How Small Businesses Can Start Adopting AI: A Practical Guide

Businesses are adopting AI to improve business outcomes and their competitiveness in the marketplace. According to Anthropic, small businesses have lagged behind larger enterprises in their adoption of AI. This article is intended to help leaders of small and medium-sized businesses (SMBs) get a start on AI adoption. No technical knowledge of AI is required for this article.

Why has AI been attracting so much attention?

AI has been getting a lot of attention. This is the result of a new kind of AI, known as generative AI, which can converse in ordinary human languages, generate content (documents, images, etc.), and take actions (send messages, etc.). Generative AI embeds within itself a vast collection of human knowledge from many knowledge domains (science, technology, literature, history, etc.). This allows generative AI to help people accomplish more in less time and become more knowledgeable.

The benefits of generative AI include:

• Easy to use - You can talk to the AI in your own natural language. AI is capable of having conversations in multiple human languages. Furthermore, AI carries knowledge across languages, so knowledge from an English-language source becomes available natively in other languages.
• Knowledge - AI has a lot of knowledge. Conceptually, AI's knowledge encompasses a very large library plus much of the public internet. AI draws on and integrates information from its vast body of knowledge when it responds to you. AI can help you discover, learn, and make connections on a variety of topics.
• Productivity - AI can save you a lot of time. Tasks that used to take hours, days, or weeks can often be accomplished in minutes.

Where we reference AI in this article, we are referring to generative AI.

How can I get started learning about AI?

The best way to learn about AI is to start using one of the popular chatbots. Ask questions and explore topics related to your business. Create drafts of documents or presentations. Upload files and request summaries or specific content. Create images (available in some chatbots). Get to know what you can do with AI.

The popular chatbots include ChatGPT, Claude, Gemini, and Perplexity. You can start using these and other chatbots for free. When chatting with AI, it helps to think of the AI as a knowledgeable and helpful assistant that you can talk to like a real person.

Some helpful tips to keep in mind as you begin using AI chatbots:

• Don't be afraid to challenge the AI if you don't understand something or agree with it, or if you think the AI left something out in its response. Remember, you are talking to a machine, not a person. Be direct - the AI won't take offense.

• Learn about how to "prompt" the AI. The "prompt" is what you say to the AI. You tell the AI what you want (the prompt), and the AI responds.

  If your request is more involved, it can help to give the AI some context around why you're asking and what you want to see in its response. In essence, AI works by connecting patterns in language (specifically, the meaning of the language, not the actual words), so giving the AI more context and detail helps the AI build responses which fit your needs.

• Encourage your employees to learn about AI and to see how AI can help them in their work. Companies with AI-savvy employees are probably going to be at a significant advantage over companies whose employees do not use AI.

• Avoid sharing any sensitive information with AI until you make sure the information will not be used for "model training" (see the section below "Can I share sensitive business data with AI?" for more information).

How can I make my website show up in AI searches?

AI is changing how people search the Internet. Rather than doing a Google search, many people are using AI chatbots to find what they are looking for. AI chatbots provide responses with recommendations, not lists of links. While traditional Search Engine Optimization (SEO) is still important, websites need to be optimized for AI search using techniques known by names such as Answer Engine Optimization (AEO) and Generative Engine Optimization (GEO).

You can try some searches yourself in the popular chatbots to see if your business gets mentioned. If the chatbot mentions your company, it's an implicit endorsement of your business. On the other hand, if your business is not mentioned, you're basically invisible.

AEO/GEO encompass technical details which are beyond the scope of this article. You may want to discuss AEO/GEO with your website professionals or with Agentricity. In general, you need to:

• Make sure your website provides content you want AI to see in server-rendered HTML - many AI crawlers do not execute JavaScript.
• Allow AI search (and possibly training) crawlers in your robots.txt.
• Use schema markup on structured data.
• Write answer-first content (question, answer, additional information).
• Keep sitemap and freshness signals accurate for recency weighting.

At the time of this writing:

• The popular chatbots do not offer "pay for mention" as this would undermine trust in their answers.
• OpenAI offers paid advertising to ChatGPT Free and Go plan users who are 18+. Ads appear below ChatGPT's answer, not integrated into it. See the ChatGPT ads documentation for more information.

What are the costs of using AI?

AI is not free. On the contrary, the underlying technology behind AI is extremely resource-intensive in terms of compute and energy. These costs are passed on to you in the form of subscription costs and/or usage-based costs. In addition, if you use any commercial AI solutions or need any custom integrations, these can incur setup costs as well as ongoing operational costs.

As mentioned above, anyone can start using the popular chatbots for free. The free subscription plans provide basic features and a limited amount of usage. It's enough to get started and learn about the capabilities of AI, but for serious business use you will need paid subscriptions.

These subscriptions can cost anywhere from around $20 per month per user to several hundred dollars per month per user. Your AI costs can add up quickly, depending on how many employees are using AI and the subscription tier they need.

In addition to subscription costs, you may incur other costs. For example, if you want to use an off-the-shelf commercial AI solution for a specific function in your business, such as handling inbound customer service phone calls, that solution will have setup/integration costs as well as ongoing operational costs. These ongoing costs may be fixed (subscription-based) or variable (usage-based).

With subscription-based pricing, be sure to understand the limits on your subscriptions. AI providers with subscription-based pricing may place limits on your usage. When the limits are reached, they may degrade or cut off subscriptions for a period of time.

With usage-based pricing, you only pay for what you use. Be sure to set a limit on your costs. AI providers with usage-based pricing should allow you to set an automatic billing limit, or at least notify you when you hit the limit.

