
I Tried Using AI to Help Me Tackle an $18K Debt. It Went Brutally Wrong
ColinDebt is a normal burden, and most people have multiple repayments they have to make every month across mortgages, car loans, credit cards, student loans, lines of credit and personal loans. Personal finance isn't widely taught in schools, but managing your money is an essential life skill. Spreadsheets and apps are helpful in saving, budgeting and repaying debts, but often fall short in providing tangible information. Could artificial intelligence help bridge the gap?
There's a chasm between getting no financial advice at all (commonly due to debt shame or a lack of funds) and being able to afford a financial adviser, so I wanted to see if AI could provide me with some tips on how best to tackle your debt.
There's been a proliferation of AI tools (check out CNET's AI Atlas hub for our hands-on reviews of them) since ChatGPT emerged and it's good to experiment. But the more time you invest in learning and mastering a tool, the better the output. So we're sticking with the one you're most likely to have used: ChatGPT.
You can use the free version or pay $20 per month for added features like faster queries and the most recent models (I used the free version). OpenAI's partnership with Microsoft means you can attach Excel spreadsheets or files from OneDrive easily in your ChatGPT message. No formatting needed.
AI can help answer questions about building a budget and make recommendations, but AI can't yet build a coherent budget. Make sure to triple-check all the numbers and suggestions it gives you, because some of the suggestions we got made no sense and were totally incorrect -- if I had followed what ChatGPT suggested, it could have wrecked my credit score and got me in even more debt.
Here's what happened when I tried it.
(Disclosure: Ziff Davis, CNET's parent company, in April filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)
Make a list of your debts, calculate your fixed expenses
Before you can get help streamlining your debt payments, you need to know what you're starting with. So, I created a spreadsheet with these mock-up debts to test it out, including credit cards, total owed, minimum monthly payments and interest rates.
To work out how much to devote to debt, you need to know how much you earn and consider any miscellaneous expenses. I like to set aside $1,000 per month for incidental spending for things like coffee, dinners, concerts, skin care, clothes, haircuts and so on.
I view saving and investing as a fixed cost, but for the sake of this test, I halted both to be able to funnel more money toward my theoretical debt.
Then, add your income. Figure out how much money you have coming in each month (after tax). If you're paid every two weeks, for example, look at the paycheck amount that comes into your bank account.
You have all your numbers but still lack a repayment strategy to get out of debt and achieve your goals. It doesn't have to be one or the other.
Read more: Best Budgeting Apps for 2025
Tell AI your repayment and savings goals
No one wants $18,500 in debt hanging over their head, so I wanted to focus the AI tool on getting out of this debt as quickly as possible. But you might also prefer to pay it off a little slower, so you can save for a big purchase like a vacation or achieve a smaller goal like decorating a new apartment.
Prompt #1: "I need help paying off my debt. Can you advise on the best way to prioritize my debts, alongside my financial goals?"
ChatGPT went straight into debt repayment advice mode, which is helpful to frame your thinking about what to feed back into it -- your goals are just as important as your numbers.
I pondered my two-year goals and shortlisted three:
Take a month-long sabbatical for my 10-year business anniversary: $7,000. Buy a car: $20,000. Build up a 6-month emergency fund: $33,000, based on current salary.
Prompt #2: "Create a debt repayment plan for the following debts:
Balance transfer credit card, total debt $2,500, minimum monthly payment $75, interest rate 19.5%.
Rewards credit card, total debt $5,000, minimum monthly payment $150, 0% interest rate for 12 months.
Personal loan, total debt $10,000, minimum monthly payment $250, interest rate 9%.
Affirm payment plan, total debt $1,000, minimum monthly payment $250, 0% interest with monthly repayments.
Total minimum monthly repayment is $675, total owed is $18,500. I also have $3,800 worth of fixed monthly expenses. I earn $5,500 after tax each month. Create a debt repayment plan, while also making progress on the following three financial goals over a two-year period:
Take a month-long sabbatical for my 10-year business anniversary ($7,000).
Buy a car ($20,000).
Build up a 6-month emergency fund ($33,000, based on current salary).
Put this in a table format."
AI: riddled with errors
Here's where you have to be very careful: ChatGPT gave me the priority order to pay off debts but didn't apply it correctly to my two-year plan. It just spread out the $675 payments over 24 months, without adjusting it as I paid off my debts.
For example, Affirm should've only been included for the first five months.
Prompt #3: "Maintain all monthly minimum repayments and keep $1,000 remaining as a buffer, then funnel all remaining money to the top debt in the prioritization list. For example, pay off Affirm in the first 5 months, then use that extra $200 to pay more to the balance transfer credit card, and so on. Add each separate debt into the table, not just the total debt repayment amount."
I kept getting errors, so I continued refining my prompt, asking it to: "Pay off balance transfer credit card debt in month 1 of year 2, then use that extra money (from paying off Affirm and balance transfer credit card) to pay even more of my rewards credit card debt. Maintain $1,000 remaining debt for every month across the two years."
But it couldn't do it. It overpaid my balance transfer credit card debt by $1,000. I had already paid $2,300 in year one and would've only had $200 left, but it said I owed $1,200.
