It has been eight years since we started Signal AI and during that time I have been learning, discussing and sharing my experiences on how to scale AI-first companies. Recently, some of my thoughts were shared in a Forbes article where I focused more on the human aspect. My intention with these two blog posts is to expand on those ideas, providing my reflections and learnings from two perspectives. Firstly, focusing on the company as a whole and what aspects are critical for scaling up. Secondly, what are the main learnings for the Data Science function, in particular to scale an AI-first company.
I believe focusing on a Data Science function (or a set of product teams using Data Science) will be incomplete, and potentially dishonest, unless we start by acknowledging that without the company changing as well, the impact of those improvements will be limited.
Clear and codified vision, strategy and values
As AI-first companies scale, they experience a number of challenges ranging from continuing to attract top talent to the growing expectations when raising capital. However, ignoring major external events (e.g., COVID), the biggest challenges tend to be related to the internal organisation of the company. With a growing number of people, communication challenges, lack of focus and misalignment between teams are the main risks to be aware of. How the company evolves its structure and defines its vision, strategy and values are critically important. Without them, even the best of technical teams will not be as impactful as they should be.
In order to minimise these challenges, the best approach is to have a coherent and clear vision and strategy and constantly communicate these company-wide.
People growth, knowledge transfer and company culture
Another factor related to a growing AI-first company is that the knowledge will necessarily become more siloed and probably also more specialised. At the inception of a start-up, a few people understand and touch almost every technical part of the system, but this quickly changes as the company grows. At this point, the internal team dynamics shift with new roles appearing and people having to adapt to these changes. This surfaces communication, cultural and context-awareness challenges. For instance, some team members will struggle to let go of specific parts of the system (as described perfectly in Give away your Lego”) and others will find their personal development in the organisation unclear or unfulfilling.
At this point, companies should invest more time on their people, infrastructure and processes in order to grow efficiently and effectively. This includes better hiring frameworks, training for new managers and clear career paths across the organisation.
Agility, measurement and Infrastructure
In general, the smaller the company, the more agile it is, being quicker to change and adapt. This usually changes as the company grows, becoming (or at least feeling) “slower than before”. This is an uncomfortable place to be as the people in the product side of the organisation (where I include Data Science) feel frustrated about their own speed while other functions (e.g., commercial) might not completely understand why everything seems to take longer than it used to.
One way of reducing this problem is to be clearer and communicate more often what is the main focus and the specific goals and metrics for the company and its teams. This will not only help company coordination but it will show the progress towards the most impactful initiatives. Another approach to multiply the impact of people in the product and development teams within an AI-first growing company is to invest in data infrastructure and automation, for example building the capability for teams to run AI experiments effectively at scale.
Scaling a company is a tremendous effort in any situation but I believe that the journey will be much easier if companies invest in three key aspects: Clear direction and measurement, upscaling people as the company grows and improving the underlying infrastructure.