Unless you have been hiding under a rock for the past decade or two, you have probably heard about Artificial Intelligence (AI) and Machine Learning (ML). It has invaded our lives, taken residence in our homes and garages, and – thanks to the advent of the smartphone, become a welcome guest to nearly every moment of our daily routine. AI has become the cornerstone of big businesses like Amazon, Google, Apple and IBM, reforming what is possible for retail, logistics and manufacturing, but did you ever stop to think how AI could be used for a mom-and-pop business or small company? You may not have, but that hasn’t stopped AI from thinking about your business!
In this edition of our Plain Talk newsletter, we’ll dig in to how even small- and mid-size businesses can benefit from AI and machine learning and explain how to take advantage of the AI that is already embedded in your business whether you like it or not. Regardless of whether you are engaging in an active project to implement the benefits of machine learning at your company or have a small website and an “open” sign that you pull the chain to turn on in the morning and pull again to turn it off at night, understanding how to take advantage of AI to increase your business’ profits is a crucial part of operations and business marketing.
Active AI Implementations
To simplify things, we are going to break down ways you can leverage AI into two categories, active implementations and passive ones. What we are labeling as “active” implementations are the types of projects that require a significant level of planning, capital expenditure, and execution. It is best to have a well-formed committee and/or top-down support to manage these types of projects, which will require significant up-front planning and project execution to ensure success.
The current state of AI is extremely well suited for anomaly detection. Whether you are training an AI model to detect anomalies in metrics, contextual data, or even with visual comparison, you can now implement a solution with a service like Amazon Lookout in hours over days (or months… or years) . Again, it is very important to do your technical/business analysis upfront to take advantage of the ease of use in these services.
Imagine your company produces a product that is primarily wood grain in appearance. It can be quite a challenge to detect defects in the product because of the variance in wood grains. With Amazon Lookout for Vision, you can connect an appliance to cameras and upload images of both good products and defective ones to train a machine learning algorithm in just a few hours that can detect defects with excellent accuracy. The real work for us is in defining what different defects look like and gathering the photographic evidence and known details over having to program the machine learning algorithms to analyze them.
Another great use for anomaly detection is assembly line quality control. You might say, “That sure sounds like big business stuff,” and you would be right in one respect, but what if you are a startup consumer electronics company, such as a boutique guitar pedal shop, who is assembling your pedals’ circuit boards manually. A circuit board with many components may be hard for a human to detect defects or missing components on, but to AI, it is a piece of cake! You might even go a step further and feed quality assurance data from the assembled product into Amazon Lookout for Metrics to guarantee that each product that gets shipped is 100% perfect.
Branding & Customer Contact Support:
That’s right, I said branding. One of the most exciting and human-like things an AI can do now is talk! Siri, Google, and Alexa have all become brand representatives for their respective companies, and your company can take advantage of that too. For businesses that cater to specific audiences who want to be greeted in a familiar voice, accent, or even dialect, this is a game changer. There is perhaps nothing more polarizing for a brand’s consumer perception than the experience their customers have with the contact center.
Imagine you have a regional bank that caters mostly to farm owners in Kentucky. How much more willing to speak to an AI for common questions and call routing do you think your clientele will be if the voice of the AI sounds like it knows you? Maybe it’s trained to make small talk about the weather and throw in some tips for farming? (In a mild Kentucky drawl). “Hey there Eli! Thanks for callin’ buddy. Bet those cows of yours are happy this weather is warming up! Now let’s get you routed to the right support person.”
The possibilities are exciting, plus the AI can transcribe the call into text and refer to it in order to present your call center personnel with helpful contextual information while they are on the call. Maybe that will end the need for having to give your information over and over again every time you are routed to a new person on the phone!
Most of us are familiar with the website chat bots, which can be helpful for answering commonly answered question in a tree format. The new generation of AI-powered chat bots do not have to follow a set of predetermined responses and can be trained to have a personality akin to their speaking counterparts mentioned above. In fact, if you think of a natural language assistant or a chat bot as a “person,” then they are technically the same person, sharing the same brain!
Keeping your customers “engaged” on your company’s website is one of the most challenging problems faced by digital marketing. In the recent past, many companies’ websites have employed A/B testing to offer up one or more choices for a customer to potentially become engaged with. Depending on popular tastes for trigger words or graphical representations of articles or products, an A/B test can provide marketers with metrics on which variants of the test are more popular with users. After the test is complete, marketers can adjust what is displayed on the site “permanently” to use the winning variant. While this type of testing is very helpful, user sentiments change over time and tests continually need to be rethought and re-run.
As a marketing agency, we often conduct consumer focus groups to get a better idea about how to position a particular campaign based upon information we derive from speaking to actual consumers about the brand, product, or tactics we will use in a prospective marketing campaign. The information derived is invaluable to creating a successful campaign, but in the end, it is limited by the size of the group polled and the passage of time, which tends to reveal changes in consumer behaviors.
