'Build in smaller chunks': How to build an AI roadmap

'If we don't have control, we'll always be lagging behind'

Nearly every IT leader is investigating AI, but where do they begin with implementation once the fact-finding is done?

Modern technology changes so quickly that it's no longer good enough to roll out new tools piecemeal; you need a strategy.

"Technology moves so fast that if we don't have control over it, then we'll always be lagging behind - we'll always be in a catch-up mode," said Chetna Bhatia, CIO at B2B energy supplier Pozitive Energy (PE).

"It is good to have a technology, but you need to have a clear roadmap of what you want to achieve and when...

"The first thing to keep in mind is: what is the objective? What gaps are you filling for your business, and what are the smart, measurable objectives that you want to achieve? You need to first come up with that."

PE has been rolling out AI for its predictive and analytical powers: identifying peaks and troughs in demands to help balance supply, and predicting fault detection and outages. Without that clear use case, says Chetna, there would have been no point.

"Just because everybody's talking about [a technology] doesn't mean you have to apply it."

Know yourself

So, you've identified your objectives. Next, you need to work out if they're achievable.

"Look at your data, look at your technical resources, look at the technology you are using, then look at the systems that are available and [find out], ‘Where is the gap?'

You should also know "what kind of technical resources you need [and] what type of data you need," as well as whether you have that data and how you will manage it. That's especially key for AI strategies, which relies on robust data sets that might contain confidential information.

"Build out your implementation plan, then build the robust data management practices. That is very important, because your data is changing every day. How will that get integrated in your AI projects? How will you make sure that any data nuances or data cleanup that is needed is all done?"

Know your projects

Every leader needs visibility of the projects in their domain: what they are, how long they'll last and what the outcome will be. AI, which (given the right data) is very good at predictions, makes that a little easier, but AI projects themselves can still be a stumbling block if you approach them incorrectly.

IT is often run as a series of big, multi-year projects, which might not produce anything for months. That's not suitable with the speed of modern tech development.

"Have smaller projects. Do it in phases, do it in smaller projects. Have your clear objectives and say, ‘Okay, in three to six months, what can I do to achieve this objective?' Start with manageable smaller projects and then slowly build over it.

"If you start thinking, 'I need to achieve this and it'll take me three years' - things might change in six months' time! [Technology] is changing so fast and so rapidly. You might have better tools in nine months that you can take advantage of.

"So, build it in smaller chunks."

Know the tech

The biggest single thing you can do with AI is to "keep abreast with the changes that are happening." PE has an entire team whose main responsibility is to research new technologies and the benefits they could bring to the business.

"Technology is fine, but what is the endgame? Every technology should be considered as, ‘What benefit will my business achieve? What KPI is there for the business?'"

It's not only the IT teams that need to know the technology. Once you've built or bought an AI tool, training the staff who'll be using it is essential.

This issue is becoming widespread: business leaders want to roll out AI, but nobody wants (or knows how) to train staff in its use. Instead, the tools sit, unclicked, in the toolbar – a waste of time and money.

Rather than solely relying on a top-down approach, where employees are formally coached on new tools, Chetna advises also adopting a cross-functional one where teams train each other.

"[IT, for example, can] give training to the business teams to say, ‘Okay, I have this intelligence, you can take advantage of this and then you can improve efficiency.'"

She says this should be a CIO's primary goal:

"CIOs need to think about how to reduce costs, how to bring efficiency, so that they are not considered as a cost centre, they are considered as revenue contributors. That is what we try to achieve."