The solution is to build a team of unicorns by upskilling from both sides of the fence, business skills for the AI talent and AI skills for the business talent, says Sid Bhatia at Dataiku.
There is no question that, lately, AI has become a lot more popular across the GCC. One recent survey found almost two in three, 65% UAE organisations are already using AI for automation and to enhance the employee experience. And 61% of Saudi Arabia’s enterprises have adopted it to bolster their cybersecurity.
And yet, we still see high failure rates among AI adopters, suggesting a need for careful consideration of all the angles before embarking on ambitious programs. What we see are a variety of points of failure and that they are distributed across business and technical teams. However, we can categorise the points of failure in other ways.
Here’s how organisations can address each category and create an environment of Everyday AI, where each employee, from the top on down, can effortlessly think in AI terms and create value autonomously.
Role problem
Most regional business leaders and all technologists are aware of the AI talent gap. But the problem does not only lie in onboarding rare analytics specialists and data scientists. Even where the talent can be recruited, it is prohibitively expensive for most organisations to hire more than a few, which will limit team size.
But additionally, it may be some time before a newly hired AI professional gets to know the business well enough to identify potential use cases. It is impractical for a regional business, looking to thrive in a competitive market, to wait around for the unicorn hire that will understand both AI and business.
The solution to this quandary is to build a team of unicorns by approaching upskilling from both sides of the fence, business skills for the AI talent and AI skills for the business talent. This approach is the most direct path to AI maturity. While the training for AI professionals will be obvious, the business side can learn AI by doing, through low-code development.
These citizen developers will be a key ingredient in propelling the AI adoption programme forward. Even before their AI training they will be brimming with ideas about how to address the business’s pain points. As they are exposed to advanced, easy-to-use AI platforms, they will quickly spot opportunities for quick-win solutions, implement them, and, under appropriate governance, add value and increase confidence in AI.
Process problem
Those that approve AI projects want something to show for it. At the beginning of the AI journey, quick wins are how we demonstrate early value and build a momentum of acceptance. What we are trying to achieve in delivering Everyday AI is a mindset that turns to AI by default when trying to overcome a challenge. Project teams must deliver efficiency boons, revenue spikes, cost reductions, and more. They cannot afford to waste time building and testing machine-learning models that amount to nothing.
AI teams must work within a methodology that slickly brings production-ready, value-adding solutions to life. There are best practices already out there that capture all of these scaling and streamlining needs. AI governance strategies that cover operational and value-based concepts such as MLOps and responsible AI are indispensable and should be established before embarking on any development.
AI governance will deliver end-to-end model management at scale, accounting for risk, compliance, value-add, and efficiency, as well as whether a solution is a proof-of-concept, a self-service initiative, or a live product.
This means an Everyday AI organisation can be comfortable with exploring various technologies because policies are in place to ensure no concrete steps are taken towards implementation without due diligence and fiscal commitments.
Right now, compliance is becoming a trickier prospect, by the day, across the region, as governments look to erect guardrails on behalf of private individuals. Strong governance is the engine of compliance, allowing for privacy, accountability, and transparency.
Platform problem
If you thought AI was a standalone technology, you would be misunderstanding the adoption journey. Think of AI as the engine for a motorbike. It needs other things to be useful. We might think of the petrol as data; the wheels are the hardware or cloud services that consume the fuel, powered by the engine. We add governance as an all-in-one clutch, throttle, and break. And behold, we have a useful platform of capabilities that we can take on the road.
A good AI platform will provide all these facilities and align with the steps of analytics and AI project lifecycles, from data cleaning to model drift. It will cover data access, cataloguing, and exploration. It will cover automatic and coded machine learning, and the explainability of models.
Operational aspects such as model management, orchestration, and exposition will also be in attendance. And the platform will take into consideration many outside factors such as performance, compute requirements, governance, project management, and collaboration.
The IT department is therefore a critical component of Everyday AI. It will ensure smooth rollouts of platforms and will oversee the governance side of AI, operating under a culture of open access balanced with due control.
Consider carefully where you want to be. AI is a business-change programme. If you are looking to scale up the capacity to capture and process data to gain actionable insights, you have come to the right place. But many twists and traps lie ahead. Failure is a painful prospect for teams that have spent arduous hours, days, weeks, even months, on a project that was supposed to deliver value but did the opposite.
While the old Silicon Valley mantra of fail fast is not always an acceptable proposition, if the IT environment allows for a staging area, it is highly recommended. Apart from that, remember to upskill, train, invest in tech, democratise the use of tools and technologies, and govern with wisdom. You have arrived. This is Everyday AI.
Key takeaways
- We still see high failure rates among AI adopters, suggesting a need for careful consideration before embarking on programs.
- We see are a variety of points of failure and that they are distributed across business and technical teams.
- Citizen developers will be a key ingredient in propelling the AI adoption programme forward.
- AI teams must work within a methodology that brings production-ready, value-adding solutions to life.
- AI governance will deliver end-to-end model management at scale, accounting for risk, compliance, value-add, and efficiency, as well as whether a solution is a proof-of-concept, a self-service initiative, or a live product.
- Everyday AI organisation can be comfortable with exploring technologies because policies are in place to ensure no steps are taken without due diligence.
- Strong governance is the engine of compliance, allowing for privacy, accountability, transparency.
- A good AI platform will provide facilities and align with the steps of analytics and AI project lifecycles, from data cleaning to model drift.
- A good AI platform will cover data access, cataloguing, exploration and cover automatic and coded machine learning.
- Consider carefully where you want to be, AI is a business-change programme.
- Failure is a painful prospect for teams that have spent hours, days, weeks, months, on a project supposed to deliver value but did the opposite.
