How Small Businesses Can Start Using AI Today

“Let me start this blog by saying that the threats of your business extinction due to not deploying Artificial Intelligence (AI) are greatly exaggerated. “

The first paragraph should, I hope, set the tone of this blog, which I hope will be a brief guide through the choices available to add AI to your business. First, AI is not new have a look at the below;

The UK government definition of SMEs encompasses micro (less than 10 employees and an annual turnover under €2 million), small (less than 50 employees and an annual turnover under €10 million) and medium-sized (less than 250 employees and an annual turnover under €50 million) businesses. (UK.Gov)

This means that in the UK, many Small Businesses are struggling to see through the constant big-tech propaganda and gain a rounded understanding of what AI can do for them.

Testing the waters

Starting with ChatGPT / Claude / Google AI / Microsoft co-pilot I will not compare the leading chatbots but take you on a very short journey through how they can help you.

The first task to run is to define what technology stack you are currently using.

Examples could be running Google Workspace for Business with Gmail and Google Drive, with the office stack being Google Docs/Sheets and Slides, et al. 

Alternatively, you could be running a Microsoft stack comprising Microsoft Office 365, including online email and the usual office suite of products.

This matters because you are looking to get a simple integration into the tools you are using. In both these cases, the respective Microsoft Co-Pilot would be best for the Microsoft stack and Google AI for Google for ease of use and rapid deployment.

The caveat for the above is that this is for full integration; my advice is always to try first. If you are looking for certain tools, this could be content creation, research or a marketing plan. Always carefully measure the impact of time saved and costs before taking the plunge.

I like Dr Who, but I don’t want a Dalek.

I prefer Darth Vader myself, but that doesn’t mean he is going to turn up in my company and enslave all the employees!

Like Cloud Computing before, AI hype is incredible. Is it a generational event? Yes, but the robots are not coming for you or the company you reside in.

The way to think about AI is to view it as a computer translator that speaks machine to other machines, a Rosetta stone for the modern age. 

“In time like the Alexa command that plays that soundtrack you demand, AI will be the bridge between that which is human, your voice.” 

At the time of writing this blog, OpenAI (ChatGPT) has released voice mode. The voice journey has begun!

What does this mean?

The technology will be simplified to the point that voice commands with screen outputs will be the way we control our devices. 

My advice is to plan for this eventuality within the next 18 months and look to understand how (if at all) this could affect your business. 

Think about voice ordering at a fast food kiosk; you speak first, and the return is always voice and pictures. (see previous blog here)

AI use cases

The easiest win with any AI is the generation of documents, plans, images and being a hella smart assistant.

1. This then has the benefit of saving time on repetitive tasks
2. Planning and using AI allows you to understand customers better
3. Use AI as a sounding board. Ask it to grab the latest thinking (not all chat-type programs can access the Internet), enabling you to make business decisions based on the latest thinking in your business sector.
4. Automate processes using AI, such as calendar booking and calling out to confirm details or appointments. (requires an API system, which we cover next)

Public LLM / Private LLM / API’s / LLM wrapper / Vector store

Confused? LLM stands for ‘large language model,’ which is what most AI tools are based on. 

The training for these tools has allowed them to predict ‘what comes next’ based on algorithms and reasoning. In simple terms, it’s a prediction machine.

Public LLM

This is a typical use case LLM such as ChatGPT or Claude. In these scenarios, you would subscribe for a monthly fee (around £16-£20) and use the interface or chat window provided. 

This can take the form of a native app or be accessed through a web browser. As this rolls out in your company, the subscription costs over time may require you to rethink your costs. 

Public models are subject to constraints such as the number of messages, even with “pro” or paid subscriptions. 

Private LLM

Thanks to the Open Source (free software) movement, a number of LLMs or, for simplicity, chat systems are available free of charge for you to download and use on your own computing hardware.

The advantage of deploying in this fashion is that nothing goes outside the server. This ensures data privacy (within the LLM, physical and system security are essential), zero licence costs, and restrictions on usage. 

Working with an IT partner and scoping correctly means that a number of secure integrations can be built from the network and hardware you control. A good resource to see the possibilities would be Langchain, which you can find here: https://github.com/langchain-ai/langchain.

Quick tips to begin your private LLM project

1. Create the use cases. These could be a customer support chatbot, data analysis, service or customer update service, document generator, project manager, or access to your CRM tools (customer management) to analyse data with graphs on a deeper level. 

2. Good hardware choices for IT hardware go with a well-supported brand with a great partner ecosystem. The specifications below are for LLama by Meta:
   • A great GPU (such as NVIDIA A100 or RTX 4090)
   • At least 64GB of RAM
   • Fast storage solutions (SSD/NVMe drives)
   • Reliable power supply and cooling systems

The initial investment in hardware may be expensive, but over the long term, if privacy is key, this is an excellent path forward.

Many companies do not retain the necessary IT skills to set up an LLM environment.

I have provided the link below, which is for TD SYNNEX, a Lenovo hardware distributor. They can independently find you a suitable local partner to work with. https://trustedadvisor.tdsynnex.co.uk/vendors/lenovo/contact/

The API conundrum

What the hell is an API, you may ask? It is an application programming interface, one would say. If that is not clearer, don’t worry—join the majority of business owners.

An API is a way for two systems to “chat” in machine language to pull information. Let me give you an example.

You go to that holiday booking website and decide, I want to go to Florida, I want the best hotels 4-5* and it has to be near the Universal Studios. The travel website then “calls” the various hotel platforms, airlines and mapping services. The call then comes back from the external platforms as an API. This completes the page you are viewing with all that glorious information, including user reviews. Then, when the card is put in, a final API is sent to the merchant to process

The API services from public LLMs are great for chaining together various actions and services. 

They are like a business canvas that slots together to form an outcome for you. 

A note of caution: Be careful and make sure you add cost limits to ensure you do not have a nasty surprise bill. If you are struggling and need help, then feel free to reach out via this website.

LLM Wrapper

You will come across this term when doing AI research on any platform.

An AI wrapper is an API that has been configured into a product and trained within that product. An example is online AI writers that are tuned to produce an outcome.

The majority of AI-driven products online are created in this form; the issue though is when the models are changed, you will find many wrappers based, for instance, on ChatGPT 3.5. 

When buying an AI service online, check the model used and always ask what they do regarding an upgrade. Be careful. 

There are a lot of “zombie” services based on an old model sitting there collecting subscriptions offering considerably less than a public LLM subscription. Email the support and ask these questions.

Vector store

A vector store is a database in the middle of public LLMs, private LLMs and the deployment choice you made. 

A vector store allows you to ‘cache’ or store previous searches and answers and add documents and learnings to create your own database.

Vector stores integrate with the majority of AI tools. An excellent use case is the enrichment of queries within private stores and the security this provides.

Conclusion

AI doesn’t have to be complicated or expensive. Even the smallest businesses can benefit from this technology by starting small and focusing on practical applications.

Plan for the future, but look at the present: where can you save on repetitive tasks, where can AI add value, what can you automate and evaluate, and where can you cut costs?

Above all, do not be afraid. Technology should serve you, not the other way around. Harness technology to become one with your business, making it personal and sound like your brand and voice. And if you get stuck, never be afraid to call on the vendors and their partners to assist.

Above all, experiment, iterate, iterate, and make it right for your needs. Be open and ready for unexpected use cases.

Thank you for reading this blog. 

[^We do not sell AI services, but if you need help, reach out using the various forms on this website, and we can point [you in the right direction.]