Machine Learning is undeniably enabled by the technological advances of the past decade. The theory and maths have remained fairly consistent for the last 50 years, but the explosion of cost effective compute power and the availability of data has breathed life into ML as a concept that can be practically applied to day to day problems within an organisation.

However, buying or developing the right technology and tools alone will not lead to instant success. In fact, we heard from one Data Science lead at a major bank that 9 in every 10 AI/ML projects fail to make it to production.

One of the supplementary enablers of ML solutions is people.

So, what can your organisation change in its approach to people to allow ML to thrive?

Establish cross functional teams

This should really be a hygiene factor for any teams, functions or organisations trying to enact technology change today.

Particularly in larger banks, we still find so many examples of technology sitting apart from the business, sitting apart from operations, with very siloed, boxed-in mentalities around roles and responsibilities.

The bank doesn’t need to have embraced agile across the board, or to have mature agile processes and ways of working in place, but forming cross functional teams with the right tools and ethos will help accelerate machine learning projects, as it will most other digital changes.

By placing engineering alongside the product owner, design next to business analysis, and testing embedded throughout, you see that:

• Communication improves dramatically
• Product decisions are better informed by technical capability, and user-centric design
• Teams are able to react much more easily to feedback
• The end product is more stable and the code of higher quality and testing is a core tenement from start to finish
• Business and customer expectations are better managed

Blur the lines between data science, analytics and technology

Following on from the creation of cross-functional teams, and very much helped by it, the boundaries between the data, analytics and technology functions need to be broken down.

The organisation needs to work to develop these core capabilities everywhere, across everyone so that functions within the business can be self-sufficient when creating ML solutions, rather than being dependent on a central, specialised and siloed ML CoE.

This journey will take time and patience, but in the long run will pay dividends. To start, firms can:

• Hire people with a mixed skillset, and create job descriptions that don’t place individuals in confined boxes
• Create flexible organisational structures, no specific, siloed functions
• Encourage and open up data, analytics and technology training for all employees
• Allow access to the right tools across the business. This requires a sophisticated level of identity and access management, but will greatly help to open up AI/ML to everyone

Develop MLOps processes and functions

It’s well understood now that a Data Science capability is essential for designing, building and enhancing ML solutions, but often underlooked is the role that an operational infrastructure team – MLOps – can play to lead to real success.

An operational team can create and maintain the technical solution, taking responsibility for:

• the underlying infrastructure
• identity management
• security and controls
• tooling
• automation pipelines and CI/CD

These platform and infrastructure engineers will know how best to optimise the machine learning workloads to keep costs down, reduce processing/training times and keep the solutions live in production.

This allows the Data Scientists to focus on what they are good at – sourcing, cleansing and tagging data, building and training ML algorithms, and constant model improvement.

To read more about how to allow AI to thrive in your organisation, and how you could manage you AI projects, download our whitepaper here.