AI, Your Business, Your Customers, and Your Team
As AI becomes increasingly integrated into business operations, organizations must ask an important question: How will you manage your business while embracing AI and simultaneously creating policies that protect your company, your customers, and your reputation?
Protective policies are essential. Businesses need to demonstrate to customers that their best interests remain a top priority. While AI can improve efficiency and reduce costs, companies must carefully evaluate where and how it is used.
We have many customers who do not want us using AI tools within their environments. This includes AI-powered meeting documentation applications. They do not want AI systems extending into their networks, communications, or business processes. Frankly, I understand their concerns.
One area that deserves particular attention is AI-generated meeting documentation. Recording meeting notes without providing an opportunity for review before they are saved or distributed is risky. Human oversight should be a required step. Unfortunately, many AI meeting-note platforms are still somewhat buggy when it comes to ensuring proper review and approval before notes are automatically shared with attendees.
The challenges go beyond simple technical glitches. Tone and messaging can easily be taken out of context when conversations are summarized by AI. In many cases, important details are omitted, key decisions are missed, or information is simply incorrect. While these tools are improving rapidly, they still have a long way to go before they can be considered fully trustworthy without human review.
Recently, I have been learning AI development by creating a private application. The process has been fascinating, but it has also reinforced an important lesson: AI is not a one-stop solution. The developer or designer must remain focused, understand the project's requirements, validate the data, and ensure the work stays aligned with the intended objectives.
For my application, I used Antigravity to develop general specifications from my sketches, use cases, and written descriptions. Then I used Kiro to refine those specifications and create a detailed project plan. Those documents were introduced to Claude, which assisted me in setting up Supabase, Expo, and GitHub to build both a mobile application and the accompanying website needed for a legitimate storefront.
The development process was both fascinating and productive. Yes, AI generated much of the code, and Claude helped guide testing efforts, but success still required dedication, persistence, and hands-on involvement. Building the project was not entirely different from managing any other IT initiative. Challenges arose, frustrations occurred, and problems still needed to be solved. The difference was that the amount of development time required was significantly reduced, which was remarkable.
My application was not particularly math-intensive. It was more of a combination of multiple AI tools working together with a database designed to capture user experiences. It did not require advanced analytics, layered truth validation across multiple data sources, or complex algorithms. The testing process was relatively straightforward, focused primarily on verifying that data associations and relationships were functioning correctly.
Even so, Claude made coding mistakes. Initially, it was reluctant to acknowledge them and often attributed problems to user error. After some discussion about accountability and error ownership, the interactions improved considerably. While this was occasionally frustrating, it highlighted an interesting aspect of working with AI systems: you can often create a more effective working relationship by establishing clear expectations and sharing your preferred communication style.
Another concern with cloud-based AI tools is data security. Organizations must be extremely careful about the information they share with external AI platforms. This is especially important when working with existing databases, application code, internal processes, or proprietary business information.
My recommendation is simple: if AI-assisted development requires exposure to sensitive databases, table structures, proprietary code, or confidential business logic, consider using AI tools that run locally rather than in the cloud. While local AI solutions may not eliminate all risk, they can provide an additional layer of protection and reduce the likelihood of accidentally exposing sensitive information through cloud-based services.
AI is a powerful tool, but it is still just a tool. Success comes not from replacing people, but from combining human judgment, oversight, and accountability with the efficiency and capabilities AI can provide. Businesses that find this balance will be better positioned to innovate while maintaining the trust of their customers and protecting their most valuable assets: their data, their reputation, and their people.
