Skip to main content

How to effectively deliver an AI transformation strategy

(Image credit: Image source: Shutterstock/PHOTOCREO Michal Bednarek)

Artificial Intelligence (AI) has evolved substantially from being a technology buzzword, to the commercial reality that it is today. Companies with an expertise in machine learning (ML) are looking to evolve to create AI-based solutions. Progressive enterprises that are yet to implement a ML learning culture are now aiming to devise their strategies.  Amidst all this hype around AI, and the fear of being left behind, how do you go out about implementing an AI transformation strategy quickly across your company? 

There are numerous challenges in establishing an enterprise AI transformation strategy, and I have aimed to encapsulate them under three main headings. Talent is the first challenge. Organisations need to hire, train, assemble and partner with the right expertise. A team of talented individuals that are capable of driving the AI transformation needs to be created and utilised fully. The second challenge is Time. It is essential to assess how fast you can achieve business results by implementing the AI strategy and create an environment where people can fail fast. Trust in your AI technologies that are underpinned by machine learning models is the final challenge.  The ability to explain the results of your ML models to regulators and stakeholders is key for drive mass adoption of the underlying technologies.

Given these challenges, how can they be addressed?

1. Building a data culture

To effectively utilise the plethora of data being generated, companies need to first build a data-driven culture across the company.  Here are four key points to enable this.

Asking the right questions:  Asking the right questions is key to building a company-wide data culture.  How do I acquire new customers, who would that customer be and how do I optimise my supply chain, are some questions businesses need to answer today. Formulating the business problem is key to any AI implementation.  To develop relevant questions, companies need people who are creative, understand the business constraints they are working in, have an analytical mindset and who can offer answers that are backed up by data, as opposed to a gut feeling.

Data collection: To build a data culture, firms need to begin collecting data proactively. Today, data can be obtained from a wide variety of sources, including the marketing and sales departments, product monitoring and customer analytics.  This data collectively forms the foundation of a data culture.

Make the data accessible: The data that has been collected needs to be made accessible to all appropriate people within the company.  This also means the data should be presented in a format that is easy for people to work with, allowing them to glean meaningful and actionable insights.

Find the right talent: Data is a team sport. This means that while companies need data experts to build models and algorithms, they also require people with different technical abilities that allow them to uncover useful insights from the data, before passing it on to the experts. Doing this can help train the existing workforce, since they have the essential domain experience for the job. ML is as much of a cultural transformation as it is a business one. Instead of rebuilding an entire team from scratch, companies should look to hire several data scientists and utilise the existing pool of experienced staff to assist them.

2. Time to deliver an AI strategy

AI and machine learning is being implemented across all industries, ranging from Know Your Customers (KYC) and Anti-Money Laundering (AML) in financial services to early Cancer Detection and Personalised Prescription Matching in healthcare to Customer Churn Prediction and Master Data Management in telecoms to Personalised Ads and Credit Scoring in marketing and retail.

Using AI across these different industries can save time and money, whilst also gaining a competitive edge.

Determining outcomes: Asking the right questions determines what outcomes can be generated from any specific application.  The main idea here is to translate the high-level goal of your company into a business problem, and subsequently determine the outcome.

Measuring Success: Companies must also identify metrics that can measure its success. The definition of success may vary for different companies, but the end goal remains the same; making a profit and delivering value.

Connect With the Community:  Community plays a vital role in driving change in any company. There are many ways to connect with the machine learning community, including online and webinars, as well as offline at meetups, when the time is right again for that.  These will enable community members to exchange knowledge and learn from each other.  Learning from each other, participating in sessions and sharing relevant insights are great ways to connect with the community, regardless of where you are. There are machine learning communities everywhere around the world, and there may be a local chapter right next to you.

3. Trust in AI and technology considerations

Machine learning models should not be seen as ‘black boxes'.  This means that we should be able to explain them coherently and identify the logic behind its predictions. Being able to describe the model’s decision adequately, having sound documentation and eliminating bias from the results are key considerations for companies, in order to instil trust in AI. Deciding what technology to use is important, as this can have a profound business impact.

Open Source or Proprietary software: When companies embark on an AI journey, they may need to decide between open source or proprietary software, or perhaps even both.  Many existing pioneers in machine learning and AI regularly open-source their technologies, which could be a good starting point for others.  As these new AI players mature, organisations may need to evolve to the commercial platforms that are made available.  

Cloud or on-premise environments: This depends on how soon you would like to begin. If you are starting from scratch and do not have an existing DevOps system in place, it’s easier to get started on the cloud. This eliminates the need for procuring and setting up software, as well as security, infrastructure and maintenance issues.  However, if you do already have a decent DevOps infrastructure in place, the on-premise option can help optimise costs. Many companies also prefer the hybrid model, where they are able to switch between cloud providers and on-prem, which is good practice.

Data: Once again, data is a critical point. Understanding how to generate, save and make data accessible is of paramount importance.  Areas such as data privacy, data governance, security and data lineage are some of the points that need to be appropriately addressed by companies.

What next?

So where do you go from here? By assessing the big three challenges, Talent, Time and Trust, and how they might be addressed, by working with your team, companies can get a sense of direction about where and how to kickstart their AI transformation journey.  Identify the problems that you are trying to solve currently and evaluate how you can leverage machine learning and AI to give you a competitive edge. An AI culture needs to be developed, and, like any important task, requires an investment of time, patience and resources.

John Spooner, head of Artificial Intelligence, EMEA, (opens in new tab)

John Spooner, head of Artificial Intelligence, EMEA,, leads the EMEA technical and data science teams, in advising organisations on democratising artificial intelligence and embedding it into their business decision-making.