AI and machine learning can be used in a method called reinforcement learning to test consumers continually via advanced multivariate testing and polling in a combined process. In what is known as a Multi-Armed Bandit solution, the AI tests visitor engagement through lots of different factors and then adjusts on the fly as it develops a high-confidence prediction about what content should work and what might not work to capture a user’s engagement. The AI can use different algorithms to continuously balance exploration and exploitation as well as optimistic “faith” that certain content will be successful based on what is derived from other inputs like polling and training by the business marketers.
Passive AI Implementations
What we are calling “passive” implementations of AI are some examples of how your business can augment the benefits of Artificial Intelligence already serving your marketing needs in a capacity that’s not under your direct control. Every size business can take advantage of billions of dollars in research and development by simply recognizing AI is out there and keeping that fact in mind while performing your normal marketing and business development activities. Here are a couple of important ways to look at this.
Tune Your Site for Voice Search:
In the same way that your company would tune your site for better Search Engine Optimization (SEO), you can adjust your website content for voice search on mobile devices and home assistants like Google and Alexa. Sites like Epicurious.com and other recipe sites have been doing this well for quite a while. Just try asking Google or Alexa how long to cook a lobster tail on the grill or what temperature to cook a steak to medium rare. You will not only get the answer, but the site will get a voice branding boost when the assistant tells you where the information came from.
You may not know it, but Search Engine Results Pages (SERP) for a mobile search and a desktop search may vary, with mobile voice searches being around 40% and home assistant searches being basically 100%. A poll by PricewaterhouseCoopers in 2018 showed 71% of consumers prefer voice search over typing. Microsoft estimated that by 2020, three quarters of American homes would have smart speakers in the house, and although that number has not been seen yet, it continues to rise every year, especially after the holiday season when the popular devices are given as gifts. On top of those amazing figures is the growing number of devices and applications that have electronic voice search assistants built into them, such as cars, phones, TV’s and other devices.
Today’s AI in voice search is used to correlate natural language terms into semantic search algorithms instead of just processing keywords. Since the “Hummingbird” update by Google in 2013, the search engine’s AI can process not only the words you search for but the intent of those words, providing more accurate results for all search requests, but most importantly for voice searches. Today you can initiate a voice search for something, someplace or someone specific and then continue to ask more questions related to the topic of that search.
Much of the way we tune sites for voice search is like the methods we implement to create better SEO in general. Some of these are:
- Making sure information is included in plain text or HTML instead of being embedded in images.
- Creating textual content that is informative, keyword rich, and has structure that is consumable by search engines.
- Following common, human-readable formats for URLs that include categorization and canonical references.
- Including long-tail search keywords within headlines and content. In other words, a group of words that are likely to be searched for together in natural speaking terms.
- Using strictly formatted Schema Metadata, which is essentially a code structure that describes a block of content in a way that the search engine uses to interpret specific content types as an object.
In addition to these (and many more) methods to increase SEO, voice searches are greatly enhanced by having an up-to-date Google business listing which has all the pertinent information about your business, such as name, location, phone, website, hours of operation and much more. Try a search for “local pet stores that are open near me.” The intent of this search activates a business category, geolocation, structured details and much more when the assistant returns its response.
Voice search trial and error is another method that can be used to discover what content you can include on your site to trigger a relevant response. Since many smaller businesses deal in specific types of goods and services not available on mainstream shopping sites, try combining the specificity of your products and services with your location. Try a voice search like this, “Hey Google, where would I find a cobweb broom in Louisville?” The result is for a cleaning service called Busy Brooms. If my company actually manufactured cobweb brooms, I’d take this as an opportunity to include some textual content on my website that parrots the question and answers it. This is a simple example, but you get the point.
Purchase Business Applications That Already Have a Track on AI:
As a small- to medium-sized business, you may be looking to implement software such as a CRM (Customer Relationship Management), IT systems maintenance scanner or even an e-commerce solution. At this point in time, you should research which of your choices incorporate a solid AI/ML roadmap. Spending capital dollars on software that does not have AI built into it is like buying a car that only burns gas, has a manual shift, and doesn’t have ABS or a rear camera. You may be the type of person who enjoys a manual transmission in your car, but are you the type of person who can analyze a mountain of data to predict the right moves to predict future sales? Probably not. Or if you are, you may have underestimated your business value!
Ready for What’s Next?
Excited to see how you can apply AI/ML to your business? Well, the robots have not invented a simple solution to every business marketing challenge (yet), so if you have any questions, we hope you’ll give us a call at (502) 499-4209 or drop us a line here. At PriceWeber, we pride ourselves on having a great team of people who are great at solving business marketing problems of all kinds, AND we speak robot as well! Let us know how we can help.