Practical Steps to Developing an AI Strategy
- August 7, 2020
What are we in the business of doing? This is the first question organizations should answer when building an AI strategy. Anyone so charged needs clarity on why their company exists in the first place. The company’s mission statement will provide the answer to ‘Why’ the company exists. Once the ‘Why‘ is understood, assess the ‘How‘: how does the company execute on its purpose?
For example, at Dell Technologies, our purpose (Why) is to drive human progress, through (How) access to technology, for people with big ideas around the world.
With the organization’s mission statement in mind, instances where the company’s operations and outcomes fall short of the mission should be identified. A senior executive leader should have broader insight on where those instances exist and could recruit business leads to scope out pain points as well. For example, a food processing company with a mission to provide superior products and services to customers, might fall short of this mission if the company has high reported numbers of degraded produce or excess waste from product defects. This may point to underlying problems in the manufacturing process. If defects could be predicted with the foresight to avoid excess waste, this would take the old saying that ‘Hindsight is 20/20’ and make it ‘Foresight is 20/20.’
All AI use cases are built on defined tasks that identify where AI can be used to improve company operations. In the case of the food processing company, the task could be to ‘predict, detect, identify or recognize,’ and the use case would be to predict or identify defected product early so that the company lives up to its mission. Some examples of AI tasks and the questions they answer are:
Using this ‘task plus use case’ formula creates a running list of use cases from which a company can downselect. If several use cases can be resolved using the same kind of task, it allows for rinse and repeat opportunities and quick AI adoption wins. To downselect, the following questions should be considered:
Waste from product defects will cost any company over time. Records of these costs serve as evidence of business value and strategic impact if money can be saved and reallocated to innovative projects.
Feasibility can be measured using the availability of relevant data to understand the problem and train AI models. What data is available, where it’s stored and how it needs to be prepared for use in AI are important factors. Relevant data provides information about factors that contribute to the desired outcome. For product defects that could be indicators of faulty processing equipment.
As mentioned, looking for external answers too early could take away the uniqueness of internal ideation from the people who know the business the best. Identifying use cases gives reference point while conducting research on what ‘similar others’ are doing with AI. External research informs on the level of effort needed to execute on selected use cases, as some might be successfully deployed already, and some might not. Web searches such as ‘AI + *insert industry*’ or ‘predict product defect + manufacturing + AI’, will yield pertinent results. Other AI tasks might come up and be applicable to other use cases.
Understanding the direction of AI adoption sets the stage for building a team of active participants to garner company-wide consensus for adoption. The following individuals would be candidates for this team:
Dell finds that successful AI projects follow a pattern like Maslow’s hierarchy of needs to reaching one’s potential. This hierarchy starts with a use case and works up to an optimized IT environment to support it.
Choosing the right solution can be daunting, so we’ve created an ecosystem of resources to help customers on their journey. Find more information here, and if you’re ready to get started, contact our Analytics/AI, HPC and Cloud team. We look forward to voyaging with you.