By: Sid Bhatia, Vice President & General Manager, Middle East & Turkey, Dataiku
The region’s artificial intelligence industry is heating up. Not only has the UAE already appointed a minister of state for artificial intelligence — and become the first country to do so — but other nations are becoming a part of the AI story. Smart city projects are ongoing and PwC estimates the Middle East as a whole could be home to a US$320 billion AI market by the end of the decade.
Figures such as these are signs that AI has gained acceptance. The need to compete, following the region’s duel with the COVID pandemic, has meant a lot more project initiation. But are the right moves being made? To what extent are enterprises succeeding or failing? The UAE’s attention to AI at the state level is an indicator of certain technologies’ promise to deliver a creator economy of sorts, but innovators need to be wary of throwing technology at challenges and expecting it to stick.
Here, we look at a sample of best practices that can support successful AI adoption. We shall see how successful AI is “Everyday AI”, and that a culture needs to be built that can sustain the onboarding and leveraging of technologies that can be so easily misunderstood and misapplied.
The short term: high impact and ROI
For those that have high expectations and understand little of the underlying tech, AI may disappoint. So, IT leaders hoping to create an AI culture must do a mixture of managing expectations and delivering a series of quick wins that excite, or at least intrigue, line-of-business executives. IT does not have years to do this. In the few months following their declaration that the enterprise is now on an AI journey, they must win over detractors, turning distrust into acceptance and skepticism into belief.
Quick wins come from selecting a modest sample of use cases that can yield high impact but are relatively straightforward to implement. This may prove tricky if an organisation is late to the AI game, as many easy-to-implement use cases will not prove impactful if they have already become standard in the industry. It is therefore critical to quantify, where possible, the level of impact an implementation will have, to manage expectations.
If you have early success, there will be an opportunity to build on those solutions, and reuse them across corporate functions, creating business impact and winning hearts and minds. If this is done well, early adopters in the enterprise become AI ambassadors, sharing success stories, and encouraging others.
The long-term: transformation at scale
Once the culture has been established, IT leaders can look to the future, identifying other opportunities for business impact. When we talk about scaling with AI, we mean that AI projects should synchronise with business strategy. Enterprises that make it to this stage have left experimentation behind them. The enterprise has already adapted to the presence of AI technologies and AI has proven itself a reliable tool for delivering impact.
Now it is time for AI to be leveraged as a conduit for long long-term value. To do this, people and data must unite to ensure that anyone in the hierarchy can add value at any time using the Everyday AI that they have come to know and trust.
Robust AI governance will be important as an organisation reaches this level of maturity. This does not only extend to its data, which must be properly catalogued, in terms of responsible parties and what they are permitted to do with it. Security, compliance, data quality, data architecture, and metadata management are all important, but data science, machine learning, and AI present a range of use cases that may not be covered by existing data governance standards. AI governance should cover responsible AI, which is as much about the algorithms and their results (and the actions emanating from those results) as it is about the data itself.
Successful AI implementation will mean finding ways for AI governance and the agile Everyday AI culture to co-exist. And part of the culture change will be the adoption of new methodologies that inject governance and policy into the development process. MLOps, for example, is gaining ground because of its ability to reduce risk and oil the gears of implementation.
Learn and grow
Of course, none of this happens overnight. Users must learn, developers must learn, managers must learn. Everyday AI requires a lot of input from an organisation, at every level of its operations. Training must be as thoughtful as the building of the culture. Each employee must learn what they need and no more, to avoid wasting time and resources. Courses must be personalised and phased and include instruction in governance standards, culture, and hard and soft skills. Non-technical staff should emerge ready to design data projects, if not implement them.
Few successes happen by chance. They usually require astute planning and a clear vision of the goal. By breaking down the journey to Everyday AI into short- and long-term phases, stakeholders can concentrate on the things AI can do for them today rather than worry about unrealistic ambitions that have little prospect of success.
The right use cases at the right time can set an enterprise down the right path. The right narrative in the right ear can also be a boon. Applying this approach as a roadmap means not wasting time and resources trying to leapfrog realities to achieve the impractical. Patience and strategy will win out.