How to implement AI – and lessons learned

We recently spoke to Glen McCracken, Head of Data, Analytics & Automation at ION Analytics, who shared his frontline experiences on what it takes to unleash the benefits of AI. 

Glen has spent more than a decade leveraging data, AI, machine learning and RPA to help drive organisational growth and efficiency across a variety of industry sectors, including investment banking, digital media, telecoms and sports entertainment. 

For the last three years Glen’s has headed up a specialist team at ION Analytics that’s dedicated to democratising data access and enabling more data-informed decisions to be made. His real world insights on how to get started with AI make for fascinating reading.

The starting point for AI is data. Can you explain?

Absolutely. I’m lucky enough to work for a company that is passionate about data, has over 35 years worth of transactional data to work with, and well established disciplines around everything from data governance to implementing reproducible data pipelines.

That said, when it comes to applying AI, automation and analytics to benefit our clients and boost our operational efficiency and growth, we had to go through much the same thinking process as organisations just starting. Because whatever shape your AI project takes, everything starts with data. And by that, I mean clean data that is in a usable state.

For us, that meant figuring out how to extract and clean and structure data, how to combine it with other data sources to derive new insights and making it accessible for users.

How do you go about making it all happen?

It seems obvious to state it, but first you need to undertake a voyage of discovery – what data do you have, where is it stored, and who has access to it?

Next you need to create a dedicated data team built around four key roles:

• Data engineers – who bring data in, make it usable, and feed output into workflows. 
• Data visualisation experts – who utilise tools like Power BI to deliver new data capabilities to users – that includes predictive insights.
• Data scientists – who focus on finding new ways to get even more insights and value from data. For example, using GenAI to enrich data and embed these insights into workflows.
• Data analysts – who tackle ad hoc requirements that include complex insight requests from senior business stakeholders.

Is this a case of cleaning the data and we’re done?

I’d love to say that’s the case, but the more we use and rely on data in our AI models, the more we’ve come to realise our data is far from perfect. As a consequence we’ve instituted feedback loops and verification checks to improve our data cleanliness. In effect, we’re applying a Kaizen-like mindset to create a continuous improvement cycle.

By working on progressively more sophisticated AI projects we’ve become more ‘consciously competent’ as we learn more about what data quality looks like for the business. In parallel, more stakeholders across the business have become more interested in data quality.

In other words, we’ve effectively crowdsourced data quality and now benefit from valuable insights from subject experts across the business. All of which helps us to fine-tune and refine the quality of our data.

How did you go about defining your AI strategy?

I often say that having an AI strategy is about as useful as having a PowerPoint strategy. AI is a phenomenal tool, but it is really just an enabler, one that needs to be applied in the context of solving real-world business problems. I suggest considering your business strategy, then look at how Data and AI could as tools best enable that strategy.

I’m a strong believer in Kotter’s Change Model, which is why I recommend starting with ‘pull projects’ that are compelling candidates for AI. These types of projects provide the ideal opportunity to develop the change vision in collaboration with business stakeholders who are more likely to provide the buy-in, commitment and support that will be needed to assure project success.

Can you give me a real-life example?

When our marketing team wanted to move away from generic messaging and tailor messages to specific customer types, we held a hackathon to find ways AI could help hyper personalise email campaigns in an extremely efficient way.

Using traditional AI to segment and classify customers was just the start. We also utilised GenAI to create highly nuanced and targeted copy. And the project outcomes have been impressive, with campaigns now generating significantly higher click-through rates – up to a seven-fold increase in some cases. 

By taking the robot out of the human, we’ve helped our relatively small marketing team become a highly effective lead and revenue generation machine that punches well above its weight.

What tips can you share for organisations just starting out with AI?

Get help. The payback you’ll get by using experienced experts to drive your first few projects will be worth its weight in gold. Rather than going through the painful trial and error process, you’ll start reaping the benefits fast. Plus, you’ll be able to maximise knowledge transfer to your teams. After that, it’s a case of getting better by doing it and we’ve found hackathons are a great way to ramp up your expertise and ensure your people build skills. 

There are some amazing AI tools out there, with more on the way. There’s never been a more exciting time to start working with AI to do some truly creative things.

Looking for insights on real-world AI use cases? We can help. To hear more from industry experts like Jon and Oliver, check out our Team Takeover podcast series.

The post How to implement AI – and lessons learned appeared first on 101 Ways.