share the thinking of Digital Hub Denmark, one idea at the time. The content is curated from leading international business thinkers and doers.

Recent posts

How to build a best practice AI function

Many of the world’s leading tech companies have become “AI first” and others are trying to follow them by setting up dedicated AI functions. But what does a best practice…

Bymarianne

27 August, 2019

How to integrate AI into your organisation

Corporations in all industries want to leverage artificial intelligence (AI) to their advantage. But how can companies integrate AI into their organisations?   At this time, AI talent is scarce…

Bymarianne

23 August, 2019

How to understand AI’s place in the world

There are clear indications that artificial intelligence (AI) can create huge value for companies across industries. But how is AI connected to the physical and digital world?   The figure…

Bymarianne

20 August, 2019

How to get the organisation right

Digital frontrunners across industries are adopting new organisational structures, often inspired by software companies. But how do they organise themselves?    Historically, companies have been organised into functional units, such…

Bymarianne

15 August, 2019

How to get competition right

In this digital age, competition between companies is rapidly changing and industry boundaries are transforming. But what does it mean for companies? To understand the competitive and industry transformation taking…

Byadmin

7 August, 2019

How to get data right

Collecting the right data and combining it in the right way has allowed digital frontrunners to unlock exponentially increasing value. But how do they do it? Historically, companies have only…

Byadmin

25 June, 2019

Related posts