AI business applications are here to stay. Businesses are using AI to improve operations, increase value to customers and reduce cost. But it can be hard to get started implementing AI when opportunities aren’t clear. This post presents a few real-world examples to help spark ideas that you can try in your organization.
Prioritizing prospects
Digital marketing tools like Mailchimp or ConstantContact provide users with various reports about their campaigns. These reports include many key performance indicators and qualitative data, like who opened an email, who clicked on a link, etc. These tools also calculate and assign a kind of “quality score” to each contact. Customer Relationship Management (CRM) systems do this too; they use the data housed in the system to score leads and surface the most qualified prospects. Sales and marketing teams could take this one step further. They could combine marketing, sales and inventory information and train a machine learning model that shows who might buy what and when. Marketers can use this information to tailor their marketing messaging.
Automating the verification of information
Many organizations rely on humans to review and categorize information. Take for example a market research firm that automatically collects web content related to their area of research. A team of human verifiers then combs through the data and flags the useful information. This job becomes increasingly unsustainable as the team gets buried under a growing pile of records. Training a classification model using historical data could help prioritize the records that are more likely to be useful, so that the team of verifiers can focus on those first.
Digitizing paper-based information
A lot of important information is trapped on paper and many scanned documents aren’t searchable. Organizations can digitize this information using commercial products like Amazon Textract or comparable alternatives like the Textract Python package. As the Textract site describes, “As undesirable as it might be, more often than not there is extremely useful information embedded in Word documents, PowerPoint presentations, PDFs, etc—so-called “dark data”—that would be valuable for further textual analysis and visualization. While several packages exist for extracting content from each of these formats on their own, this package provides a single interface for extracting content from any type of file, without any irrelevant markup.”
Categorizing sentiment and opinions
People who analyze surveys spend a lot of time categorizing answers to open-ended questions. Not only a time-consuming issue, humans are prone to bias when interpreting ambiguous information. This becomes even more problematic as multiple people—each with their own biases—divvy up the work of categorizing a single set of responses. Topic classification algorithms can cluster responses into logical categories. Sentiment analysis tools can determine whether a response is positive, negative or neutral. Other natural language understanding methods can flag threatening, abusive or inappropriate comments. These and other text-analysis tools can dramatically reduce the time spent on these tasks and increase the quality of the analysis.
Automating first line of support
Chatbots are everywhere. These virtual agents use Natural Language Understanding (NLU) and machine learning models to match a user’s question to an existing answer. While the technology isn’t perfect—chatbots still feel a bit clunky—they will become very good sooner than later. Eventually, these tools will provide a good-enough experience and answer most questions or route the more nuanced ones to the appropriate human agents.
Improving the user experience of marketplace applications
To a large extent, the defensibility of a marketplace-style application depends on its ability to create and sustain network effects. Further, what makes a marketplace sticky and valuable is the frequency with which both sides of the market come together in a transaction. Following common business models, the higher the number of transactions, the more revenue the marketplace generates. Machine learning can help entice the demand side to transact. A machine learning model could present specific supply goods (e.g. Airbnb apartments) to users who are likely to be planning a visit to the destination and that are likely to value the amenities of that property. Also, the site’s search engine could rank results based on the likelihood that the user performing the search would find the options appealing.
The path forward
We hope that this post helped you picture how your organization could start using AI to improve processes and make products and services more valuable to customers. You might be wondering what to do next. Reach out to us for a consultation. We are here to give you our advice and share ideas with you.
