Why data readiness is holding back AI adoption
Organisations need to get their data house in order, and that starts with identifying value and outcomes.
Data readiness is holding back AI adoption and business-critical outcomes, according to Jason Frost, GM data and integration, Atturra.
“I've been in so many conferences lately where we’re talking about ‘we almost got there but the data failed us’,” he said.
But things are changing and in the last three months, Frost is finding more organisations are realising there’s a requirement to solve a data problem as part of their AI project.
“Everybody knows they've got what we call ‘ROT’, redundant, old, trivial data that AI doesn't know whether it's redundant, old or trivial, it knows it's data,” said Frost, speaking to CRN Australia from the Boomi World Tour event in Sydney.
However, Frost believes that labelling the problem ‘data readiness’ creates something that’s too big to address because the goal should be to find value. “Everybody will have a data readiness problem, but do you have an initiative or what area [of the business] needs value from the data?”
Moving faster shouldn’t be the goal because if the data's still not right, it won’t deliver the outcomes that will make a difference. Trust is vital.
"I've been using the word trust a lot lately, which is really the new word for governance. Everybody wants trust in their data, which really means you're going to govern it, clean it, secure it,” he told CRN Australia.
Determining the ROI of data not AI
Frost wants organisations to focus on ROI before trying to speed up adoption of AI and decision making. “There's no point being much faster developing something if at the end of it the data's still not right,” he said.
“Knowing the ROI of the data decision for us is critical. Which then rolls back to discovering the data, governing it, which is trusting it, and knowing the boundaries around that,” he said.
It’s a matter of finding the business reason for change and from that finding the value in the data that’s going make a difference. “You've got to work out what the business benefit is and then it becomes an ROI and a business case that stacks up,” he added.
Too often, organisations are getting to the final part of AI going live and realising it's making the wrong decisions on the wrong data. He sees too common problems that lead to the outcomes.
The first is that data that is poorly defined. “AI doesn't know the difference between structured and unstructured data,” he explained.
The second is ensuring data is accurate, trusted and comes from the right systems of record. “You’ve got to make sure you’re using the right systems of record to make the right decision,” he added.
His advice is to work backwards from the value the organisation is trying to create and make sure the data doesn’t block that desired outcome. And there’s an important role for partners.
“As partners, or data leaders, we've got to be more about value assessments for data: ‘What are you trying to achieve and what data do you need?’ Because everybody knows their data's not ready,” he concluded.