How to start building your AI capability – Tips for the CEO

Recent months have seen apocalyptic warnings on the dangers of artificial intelligence. But “AI transformation expert” Dean Zia Dar isn’t panicking yet – and says business leaders shouldn’t either. He offers these tips to help you standup AI capability in less than a year.

1. This is how AI helps you

“Don’t let the technology and the hype scare you from investing in AI and reaping the benefits,” says Dean. Fundamentally, these are the same statistical models we’ve been using for 150 years, but because of advances in computing power, we can now process so much more.”

Even if you don’t understand how AI or machine learning works, you at least need to be very clear in your understanding about what it can do. And that, says Dean, is about maximising  your understanding of customer behaviour. “It’s about using historical data to see what opportunities you’ve missed and how you can respond. Then, it’s predicting at a granular level how those customers are going to behave in the future.

“Financial forecasts, supply chain optimisation, product or service pricing – they all come off the back of this, but it’s fundamentally about generating clear insights and actions based on behaviour.”

Dean has been leading an AI transformation programme at one of the UK’s largest providers of safety equipment. They operate in the B2B and B2C markets and by using AI to connect data across sales, costs, inventory, etc, the company can now always offer the optimal product price for each micro-segmentation. It’s also driving product rationalisation, by helping the firm know exactly which lines to focus on to boost revenue.

2. Start with a value creation plan – not the tech 

“When I go in and talk to CEOs, they’ve often been sold an AI platform and now they’re looking for a problem to solve – it’s the tail wagging the dog. Don’t start with data or technology; start with a value creation plan and what you’re looking to achieve as a business. Is it revenue growth, market expansion or  increasing your margins? What are the levers you can pull to get there and where is the data that can enable decision making?”

AI can be a powerful aid to achieving what every business wants: to understand their customers better.

3. It doesn’t take many people

Dean says a common mistake is to sit AI with the IT team or under the CIO. He argues that companies should have their own dedicated team under the leadership of a chief data (or data and analytics) officer, who in turn works closely with the CEO. Under the CDO there should be a team including data scientists to write the models, data engineers to build the pipelines and systems to connect to your data lake, and platform engineers to create the fundamental scalability, security, storage and necessary integrations across the business.

“You don’t need lots of interim managers or consultants on this,” he says. “They’re fine to help you get started or to accelerate the pace to a point that the business can use AI as a tool, but you will eventually need a permanent team who understand your business and can give you the clear, elegant insights that allow you and your team to make quick, nimble decisions. For a 1,500-person organisation, you probably need 20-30 people, but for 10,000, you only want 50-60 – it’s not a linear scale.”

4. Build in Go / No go gateways

Embracing and capitalising on AI doesn’t have to come hand in hand with huge commitments and huge budgets. The key is to build in “go” and “no go” Gateways and take it one step at a time. This will allow you to prove concept and take benefits along the way, these benefits will then add more weight to the business case.

If done correctly AI and machine learning will become an integral part of your business but remember that the fundamentals are based on statistical models. Some of your initial spend will be to improve both the tech stack and models and this will inevitably bring the early business benefit before AI and ML come into play

5. Test, learn and refine

AI takes data, connects it and gives you insights and actions. But, he says, models need testing and assessing. What happened when you changed the price of a product for a certain customer segment? Were they happy? What were the unintended consequences?

“When I was working with a large mobile comms company, they had models that had been running for 10-15 years but hadn’t been tested for a long time. Customers weren’t being given the right offers, so by rebuilding the models and testing them, we were able to jump from 50% confidence in the models to 80%.”

6. Don’t model on bad data

“If you have poor-quality data held in different departments under different names, it doesn’t matter how good your models are. This is something your data team should be able to untangle, but it requires executive sponsorship to overcome any barriers.”

7. Don’t wait to do this or you will lose ground

Dean says to stand up a team and AI platform should take three to six months, and up to another six to start deploying and testing the first models. By following the right processes, building the right capability and keeping things simple, AI can be a powerful aid to achieving what every business wants: to understand their customers better.