Crafting a Business Plan with ChatGPT: A Cautionary Experience

Introduction: The Challenge with Generative AI

It's widely recognized now that generative AI, like ChatGPT, can be unpredictable. When employed for uncertain tasks such as writing a book or designing an image, the AI can introduce elements that disrupt the cohesiveness of your work.

While minor unpredictability can be beneficial for brainstorming, it becomes problematic for meticulous activities like financial planning. Be ready to allocate substantial time to reviewing and correcting outputs; otherwise, errors could mislead you.

The Experiment: Using ChatGPT for a Business Plan

I embarked on an experiment with ChatGPT to formulate a business strategy, examining how effective it was at constructing hypothetical business plans over several weeks. Though the assistance was informative, inaccuracies were frequent.

(Note: Ziff Davis, the parent company of ZDNET, initiated legal proceedings against OpenAI in April 2025, claiming copyright breaches by the AI's training processes.)

The pivotal insight was simple: Extended interactions with ChatGPT increase the likelihood of error accumulation, which can turn the process frustrating.

Practical Application: Building the Business Model

In my project, the goal was to gather subscribers for a nascent newsletter by investing in advertisement campaigns. The process involved creating detailed spreadsheet models showing subscriber numbers, revenue, advertising costs, and financial forecasts.

ChatGPT was instrumental in generating Excel templates, offering flexibility to adjust assumptions like subscriber growth. It started with a prompt: “Please outline a straightforward plan to expand a subscription-based service from 250 to 10,000 patrons over three years, assuming an annual churn of 16%.”

Encountering Errors: An Inevitable Aspect

Initially, it was constructive. Errors surfaced as the journey progressed, primarily when computations became inconsistent. For instance, an incorrect assertion was made about turning profitable in month 11 when financial tables displayed profitability beginning in month 10.

After identifying the mismatch, ChatGPT acknowledged the mistake and suggested visualizing the profitability graphically.

Later, ChatGPT repeated similar errors, failing to incorporate foundational data such as the initial 250 subscribers count. This oversight indicated that vital assumptions could vanish unnoticed.

Consequently, it became evident that one must have a keen memory for essential details as AI technology can often overlook them.

Navigating Through ChatGPT's Inconsistencies

Odd interruptions, like calculating business lifetime value inaccurately, highlighted the need for manual oversight. ChatGPT would sometimes incorporate incorrect subscriber numbers and revenue figures.

For instance, the AI suggested a total of 9,200 subscribers at the end when previously the number was over 10,000. These kinds of shifts necessitated continuous verification.

Consistent errors necessitated persistent adjustment, akin to a game of error elimination with varying abstractions.

Conclusion: Striking a Balance with AI Assistance

The AI can provide groundbreaking solutions and insights without needing to search for external information, maintaining conversational continuity across different data sets. Still, its folly in logic and assumptions persists.

Despite AI's substantial aid in reducing the time for creating preliminary drafts, the time spent reconciling AI's mistakes dilutes this advantage. A personalized productivity measure might reflect reduced hours but increased error-checking.

Unquantifiable stress arises from continuous vigilance required to prevent incorrect assumptions from derailing the entire task.

As AI continues to evolve, users can look forward to enhanced consistency and better tools for maintaining variable fidelity within the system, though for now, vigilance remains key.