I Let AI Guide Every Financial Decision for 30 Days in Kenya: Here Is What Actually Changed And It Was Not What I Expected

Experiment disclosure: This was an informal 30-day personal experiment. The AI did not have access to my accounts or make any decisions on my behalf. All decisions were ultimately mine. I used the AI as a mandatory thinking partner I had to consult before acting. Figures are accurate to my personal experience.

I gave myself a rule in June 2026: for 30 days, every financial decision above KSh 500 had to go through a free AI tool first.

Grocery budget allocation. An investment reallocation I had been considering for two months. Whether to pay an insurance policy annually or monthly. Whether to renew a subscription. Every single one before I acted, I had to open ChatGPT, describe the decision and my reasoning, and ask it to evaluate both.

I did not give the AI access to my accounts. I did not let it make any decisions. What I did was use it as a mandatory thinking partner, something I had to explain my financial choices to out loud, something that would ask follow-up questions I had not considered, something that had no emotional stake in telling me what I wanted to hear.

I expected to save some money. What I did not expect was what the experiment revealed about how I had been thinking about money for years.

📊  THE NUMBERS:  Over 30 days: 23 financial decisions above KSh 500 run through AI. Changed decision after AI response: 14 of 23 (61%). Estimated annual financial impact of changes made: KSh 68,400 redirected from waste and suboptimal choices toward savings and investment.

1. The Setup - The Exact Rules I Used

The experiment needed clear rules to be useful. Here is what I committed to:

• Every financial decision above KSh 500 required a ChatGPT consultation before acting
• I had to state the decision, my intended action, and my reasoning not just ask 'what should I do?'
• I tracked every decision and whether my action changed after reading the AI's response
• I could override the AI at any time but had to write down why in a notebook
• Spontaneous small purchases under KSh 500 were exempt  the rule was for considered decisions

The last rule was important. This was not about making AI approve every cup of coffee. It was about applying deliberate analysis to every decision that involved enough money to think about which turned out to be more decisions than I expected.

2. Week by Week - What Actually Happened

Week 1 - The uncomfortable audit

The first real test was running my monthly budget past the AI. I described my income, fixed expenses, and variable spending honestly like food delivery, transport, subscriptions, entertainment.

The AI's response identified something I had been blind to: my food delivery spending was KSh 3,200 higher than I had estimated because I had been mentally categorising delivery fees as 'food' rather than 'convenience premium.' My transport costs were above average for my route because I had defaulted to more expensive options out of habit. I had no allocation for irregular annual expenses like insurance renewals, vehicle service, family contributions, that were guaranteed to arrive every year but never planned for.

None of this was hidden information. I knew all of it, vaguely, somewhere. What I had never done was state it clearly to something that would ask follow-up questions and refuse to be embarrassed into accepting a charitable interpretation of my numbers.

💡  AI PROMPT:  "Here is my honest monthly budget for Kenya in 2026: salary KSh [amount], rent KSh [amount], food KSh [amount], transport KSh [amount], subscriptions KSh [amount], M-Pesa transactions approx KSh [amount], savings KSh [amount]. Please evaluate this budget honestly, where am I likely overspending without realising it, what categories are missing, and what is the single most impactful change I could make this month?"

Week 2 - The decision that changed

I had been considering moving a portion of my MMF balance into a longer-dated treasury bond I had seen advertised. The rate was attractive. Before acting, I ran the decision through the AI.

The 12-minute conversation surfaced three things I had not fully thought through: the liquidity difference between a 2-year bond and a rolling 91-day T-bill strategy, the inflation risk of locking in a fixed rate if Kenya's rate environment shifted upward, and the opportunity cost of removing funds from a liquid position I might need for an investment opportunity.

I did not make the bond investment. Not because the AI said no. it did not. It said 'here are the three trade-offs you have not named.' Naming them changed the decision. That is what I had not expected: the AI's most valuable contribution was not recommendations. It was questions.

Week 3 - Smaller decisions, bigger patterns

Week 3 covered smaller decisions, insurance renewal payment structure, a family contribution request, a subscription I had been meaning to review. What emerged was a pattern: most of my suboptimal financial decisions were not the result of not knowing better. They were the result of never stopping long enough to ask the right question before acting.

The AI was particularly good at the insurance decision: I was about to renew a policy monthly because I was used to monthly payments, without calculating that the annual rate was KSh 1,200 cheaper. A calculation I could have done myself in 30 seconds but never had, because the habit of monthly renewal meant I never thought to question it.

Week 4 - The portfolio review that revealed a gap

The final week included a full portfolio review. I listed every investment account, current balance, and monthly contribution and asked the AI to evaluate whether I was on track for my stated goals.

The AI identified that my emergency fund buffer was under-target relative to my current monthly expenses, a gap that had opened as my rent increased without a corresponding adjustment to my emergency allocation. I redirected KSh 5,000 from a planned discretionary purchase to the MMF. A gap I had not noticed for three months, caught in 8 minutes.

3. What Changed And What Did Not

What genuinely improved:

• My awareness of spending patterns I had been ignoring not because I did not know, but because nobody was asking me to name them out loud
• Decision quality on medium-stakes choices - the AI's questions forced me to name trade-offs I had been avoiding
• My ability to explain my financial decisions clearly - 30 days of articulating choices to an AI made me significantly better at thinking about money out loud
• Approximately KSh 68,400 per year in redirected spending and investment improvements

What did not change:

• My emotional relationship with money -  AI cannot replace the psychological work of changing financial habits
• The quality of my instincts on small decisions - KSh 500 threshold meant I still made many unconsidered choices
• My fundamental income - AI can optimise what you have; it cannot earn more for you

⚠️  IMPORTANT:  The AI was wrong three times in 30 days. Twice on specific current rates that had changed since its training data. Once on an assessment of a financial product I had more context on than it did. Always verify specific current figures with official sources and always trust your own judgment on decisions you have more context on than any AI tool can.

4. The Unexpected Finding - What Changed Was My Thinking

I went into this experiment expecting to save money. I did save money. But the more significant outcome was something I had not anticipated: being forced to explain financial decisions to an AI changed how I thought about those decisions before making them.

The act of articulating a financial choice stating what I was doing, why I was doing it, and what I expected to happen, created a pause that most of my financial decision-making had never had. That pause, more than any specific AI recommendation, was the most valuable thing the experiment produced.

It is a form of what psychologists call 'implementation intention', the research finding that stating a plan out loud increases the probability of following it. Applied to money, having to explain a decision to something that asks follow-up questions forces the kind of deliberateness that most financial decisions never receive.

AI was the mirror. The reflection was always mine.

✅  KEY TAKEAWAY:  Use AI as a mandatory thinking partner for financial decisions above a threshold you set. Not to make decisions for you but to force you to explain them clearly before acting. That pause, applied consistently, is worth significantly more than any specific recommendation the AI gives you.

Have you ever run a financial decision past an AI before acting? Share what happened in the comments.

📖  RELATED READING:  Curiosity Earned the Cash: 10 Questions to Ask AI Before Any Investment Decision  https://www.thenetworthshift.com/2026/08/questions-to-ask-ai-before-investing-kenya.html the framework for using AI as a thinking partner on investment decisions specifically.

Disclaimer

This article describes an informal personal experiment. All figures are personal and approximate. AI tools can and do produce incorrect information, always verify specific rates and details with official sources. Nothing here constitutes financial advice. Consult a CMA-licensed professional at cma.or.ke.