There’s been a lot of hype around artificial intelligence (AI) over the last several years, and while some retailers and brands have experimented with it, others are wondering if AI is the right solution for them. Those unsure about whether to apply AI or not, must remember that there is often a learning curve to getting it right and thus, it is beneficial to learn from the successes and failures of others before walking down this path.Let’s first review the successes. AI has been successfully deployed by Netflix, who uses data on its subscribers to recommend movies or shows they may want to watch in the future. And by Coca-Cola, who has smart vending machines that inform the company on when a machine needs a refill. Cover Girl has also implemented an AI solution that helps consumers pick the right foundation for their skin.
Now the failures. Microsoft did not consider the amount of negative sentiment used on Twitter when it released Tay, an AI chatbot intended to be an experiment in “conversational understanding.” Within a day, the Twitter bot, turned into a racist chat bot and had to be shut down. Facebook tried a similar experiment and found their AI chatbots started speaking to each other in only a language they understood, requiring human intervention.
In a recent presentation at the 2nd annual Ugam Customer Summit, our Guest Speaker, Forrester Principal Analyst Brandon Purcell, discussed his recommendations for businesses looking to successfully apply AI. The following represent our takeaways from his presentation.
1) Start with a business objective, not a technology – Don’t do AI for the sake of AI, warns Purcell. It’s important to start with a business problem. What are you trying to solve by using AI? Is it managing your inventory better like Coca-Cola did with their AI-controlled vending machines or do you want to win more customers or improve customer loyalty, etc.? Having a clear objective going into the project will help guide the results.
2) Inventory your available data – can it be used for training? – Before taking on an AI solution, you need to understand the data you have so you can properly train the AI system. If you don’t have a lot of data to work with it’s better to go with a well-trained machine, if you have plenty of data to train the machine, then you will want to look at a system you can train. If you can, Purcell recommends training a system with your own data over going with an AI system that is already trained, as the pre-trained system will be more general in nature, and not necessarily support the unique needs of your business.
3) Select a narrow use case based on feasibility and perceived ROI – It’s always good to start with a small business problem and expand out as you go along. As Purcell notes, try to prioritize your efforts based on what you think the overall business impact will be.
4) Build if you have the right skills and data, otherwise buy – Not everyone is lucky enough to have a data scientist working for them, and if you don’t, a pre-built solution that you can buy may be better for your business. This is an important thing to consider before you buy.
5) Continuously measure impact against KPIs and customer experience – Keep an eye on your AI solution and continuously measure how its performing and make sure it’s not going off the rails and remains in line with the objectives you’ve set for the project.
6) Communicate results and build on your success! – Develop metrics for success before implementing the project and promote the project within the organization. And continue to champion the successes as you reach and surpass your goals – communicating your project and championing it across the business will only help to open more doors for how you can leverage AI for additional projects in the future.
2) Inventory your available data – can it be used for training? – Before taking on an AI solution, you need to understand the data you have so you can properly train the AI system. If you don’t have a lot of data to work with it’s better to go with a well-trained machine, if you have plenty of data to train the machine, then you will want to look at a system you can train. If you can, Purcell recommends training a system with your own data over going with an AI system that is already trained, as the pre-trained system will be more general in nature, and not necessarily support the unique needs of your business.
3) Select a narrow use case based on feasibility and perceived ROI – It’s always good to start with a small business problem and expand out as you go along. As Purcell notes, try to prioritize your efforts based on what you think the overall business impact will be.
4) Build if you have the right skills and data, otherwise buy – Not everyone is lucky enough to have a data scientist working for them, and if you don’t, a pre-built solution that you can buy may be better for your business. This is an important thing to consider before you buy.
5) Continuously measure impact against KPIs and customer experience – Keep an eye on your AI solution and continuously measure how its performing and make sure it’s not going off the rails and remains in line with the objectives you’ve set for the project.
6) Communicate results and build on your success! – Develop metrics for success before implementing the project and promote the project within the organization. And continue to champion the successes as you reach and surpass your goals – communicating your project and championing it across the business will only help to open more doors for how you can leverage AI for additional projects in the future.
AI is definitely a hot commodity – and it’s worth trying – but ONLY if you have the right plan in place for what you hope to achieve.
For more insights on AI from Brandon Purcell, view his presentation from the 2018 Ugam Customer Summit Click here.
For more insights on AI from Brandon Purcell, view his presentation from the 2018 Ugam Customer Summit Click here.
