My father, now in his 60s, has lived in the same small town his entire life, and even continued working at the same family business since he was twelve. Despite this evidence, he actually has a very high need for variety, and enjoys being an “early adopter” for new technologies. This is the story of how a small community radio station uses AI to inform and drive sales.
When many people think of AI, robot butlers and self-driving cars come to mind. These technologies are “stand alone” solutions that do not require much human interaction but do require a huge up-front cost. However, there is another branch of AI that augments human capabilities rather than replaces it, and often for a fraction of the cost.
Many businesses wait for customers to walk in the door, but in the world of advertising, sales people are out talking with local businesses, trying to find a way to create a strategy that will move listeners to action. Any business that has a sales team knows that simply calling on more potential customers, and asking for more business can be a great driver to increase sales. However, doing this “blindly” without a coherent strategy can also be a huge waste of time, or worse yet, backfire and end in a lost account. Previously, navigating that nuance relied on experience and intuition, but now my father augments those skills with machine learning & automated accountability.
Let’s examine the sales cycle…
When managing a team, and especially when training new employees, being able to keep accurate measurements of how each person is doing, along with each account, is crucial. Being able to offer data-driven incentives on how changes to behavior can change outcomes takes it to a whole new level.
With state-of-the-art machine learning tools, an incoming email -or customer review, or any bit of text- can be quickly analyzed around distinct topics -price, calendar, marketing strategy, etc.- to extract sentiment or relevant facts from the conversation.
Here is a visual example of one aspect of parsing written communication, extracting named entities (aka: important nouns that likely have a large impact on the interpretation of the text):
This automatic extraction can then be linked to each other in meaningful ways, for example, it could suggest adding a calendar event on Tuesday morning at 10:30 AM (many email & calendar platforms can already do this). It could also be combined with sentiment scoring to better understand communication patterns of your team and their contacts.
Now we’ll compare it to a follow-up email from a different coworker.
Sentiment for this second email is noticeably different:
A manager could set a threshold for the “Overall Score” that would alert them when a conversation might have taken a bad turn, and swift intervention could address the issue. There are many more fine-tuned adjustments and automation tools that could be built in as well, especially if there is a large collection of documents to train the AI model for better future performance.
When designing an AI tool that is meant for augmenting human ability (especially if it will be imposed on people who were not part of the design process), it is very important to keep things simple and intuitive. A large draw for why my father uses this kind of tool is because his sales staff barely have to change the way they interact with their technology & reporting system at all.
Having a “Guardian of Simplicity” in the design process will increase the liklihood of adoption when the tool is rolled out, and allow focusing on the distilled-down version of the data that contains the important parts, from the manager’s perspective, while still encouraging the human connection and relationship-building that roots so many good sales interactions.
David Kaspar is a Data Analyst at Pandata.