Either way, you will need to consider how your business will handle hitting a limit and losing whatever capability the AI provider is giving you.

Businesses need to think about their "AI budget" in their overall expense picture. When sizing your AI budget, bear in mind that understanding ROI for your AI spend is a challenge businesses face today. The best you may be able to achieve is a qualitative assessment based on employee/customer feedback and overall business performance.

Can I share sensitive business data with AI?

You can share sensitive business data with AI, but you need to make sure the AI provider will not use your data in "model training".

When you have a conversation with an AI chatbot, depending on your subscription plan and your account settings, anything you say to the chatbot (or anything you upload) may be used in model training. This means your data can get "baked into" the AI's knowledge.

Once your data is in the AI's knowledge, your data can get woven into responses the AI makes to anyone. It is important to note if your data gets used in model training, the action is irreversible. Your data cannot be deleted from the knowledge.

While the risk of data disclosure is low, it is non-zero. Still, it is best to avoid a publicity nightmare with damage to your business's reputation.

For personal subscriptions, most popular chatbots enable the use of your data for training by default or via a pre-selected setting. Generally, there are account settings which allow you to opt out, but this must be done manually.

For business subscriptions, the popular chatbots do not (by default) use your data to train their models. For this reason, it is highly recommended you provide employees with accounts on a business subscription plan.

It is recommended that you verify your providers' current policies, as they are subject to change.

If you use commercial AI solutions within your business, you should have assurance in writing from the solution provider guaranteeing your data will not be used by any downstream AI providers for model training.

Regulated data (such as Protected Health Information, or PHI) should not be shared with AI until the AI provider's support for the regulated data has been verified and any legal agreements (such as a Business Associate Agreement, or BAA) are in place with the AI provider.

How do I figure out where to use AI in my business?

The number of ways you can use AI is virtually unlimited. As you learn more about using AI, and see how other businesses are using AI, you can begin to envision ways AI can help your business.

To help you come up with ideas for applying AI in your business, here are some basic building blocks describing the types of work where AI excels:

• Drafting - Creating a first version of anything written. Emails, proposals, job postings, marketing copy, social posts, standard operating procedures, review responses, and more. AI is much better at going from a blank page to 80% than humans are, and humans are better at the last 20%. Question to think about: "Where do we write similar things over and over?"
• Distilling - Turning a lot of text into a little. Summarizing meetings, long email threads, contracts, customer reviews, reports. Question to think about: "Where do people spend time reading things just to extract the few points that matter?"
• Transforming - Turning information into a different form, in whole or in part. Extracting structured data fields from invoices or PDFs, rewriting for tone or audience, translating to another language, converting a phone-call transcript into a CRM entry, turning bullet points into a polished document (or the reverse). Question to think about: "Where do we retype or reshape information that already exists somewhere else?"
• Retrieving - Obtaining information from everything your business already knows. Your business's knowledge lives in two places: words (policies, contracts, past proposals, product docs, CRM, ticketing system) and numbers (your POS system, your accounting system, your spreadsheets). Question to think about: "What questions get asked over and over where the answer exists somewhere - in a document, a drive, a database, in systems - but takes real effort to dig out?"
• Classifying and triaging - Making judgment calls at volume. Routing support tickets, flagging urgent emails, tagging customer feedback by theme, scoring leads, sentiment analysis. Question to think about: "Where does a person look at incoming items one at a time and sort them into buckets?"
• Advising - Being a thinking partner. Brainstorming, critiquing a plan, reviewing a document, role-playing a negotiation or a difficult customer, explaining unfamiliar topics (a contract clause, a competitor's strategy), pressure-testing a decision. Question to think about: "Where would a smart and knowledgeable colleague be useful to bounce ideas around with?"

Real workflows are usually chains of these. For example, a workflow to prepare a daily picture of the business for a morning meeting might be Retrieve → Distill → Classify → Draft → Advise.

1. Retrieve the needed information from business systems
2. Distill the information into what is needed for the analysis
3. Classify the distilled information into "routine" vs "needs attention"
4. Draft the daily briefing
5. Advise management in the meeting as a thinking partner (for clarification, options)

Chaining building blocks together is called orchestration. Today, you'll usually be the orchestrator - driving each step in a chat session. Orchestration can be done with traditional software, invoking AI at each step. Increasingly, AI is orchestrating the steps itself; this is what people mean by "AI agents".

As you look at your business and business processes, and consider the building blocks with the questions to think about, you can begin to formulate a strategy and roadmap for the use of AI.

You may want to focus on using AI internally first, where AI is assisting employees, and the only touchpoints with AI are with your employees. Then, over time and when you feel comfortable with it, begin to use AI with external touchpoints, such as customers, suppliers, and partners.

Keep in mind today's AI is like a knowledgeable human intern who is great at the cognitive work described by the building blocks, but needs human guidance and checks on anything where absolute accuracy matters.

How can Agentricity help with my AI adoption?

Agentricity can help with all aspects of AI adoption, including learning about AI, understanding the costs of AI, data protection with AI, figuring out where to use AI in your business, and optimizing your business website for AI search. Agentricity can also help with integrating AI with your business systems, customizing AI for your business, and training your employees.

SMBs can likely achieve compelling benefits from the use of AI even with modest AI budgets. The challenge is where to apply AI to get the most benefit for the cost.

Contact Agentricity for a free initial consultation on how your business can benefit from AI, or to find out more about our AI engineering and implementation services.