What to consider before investing in AI
AI is one of the biggest buzzwords of the decade. Most companies think they should be integrating AI into their daily business practice, but many don't understand what AI is or what it can do. While AI can be a game-changer for companies, it can also be a wasted pit of money if you don't know what you are doing. The number one rule of money pits, they are only fun if you are diving into them, not if your company is propagating them.
When companies correctly implement AI, they see significant benefits. According to the study, "The Enterprise Powered by AI," conducted by the Capgemini Research Institute, 97% of respondents reported 'measurable benefits' from incorporating AI into their business. Even with 2020 being a challenging year, one in five respondents said they plan to increase their AI investment, and more than three out of four said they plan to continue to invest as they did before Covid.
While there are many benefits of AI, that doesn't mean that it's always easy. If you aren't in the tech or AI industry, it can be challenging to parse fiction from reality when determining what is possible. Most people envision AI as something between incompetent chatbots and futuristic super-computer murder machines.
In reality, AI is neither or both, depending on whom you ask. We are at a precipice where technologies are rapidly changing, but there are still some kinks to work out. While tech companies are doing some amazing things with AI, most companies trying to add automation or machine learning are also not themselves tech companies. Their priority is still their actual industry, and tech comes second, often leaving less room and resources for these projects.
So what should you really consider before proposing AI?
Finding your use-case
Every company has things they excel at and others that are more difficult to get through. In order to help identify which projects could be the most valuable to your company, ask-
What does my team feel is the most draining or unfulfilling part of their job?
What takes up the most time during the day?
Where do we continuously face bottlenecks?
Where in the pipeline do we tend to lose clients?
The key for non-tech companies looking to implement AI is to start small and specific. Projects that pinpoint a specific task are far more likely to succeed and show tangible results than something broad or wishy-washy. Make sure you assess every possibility's value chain and determine exactly how it will impact customer expectations and experiences. If you can't clearly show the benefits of your project to your team and higher-ups, there is little chance you will be able to get anyone on your side for future proposals. Some examples of common use cases include-
Chatbots and prioritization tools
An inordinate amount of customer service representatives’ time is spent answering repetitive customer questions, backing up the line for clients with more important, or at least more complex, questions. By utilizing chatbots or tagging systems to prioritize specific inquiries or clients, CSRs can instead use their time where it can create the most value. Chatbots and prioritization filters are also relatively simple to implement because tech giants like Google, Facebook, and Zoolando have already done the hard work of teaching computers to understand human language. Companies can now purchase a natural language processing (NLP) system and use it right out of the box for most projects.
Other AI tools in customer service include those used to measure important KPIs such as CSAT, resolution times, and ROI. AI can also improve customer relationships by tracking incident history, lifetime customer value, and churn probability.
Customer journey mapping and product recommendations
Readily available tools such as Google Analytics or Adobe Analytics can show how customers move through your website and identify the Struggle Score (number representing how difficult users find that page to navigate) of each website page. Once you can keep customers from closing that browser tab, product recommendation tools such as Adoric, Monetate, or Barilliance can show them specially selected products that will hopefully catch their eye. Amazon claims that 35% of their sales come from product recommendations, a huge total number of sales to be leaving on the table.
Market Research
AI systems such as IBM's Watson can conduct market research, compare competitors and create detailed reports, all in a fraction of the time (and potentially with a fraction of the errors) that a human could. Other useful tools for market research include advanced data analysis, automated stats analysis, NLP, and prediction tools. Combined, these tools can not only tell us how everyone in the current marketplace is performing but also give insight as to how future products will do before they even launch.
Find the right tool
AI is far from one thing. Instead, it's a whole tool kit with all different kinds of precision utensils. Sure, you could jam the screwdriver in there and kind of make it work, but if you put in the effort and search for the right tool, the result will be easier to use, more stable, and function more smoothly. One good thing about finding the right tool is that there are often many choices for any particular issue. Which you choose to use will depend on your exact needs, cost, and simply which tools your data scientist prefers.
Assess your data landscape
Chances are your company has some data, but the amount, type, and quality will all be factors in what kinds of projects are possible, or at least relatively easy to accomplish. If you don't readily have the right data available, that doesn't mean your project is doomed to fail, but rather that extra steps or tools need to be incorporated.
Measure expectations
A common reason AI projects don't live up to their expectations is that those expectations were unrealistic to start with. While you might have an educated understanding of the goals and processes, others on your team might not. Clear communication is key in ensuring that every person and department involved knows exactly what the project entails and the predicted outcome. If the boss expects a fully customized, high-level system for only $3,000, there is a good chance everyone will be disappointed in the end.
It’s difficult in the end to know which project to undertake, or when you are ready to launch. Finding the right project, which tools will help you reach your goals and how to proceed, will at least ensure that the project gets started on the right path.
To read this article in its original publishing, click here