Those who do not know if the AI should be used should know that, as in any situation in life, one can always look at the successes and failures of others to learn from them and follow this path independently.
Some successes:
- US entertainment company Netflix has successfully deployed AI, which, based on subscriber preferences, recommends movies and TV shows they may want to watch in the future. Something similar is used on YouTube. Obviously, for such companies, the use of artificial intelligence is a great way to retain a new audience, which also means increasing your own profits
- Coca-Cola has smart vending machine that can signal that they need to be replenished.
- The American cosmetics brand CoverGirl has also introduced AI, which helps consumers choose the most appropriate foundation for their skin.
Some failures:
- "Microsoft" did not take into account the number of negative moods on Twitter when publishing Tay – a conversational agent designed to conduct "conversational comprehension" experiments. In just one day, the robot became a racist dictator and was immediately turn off.
- "Facebook" also tried to conduct a similar experiment, but it gave much stranger results: their their chat bots started to communicate only with each other and in a language incomprehensible to humans.
At the second annual Ugam Customer Summit in 2018, Brandon Purcell, senior analyst at Forrester, presented his own recommendations to companies seeking to a successful use of AI in their businesses.
You will find below the conclusions of his presentation:
1. Start with a goal, not a technology.
Do not engage in AI for AI. It is important to start with the problem.
What are you trying to solve with the AI? Do you want to manage your inventory like Coca-Cola with your vending machines? Or do you want to attract more customers? Or increase their loyalty? Etc.
Having a clear goal that is part of the project will help direct you to desired results.
2. Make an inventory of your data - Can they be used to train the AI?
Before you begin to solve the problem, you need to understand what data you have in order to properly prepare the AI.
If you do not have a lot of data to work on, it's better to use an already trained AI.
If you have enough data to configure the AI, then you should consider the self-learning option and decide on the type of AI used.
Purcell recommends teaching AI with your own data, rather than using an already pre-trained AI, as such an AI will be more general in nature and will probably not take into account the unique characteristics of your business.
3. Choose a limited use case based on feasibility and estimated return on investment
It's always good to start with a small business problem and expand as you go. Purcell emphasizes that it is helpful to set priorities in your efforts, depending on the overall impact on your business.
4. Create, if you have the necessary skills and data, otherwise - buy.
Not everyone is lucky enough to have a big data specialist and machine learning. If it is your case, a ready-made solution may be better for your business, however it is very important to understand what you are buying before buying it. At AZN, we can offer you a live demonstration of our solutions.
5. Constantly measure impact on KPIs (KPIs) and quality of customer service
Monitor your AI and constantly measure its effectiveness. Make sure that it does not go astray and that it respects the objectives that you have set for the project.
6. Report your results and consolidate your success.
Develop indicators of project success before putting the AI into service and, even better, before it is implemented. Follow the evolution of these indicators as you reach and exceed your goals. Report on your accomplishments and encourage your decision. You will have even more possibilities to use AI, for example for other projects in the future.
«Artificial intelligence is definitely a useful technology, and it is worth trying, but only if you have a clear plan in line with your goals»