Even after years of excitement and enthusiasm, just one in four organizations with C-level analytics leaders have a fully unified, enterprise-standard approach to delivering AI projects
Despite the huge amount of attention AI technologies have received in recent years, we are still a long way from the sort of broad, human-level AI that was envisioned when the term was first coined in 1956.
“I think ‘general AI’ is still one or two generations ahead of us,” says Bart Pietruszka, CDO and Head of Analytics at HSBC. “We have to be mindful of the fact that AI has been a widely discussed topic since the 20th century and it hasn’t been [achieved] yet.”
AIs that are more intelligent than humans may still be the stuff of science fiction. But our survey of 104 CDOs, CAOs and CDAOs shows just how far the discipline has come.
“We have to be mindful of the fact that AI has been a widely discussed topic since the 20th century and it hasn’t been [achieved] yet”Bart Pietruszka, CDO and Head of Analytics, HSBC
More than 99% of the executives we surveyed are interested in developing AI capabilities. Meanwhile, 87% have at least started developing their ‘proof of concepts’ and 69% have deployed AI capabilities in specific regions or business units.
What’s more, 63% of these executives have started scaling AI within their organizations and 45% have established efficient processes for developing, testing and scaling AI capabilities.
“We say ‘the future of AI’, but we have it all around us every day,” notes Nirali Patel, CDAO at AXA PPP Healthcare. “It’s not something new. It’s just something that’s evolving and helping organizations and helping people live better lives.”
In short, AI is entering its adolescence. Many enterprises have now succeeded in preparing the ground for AI to flourish and are applying the technology to drive real ROI. But with three in four respondents saying they still lack a unified, enterprise-standard approach to ‘doing AI’, true AI maturity is still a way off for most.
Enterprises are Laying the Foundations for Operational AI
While virtually all enterprise data and analytics executives are interested in AI, some are further down the path to AI maturity than others. Our research shows that CAOs are the most likely to have started scaling AI capabilities in their organizations, compared to CDOs and CDAOs. A full 71% of CAOs have reached this stage of the AI journey, compared to 64% of CDOs and 55% of CDAOs.
“The CDOs I interact with often have more of an IT and architectural focus, with less of a focus analytics,” says Scott Zoldi, CAO at analytics company FICO. “But if you go to a CAO, they tend to be experts in AI and machine learning.”
Similarly, different industries have different appetites for AI adoption. Financial services AI leaders are the most likely to have established a unified, enterprise-standard approach to AI, closely followed by those in retail, telecoms and consumer goods.
“There has to be a standard of responsible AI development for the enterprise, which should be sanctioned by the CAO, and there should be checks and balances to ensure that it’s followed”Scott Zoldi, CAO, FICO
It will take time for the remaining 12% of AI leaders who are interested in AI but yet to start their AI journeys to put the right data foundations in place. Enterprises that launch into AI development without sound data governance, data management and data quality supporting processes in place often run into serious trouble.
“The architecture to scale AI is complicated,” explains Dante Tellez, CDAO at insurance giant Chubb. “If you want to scale AI, it must be embedded in your operations. So, if you’re having problems with your transactional systems, it is hard to think about end-to-end AI models or algorithms for your operation.”
At the same time, there is no one data architecture template that will work for every organization. Plus, AI leaders will naturally need to make different decisions based on whether they’re prioritizing long-term scalability or driving value in the short-term.
How to Ensure AI Teams Drive Value Efficiently
Another sign of the general immaturity of enterprise AI is the lack of consensus about the best way to structure an organization’s AI teams.
Our research reveals that 85% of data-driven enterprises have a dedicated AI function. Where there are dedicated AI teams, 74% are either organized as one centralized AI team or with a single center of excellence coordinating decentralized teams. However, the other 26% have chosen to embed AI teams into specific business functions instead.
“Developing enterprise AI is hard,” notes Zoldi. “There need to be standards to support responsible AI and those standards must be applied and governed equally across teams.”
“So, I have a view that it all needs to be centralized,” he argues. “There has to be a standard for responsible AI development for the enterprise, which should be sanctioned by the CAO, and there should be checks and balances to ensure that it’s followed.”
Three quarters of the executives we surveyed say they have developed streamlined model development-to-production environments to ensure they’re using AI efficiently. Meanwhile, 48% have established standardized processes to help them ‘get it right first time’. What’s more, 45% have defined clear ‘dotted line’ reporting structures using an AI skills and resources matrix.
“I’ve seen some organizations take a start-up mentality to their digital and AI parts,” adds Patel. “So, they accept that their historical or legacy business sits on one side of the fence. Then, they’ll start a new thing that’s AI-driven or AI-embedded and slowly wind one down and the other one up.”
However, given that our respondents said 43% of the AI projects they have identified a need for are not yet fully deployed, some might rightly question how many AI teams have the right mix of processes and organizational structures in place at present.
Today, companies like Facebook and Amazon offer a glimpse of what the future might have in store for AI-driven enterprises. These businesses represent pockets of excellence in a world where AI is still broadly going through its adolescence.
The findings of our research should help AI leaders see where there are opportunities to improve elements of their AI functions to accelerate their journeys towards AI maturity.