The reality of deploying AI for Businesses

I saved significant money and resources in my small business by leveraging the latest AI models. As a Software Engineer and Business Owner I’ve spent 1000s of hours experimenting through trial and error on what works and what doesn’t.

Personally, I don’t know many small to mid-sized businesses leveraging AI properly because they lack the expertise and workforce to set it up in a usable and secure way. On LinkedIn, X, Facebook, and other platforms there is no shortage of bullshit from snake-oil salesmen claiming obscene things.

To counter balance the hype I decided to write this blog series called “AI for Businesses” where I post examples of AI automations in my business, work I’ve done for my clients, general tips and learnings, and other reflections and thoughts.

The purpose of this first post is to set the correct framing. I ask you to leave your baggage of assumptions at the door. The current form of AI isn’t going to replace your workforce and run your entire business while you sip an aperol spritz on a sunny beach in Italy (although that does sound nice).

The Non-Hype Reality Check

To start with: AI can be dangerous as seen with the recent hack involving OpenAI, the creators of ChatGPT. When set up with incorrect permissions and insecure sandboxes it’s a security nightmare. AI agents will happily delete files, send clients unauthorised emails, expose confidential materials, and perform a multitude of other nasty things. Before starting to use AI automations on a deeper level for business you must first understand its limitations and the concept of “skin in the game” (who’s responsible when shit goes wrong?).

When I explain it to my clients I say:

Who will you blame when AI sends $1 million to the wrong bank account?

AI agents have no skin in the game, there is no punishing or firing them for misconduct. Remember this before you jump head over heels into AI utopia. A human and business must take responsibility for the actions and outputs of these AI agents, as seen in the recent case of a lawyer in Australia who received sanctions for fabricated AI quotes in court filings or Deloitte who reimbursed the Government for mistakes in AI generated reports.

Therefore a company that adopts AI without the correct knowledge, procedures, security, and other expertise risks severe consequences. The race to save time and money with AI instead leads to poorer work output, expensive litigation, reputation loss, professional disbarment, lower quality of service, and so on.

I would deem it absolutely reckless for any company to rush into adopting AI without professional assistance from an expert Software Engineer or legal consult. Someone must set up secure environments, provide technical consultation for automations, and assist with creating safe procedures to use the tech.

You must understand, this type of AI is not human-level intelligence, it’s called an LLM (Large Language Model), a subset of AI. I won’t bore you with the technical details but, in general, an LLM is an intelligence with great capability but, like everything in this world, it has its limitations. When compared to humans it excels at several tasks but is useless at others.

A person that thinks that an LLM in its current form will replace all workers is either misinformed, incompetent, or experiencing AI psychosis.

The non-hype truth is that the current form of LLMs helps to:

• automate repetitive tasks
• enhance skilled workers
• build custom automations that were almost impossible or overly expensive without AI
• free up team members for more critical tasks
• extract incredible analytical data that previously required a costly data scientist

AI is no longer a hype or a joke, it’s a reality that businesses are already leveraging today to save a ton of money and resources. I’ve personally done the same for my own business over the last year. I’ve automated things such as:

1. Raising invoices for my teams tracked hours from Everhour to Quickbooks.
2. Assigning work requests from a client email into actionable todos in my project management system Basecamp.
3. Transforming meeting transcriptions into assigned and actionable tasks for team members.

Although the above examples sound trivial they saved my business precious work hours that I now redistribute into higher value tasks. I continue to find useful automations like the above throughout my business and those of my clients that previously required expensive manual processing and costly software integrations which made them financially unviable.

There are risks with AI, sure. There are people over-hyping the tech, hell yeah. But, these aren’t excuses to dismiss or fear the technology. Serious technology requires serious professionals and guard rails. AI, when set up thoughtfully and methodically, is an absolute boon to businesses. One of the best technological breakthroughs I’ve seen in over a decade.

Next month I will publish the second post in this series that goes through an example AI automation in my own business and how I set it up.