Why set up an AI lab?
What is an AI lab?
AI labs turbocharge value creation by defining AI use cases. An AI lab brings together:
- Multidisciplinary skills: experts who are both business minded (product owner, business owner etc.) and technical (data scientist, data engineer, data architect etc.) and work together to achieve use cases.
- Methodologies: Design Thinking and Agile. This allows quick movement from idea to concept, from concept to prototyping, from prototyping to production and from production to scale deployment
- Technologies: tools that allow for the creation of efficient and scalable technical solutions
The major issues
Artificial intelligence initiatives are often dispersed throughout organisations. Most remain at the proof of concept (PoC) stage, ultimately delivering no value.
These PoCs are rarely part of a wider strategy, meaning that potential value is left unexploited.
AI labs have two major challenges:
- Creating value at scale by prioritising use cases intelligently, accelerating their industrialisation and scale
- Promoting knowledge sharing and increasing skills among employees.
Artefact’s Top Tips for establishing an AI lab
#1 Choose the right use cases
Balance different business areas and different value layers.
Go for quick wins – both to ensure a true incremental value and to facilitate implementation. It is more important that the first use cases succeed than that they deliver the greatest value.
Be pragmatic – choose the easiest option on both a technical and organisational level (choose use cases with the least friction).
#2 Organise feature teams
Establish agile teams, each in charge of a major business problem, driven by a KPI. For example a retailer might create an agile performance store team and an agile performance supply chain team.
Divide these agile teams into feature teams, each in charge of a sub-problem. For example, the agile supply chain team can be sub-divided into a sales prediction team, a team focusing on automating warehouse work etc.
#3 Break down the complexity
Segment each sub-problem into basic sub-units.
For example, the feature team in charge of sales prediction is initially interested in predicting sales of fruits and vegetables and more specifically in the prediction of tomato sales.
#4 Build Skills
Establish processes which promote sharing knowledge and skills, for example:
- Tech talks: organise weekly events for feedback from a team on a specific point (a tool, a challenge etc.)
- Pair programming: set up pairs within the feature teams, working together on the same code
Create a lab academy to run a training program within the lab.
#5 Make it scalable
Be able to rapidly increase the capacity of lab teams by adding feature teams or creating new feature teams.