All Posts
Computer motherboard with AI chip. AI implementation for small business isn't out of reach.

AI Implementation for Small Businesses


12 min read

Unless you’ve been hiding under a rock for the past decade or two, you’ve probably heard about artificial intelligence (AI) and machine learning (ML). It’s 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 is now the cornerstone of big businesses like Amazon, Google, Apple and IBM. It continually reforms 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 into AI implementation for small businesses. Learn how small (and even mid-size) businesses can benefit from AI and machine learning. We’ll explain how to take advantage of the AI that’s already embedded in your business (whether you like it or not).

Whether you’re actively involved in a project to integrate machine learning into your company or simply managing a small website with a manual “open” sign, comprehending AI’s potential to boost your business profits is essential. Understanding how to leverage AI effectively can significantly impact your operations and marketing strategies, regardless of your business scale or industry.

To simplify things, we will divide ways to leverage AI into two categories: active implementations and passive ones.

Active AI Implementations

“Active” implementations are the types of projects that require a significant level of planning, capital expenditure, and execution. For managing such projects effectively, it’s advisable to establish a well-formed committee and secure top-down support. These initiatives demand thorough upfront planning and meticulous project execution to guarantee success. With a structured approach and strong leadership backing, you can navigate the complexities of AI implementation and achieve desired outcomes for your business.

Anomaly detection

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. Amazon Lookout for Vision offers a streamlined process. You can link a device to cameras and upload images, encompassing both quality products and defective ones, to swiftly train a machine learning algorithm. In just a matter of hours, this algorithm can achieve high accuracy in detecting defects, facilitating efficient quality control processes.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.

Assembly line quality control

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 feed quality assurance data from the assembled product into Amazon Lookout for Metrics. This guarantees that each product that gets shipped is 100% perfect.

Branding and 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. Your company can take advantage of that, too. For businesses that cater to specific audiences who appreciate greetings 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. Imagine how much more willing your clients would be to engage with an AI for routine inquiries and call routing if the AI’s voice sounded personalized and familiar to them. 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.”

Transcriptions

The possibilities are exciting! Plus, the AI can transcribe the call into text. This text can then be used to provide your call center personnel with relevant contextual information during the call. Maybe that will end the need to give your information over and over again every time you are routed to a new person on the phone. (Fingers crossed.)

AI-powered chatbots

Most of us are familiar with the website chatbots, which can be helpful for answering commonly answered question in a tree format. The latest AI-powered chatbots aren’t limited to predefined responses. They can be trained to have personalities similar to human counterparts, as mentioned earlier. In fact, if you think of a natural language assistant or a chatbot as a “person,” then they are technically the same person, sharing the same brain!

Engagement

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. An A/B test can offer marketers valuable insights into user preferences. It can assess the popularity of different variants, such as trigger words or graphical representations, depending on the popular tastes. After the test is complete, marketers can adjust what displays 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. We use them to get a better idea about how to position a particular campaign. This is based on 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. However, in the end, it’s limits are the size of the group polled and the passage of time, which tends to reveal changes in consumer behaviors.

Reinforcement learning

AI and machine learning can be used in a method called reinforcement learning. This tests 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. Then, it 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. It also has optimistic “faith” that certain content will be successful based on what is derived from other inputs like polling and training by the business marketers.

A brain with connectors and a robot grabbing a box representing AI implementation.
AI can use different algorithms to continuously balance exploration and exploitation.

Passive AI Implementations

What we’re terming “passive” AI implementations showcase how your business can augment the benefits of AI already serving your marketing needs. These operate independently of your direct control, offering an effective way to enhance your marketing efforts. Every size business can take advantage of billions of dollars in research and development. How? 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.

Just as your company tunes your site for better search engine optimization (SEO), you can also optimize your website content for voice search on mobile devices and home assistants like Google and Alexa. This ensures your content remains accessible and relevant to users across various platforms and devices. 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. Mobile voice searches are around 40% and home assistant searches are 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. Although that number has not been seen yet, it continues to rise every year. This is especially true after the holiday season when the popular devices become gifts. There are also growing number of devices and applications that have electronic voice search assistants built into them. These can include cars, phones, TVs, and other devices.

Natural language and semantic search results

Today’s AI in voice search correlates natural language terms into semantic search algorithms instead of just processing keywords. Since Google’s “Hummingbird” update in 2013, the search engine’s AI can process not only the words you search for but also the intent behind them. This results in more accurate search results for all queries, particularly for voice searches. Today you can initiate a voice search for something, someplace or someone specific, then continue to ask more questions related to the topic of that search.

How we tune sites for voice search are the same methods we implement to create better SEO in general. Some of these are:

  • Making sure to include information in plain text or HTML instead of embedding it 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.

Local searches

In addition to these (and many more) methods to increase SEO, you can enhance voice searches by having an up-to-date Google business listing. This 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.

Trial and error

Voice search trial and error is another method that can 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 might enjoy 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 for future sales? Probably not. Or if you are, you may have underestimated your business value!

Get Help With AI Implementation for Your Small Business

Excited to see how you can apply AI/ML to your small business? Well, the robots have not invented a simple solution to every business marketing challenge (yet). So, we hope you’ll give us a call at 502-499-4209 or drop us a line here. At PriceWeber, we have a great team of people who are great at solving business marketing problems of all kinds. (We speak robot as well!) Let us know how we can help.