Everyone’s talking about machine learning (ML) and Artificial Intelligence (AI) these days. If you are a CxO or work in IT or marketing, I’d bet that you hear these terms more than you probably want to. It feels an awful lot like the early data of Big Data or Business Intelligence or the days when the “Intranet” was first making waves within organizations.
Like most new technologies (ahem…buzzwords), machine learning and AI can seem like solutions looking for problems. While I would argue there are people / companies looking for problems to throw their experience with AI and machine learning at, there are some viable problems out there for ML/AI. That said, I still stand behind my argument that you probably don’t need machine learning…but every organization should investigate the use of ML/AI.
Rather than buy a solution and then look for a problem to through it at (like many vendors / consultants are pushing these days), its worthwhile for every company to spend some time looking at a few important areas within their businesses to see if there’s anything that ML/AI can do to help.
Below are a few examples I’ve helped organizations with over the last few years.
Areas to start investigating the use of Machine Learning
Improving/Personalizing Customer Service
Customer service is one of those areas that you either immediately think “yes…that’s a perfect place to use ML/AI” or “uh…what?”. Hopefully you fall into the former category because customer service is an ideal space for implementing machine learning and artificial intelligence to help improve service, better understand your customers and personalize interactions. Why’s it an ideal space? Because you have a lot of data – some of which is structured and some of which is unstructured. What better place to start with machine learning than a place that you have a long history of data and have multiple types of data? It’s a perfect problem for a machine learning solution.
Additionally, the use of AI for things like chatbots can drive a great deal of value for your organization. In a reported described by Business Insider, 44% of consumers surveyed stated that they would use chatbots if the experience could be perfected/improved. That’s an impressive number given that these chatbots are automated and people claim to want to speak with ‘real’ humans when contacting an organization.
Fraud Detection and Analysis
You don’t have to be a large credit card company to benefit from machine learning for fraud detection. While those organizations do benefit greatly from implementing ML / AI systems and approaches, any organization that has large enough volumes of transactions can use various machine learning approaches to detect fraudulent activities. How much is ‘large enough’? I can’t tell you that…but if you have transactional data covering multiple years, you should have plenty of data to build an anomaly detection algorithm to see those transactions that are out of the ordinary. Fraudulent activity detection isn’t something every organization can benefit from, but it is a large area the lends itself well to machine learning approaches.
Supply Chain Management
Another area ripe for machine learning is the supply chain. If you sell products and manage logistics, you have a great deal of data just waiting to have machine learning turned loose on it. You can find new efficiencies in your supply chain, find areas that can be improved upon and find new avenues for cost cutting as well as revenue. The supply chain has a great deal of both structured and unstructured data as well as many different types of data that cover many different types of metadata (e.g., costs, times, production requirements, etc). The large amount of data as well as the various types of data provide an ideal base of data to apply ML techniques to better understand and manage the supply chain.
Measuring marketing ‘reach’ / brand exposure / campaign success
One of the first things that many organizations want to do with machine learning is to throw their marketing data at it to ‘do things better’. While I find this fairly naive, I also love the enthusiasm. Marketing and the data that marketing groups have is an ideal place for organizations to start investigating the use of ML/AI as there is generally plenty of data of varying types throughout every marketing organization. Using ML, organizations can get a better feel for who their customers are, how to reach them quicker and more effectively and how well their campaigns have performed.
Creating a better hiring process
When I was first approached by a client and asked if I could help them ‘improve their hiring process’ using machine learning, I was skeptical. I’ve always been skeptical of most hiring processes and have rarely seen an automated hiring process within HR that I would consider to be ‘good’. I shared my concerns with them – and they agreed completely with me – so I agreed to help them build a proof of concept system that used machine learning and natural language processing to sift through resumes to fitler the ‘best’ ones to the top. Our first attempts were no better than their existing keyword search systems but we quickly found an approach using keywords combined with other ‘flags’ that could find those types of people that this organization liked to hire and filter them to the top of the queue.
Using Machine Learning / AI during the hiring process is still a tricky concept because a human with domain experience will generally find the best candidates for a position, but ML can help filter the candidate pool.
What are you doing with machine learning?
There’s a lot of buzz about machine learning and AI these days. Most of that buzz is because of the real value that can be found with properly implemented machine learning/AI using quality data.
What cool things are you implementing with machine learning?