Designing AI initiatives that enable your organization to deliver value through high stakes use cases is no mean feat. However, there are sensible steps that any skunkworks project can incorporate to increase their chances of success.
Although discrete use cases can provide incremental efficiencies, they are too constrained to drive material change in both the operation and bottom line of an organization. Moreover, small scale projects are inherently difficult to scale, especially when they were not designed with the wider company in mind. At the same time, enterprise-wide efforts face risks from too many moving parts, shortfalls in support from leadership, unsuitable education programs, and too many cooks in the kitchen.
Navigating these waters is challenging, but success can be achieved through communication and a sensible process that aligns all the moving parts. As a starting point, organizations should look for areas that will solve systemic business problems like process inefficiencies and bottlenecks in product delivery. Often these steps will not require the immediate implementation of advanced technical tools, however they provide a no-code value that will both increase the odds of project success and improve the general function of the business along the way.
Finding oxygen throughout the process means having support from both leadership and a dedicated team. I use the word dedicated here to refer to the business owners that will fulfill certain key roles in the brainstorming and development of your AI initiative. This may include a product owner or department head that could very likely be the individual responsible for pursuing a solution using advanced technologies. There is also the need for team members focused on the change management side of the operation, including any requirements for education and certification. Finally, there are the frontline ‘users’ of the solution. Do they currently exist? Or are these roles that need to be created? If so, will it make more sense to reskill internally or look outside the business for new talent?
Finally, have you selected a domain where the data and technology components necessary to run the AI models overlap, so each new AI project within the domain can build off past work, rather than start from scratch every time? Put another way, is this project going to enable continuous improvement or create a new inefficiency that someone will have to clean up later? Organizations are successful when they focus on a few priority domains to start their initiatives based on value, feasibility, and leadership support. Ensure you are capable at one thing before you take on another. This is the meathead equivalent of not overloading the bench-press on your first day back from vacation.
If you are looking for a basic structure to help frame your next automation project, a three-step process can help:
- Set expectations early and look for process wins that maximize value toward understanding data whilst minimizing exposure toward AI over-application.
- Ensure your development team (internal or external) have a sensible agile based approach which incorporates risk assessment alongside milestones. This can be achieved through:
- Pressure testing use cases
- Scaling a prototype through multiple smaller scale proof of concepts
- Assess the needs of IT infrastructure, and kick-start transformation. Ensure you build the capabilities required for continued success. Short-changing your infrastructure needs is never a great idea.
After an organization witnesses success within the first few use cases, it will have a repeatable playbook of methodologies and a mindset for reusability that enables the acceleration of AI innovation. Eventually, this upskilling and process improvement will enable organizations to take on multiple use cases at a time without fear of overloading. Ultimately as firms move from one AI pursuit to another, their pace will quicken, their AI capabilities will rapidly compound, and they will find that their goals are more attainable than what initially seemed possible.