- Data Science: the business pictureJanuary 23, 2019
- Advanced analytics and artificial intelligence (AI) allow discovering new sources of growthJanuary 7, 2019
- Data Science: the Science is in the DetailNovember 15, 2018
Creating AI Solutions with the AI Model in Mind First – Not Users
- May 10, 2018
- By Franki Chamaki
Don’t start your AI project off on the wrong foot.
When designing AI solution, it’s easy to immediately get caught up in all the ways it can service your customers’ needs. Media hype about digital assistants like Alexa and Siri focus on user-facing features. So, it’s hardly any wonder why we think in such fashion, yet it’s the wrong approach to design a functional AI platform.
Henry Ford is famously credited for saying that if he asked people what they wanted, they would say faster horses. Although we could never be certain if he actually uttered those words, the fact is, customers don’t always know what they want, especially if the product you are considering is never-heard-of. While addressing customers’ needs is certainly important, it’s secondary to building functional tools that will actually help them. Key Performer Indicators (KPIs) are important in business, and you need to set quantifiable metrics and goals before unleashing a buggy program on your customers.
Just like you wouldn’t throw a new hire on the floor without providing them with orientation and training, AI needs to be trained on business rules, policies, and procedures. A well-honed AI will inevitably serve your customer’s needs but things will continue to change throughout the project cycle.
This is why user design only comes after a solid foundation is set in place.
Onboarding AI in any facet of business operations
With design thinking in place, algorithm development of AI applications is tested and validated before its customer interactions are refined. We learned this lesson when implementing HIVERY’s Vending Analytics and Promotional Effectiveness Calendar tool and our Category Management application. All AI applications are done in an iterative fashion.
As we focused on refining the machine learning algorithm first and foremost, after the deployment of Vending Analytics solution we saw a 15 percent increase in sales and an 18 percent reduction in re-stocking visits. Once the model was sufficiently in place, we designed a user interface to leverage the model’s thinking.
This design thinking method is important. A checklist should be set in place to replicate the process for any new technology implementation. Google increased its employee productivity by 25 percent with a simple five-step onboarding checklist similar to the one below, so there’s no reason to skip it.
1. Define your business KPIs
What is the purpose of your business? Probably, to generate profit while providing quality goods and services. That part is easy, but setting goals that quantify the profit is harder.
AI applications are intelligent, but they can only do what you train them to do. Defined KPIs regarding production and quality as well as business policies and procedures are, therefore, necessary.
2. Determine data sets that quantify goals
After you develop a clear definition of your goals, you must determine what data sets should be analyzed to evaluate your progress towards these goals. Computer software requires business rules. While your AI doesn’t need to know a dress code or how to clock in (unless it’s being used as an employee onboarding platform), it definitely needs to understand the data that has to be looked at.
3. Develop a capable algorithm
With goals and datasets in place, algorithm development is much more efficient. Machine learning builds over time, but only when a solid foundation is built for it to grow on.
Your employee of the month didn’t get there by fumbling their way through the business. Company orientation programs, departmental trainings, managerial one-on-ones, and more were crammed into busy schedules, all culminating to that one milestone. Develop AI applications that can intuitively navigate large data sets within company policies.
4. Design the user interface
Finally, you have a viable AI and it’s time to determine the user-facing side. Will it operate as a chatbot that answers customer and employee questions on social media? Or as a data verification app in legal? A handheld scanner on front-end employees, or even a robot in the warehouse? This is the icing on the cake as your employees graduate from the company training program and prepare to hit the live floor.
AI is a hot technology that is disrupting every industry. Over half of the largest businesses in the world have an AI strategy, and that number is only going to continue growing. Get on board now with an intelligent user design based on a solid backend.
Study: Google Assistant most accurate, Alexa most improved virtual assistant
Henry Ford, Innovation, and That “Faster Horse” Quote
What is Design Thinking and Why Is It So Popular?
Google Increased New Employee Productivity by 25 Percent With 1 Email. Here’s What You Need to Know
Reshaping Business with Artificial Intelligence