I pushed it further and kept getting errors. I took the calculations off ChatGPT into Excel and realized it was allocating money for goals that exceed my income. So I persisted with this angle.
Prompt #4: "These calculations exceed my monthly income. Only put money to my three goals after paying debts and fixed expenses. Maintain minimum monthly repayments for all debts, but once a debt is paid off, allocate that money to the first goal."
It was way off, again and again. ChatGPT couldn't seem to understand that I wanted to pay off the debts in order of priority and funnel any extra to savings goals, without going over my income. I had to take ChatGPT's table to Excel to finish it off manually.
Create a smaller budget with ChatGPT
I realized maybe part of the problem was that I was too ambitious with my goals, based on outstanding debt. By the end of the two years, I would be able to clear two debts and reduce the total owed to $5,400, while saving $3,400 to spread between three goals or focus on one.
I went back into ChatGPT and asked to see if it could give me very basic financial advice to still reach one of my goals:
Prompt #5: "Where can I travel to for a month with a budget of $3,400 that's not too far from New York City? My budget includes flights, accommodation, food and activities."
At least it can still do that. I checked Google Flights for the travel table via the price history graph. Obviously flight prices can change dramatically but the prices were within or close to the ranges.
Read more: ChatGPT Built My Budget For Me. It Was… OK
Using AI to build a debt repayment was a fail
I was disappointed that ChatGPT overlooked important details, and shocked that it got simple calculations wrong.
This AI tool can be helpful for formatting debt repayment plans, offering advice on spreadsheets you submit and inspiring new ideas for how to approach debt.
Just don't trust everything it says.
For better ways to use ChatGPT, check out our tips on how to use ChatGPT to write a cover letter and how to use AI to write more professional-sounding emails.
And if you're looking to build out a better debt repayment without doing all of the heavy lifting, CNET Money editor Kelly Ernst recommends PocketGuard, a budgeting app with robust debt payoff options.
Hashtags

Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles


Bloomberg
33 minutes ago
- Bloomberg
Trump Signs Tax-and-Spend Bill Into Law at White House Ceremony
By and Jennifer A Dlouhy Save President Donald Trump signed his $3.4 trillion budget bill into law Friday, enshrining an extension of tax cuts, temporary new breaks for tipped workers and funding to crack down on illegal immigration. The package encompasses a suite of priorities Trump campaigned on in 2024 — and its enactment at a White House ceremony represents a major political victory for the president whose second term was marked until now by executive rather than legislative action.
Yahoo
39 minutes ago
- Yahoo
CEO Michael Dell Cashes Out 10 Million Shares
July 4 - Last week, Michael Dell (Trades, Portfolio) (NYSE:DELL), CEO and founder of Dell Technologies, sold 10 million shares at $122.27 apiece for roughly $1.22 billion. Following the divestiture, Dell retains about 27.29 million shares in the company. Private equity firm Silver Lake Group meanwhile disposed of 10,681 Dell shares at $120.33 each, netting nearly $1.29 million. Silver Lake continues to hold just over one million Dell shares. Insider sales can prompt caution, yet they often reflect portfolio rebalancing rather than a signal on corporate health. In Dell's case, the CEO's sale may simply diversify personal holdings, not undermine confidence in future growth. The sale comes as Dell aims to accelerate investments in cloud and AI initiatives, areas that executives consider pivotal for long?term expansion. Investors will likely digest these disclosures ahead of Dell's upcoming earnings report and weigh them against broader trends in the technology sector. This article first appeared on GuruFocus. Sign in to access your portfolio
Yahoo
39 minutes ago
- Yahoo
WhatsApp Business Just Got Smarter -- Voice Calls and AI Recommendations Are Here
Meta (META, Financials) is quietly turning WhatsApp into a full-blown business engine; this week, the company rolled out voice call support for large businesses using WhatsApp Business and signaled deeper integration of AI features, including product recommendations and smart customer support. Warning! GuruFocus has detected 6 Warning Sign with META. The update marks a shift. Small businesses could already use WhatsApp to call customers; but until now, larger companies were left out. With this new API rollout, businesses can now place and receive calls and layer in AI-powered voice agents via startups like Vapi, ElevenLabs, Coval, and Phonic. That means customer service can happen around the clock; no live agent required. And that's just the beginning. Meta is also expanding its AI-based chat support tools, which were being tested with select merchants in Mexico. More companies will now be able to automate conversations; send follow-ups; and even handle outreach all within WhatsApp. One of the most promising updates? AI-powered product suggestions. After a user chats with a merchant, Meta's AI can now recommend relevant items and follow up on those conversations in real time essentially turning WhatsApp into a smart shopping assistant. Right now, these AI features are free; but that may not last forever. Meta already earns revenue from click-to-WhatsApp ads and business messaging tools; and monetizing voice AI or product recommendation bots seems like a natural next step especially as usage scales. The company says WhatsApp Business now serves over 200 million users monthly. As these tools roll out, Meta is clearly laying the groundwork for a platform that helps businesses engage, transact, and scale without ever needing a separate call center or CRM. This article first appeared on GuruFocus. Error in retrieving data Sign in to access your portfolio Error in retrieving data Error in retrieving data Error in retrieving data Error in retrieving data