Written by —
Our Marketing Lead Michaela tells you how she used AI to buy a house and how prompt engineering is a useful skill even in your personal life.
Written by —
🇫🇮 This blog is also available in Finnish; please find it using the language switcher on the top right corner.
Why prompt engineering is a useful skill
I couldn’t believe it was true; I’d just found a lovely property on the market for a price that was actually humane. Could it be real?
I quickly wrote a message to the real estate agent and booked an appointment to visit the property. The agent mentioned that it was highly sought after, with the listing gaining ~120,000 views within its first week and her receiving calls constantly. Then and there, I decided to place a bid. (A little bit struck by FOMO.)
Then the difficulties began. The house, built in 1956, presented financing challenges. Despite sufficient capital and securities, the largest banks in Finland hesitated to approve the loan because of the property's original ‘50s condition. Although I explained my past experience with real estate, banks have their own standardized methods to estimate renovation costs per square meter, leaving little to no room for negotiation. Initially, I trusted my guts and wrote down a renovation plan for a modest budget. However, the feedback I received from the bank was nothing shy of discouraging.
How prompt engineering helped me, step-by-step
- Inspection report summary: I uploaded the house inspection report to ChatGPT, prompting it to summarize the document and categorize its key findings. The resulting summary neatly organized the recommendations and issues needing immediate attention.
- Detailed renovation plan using property data: I provided ChatGPT with detailed inputs such as the house’s floor plan, location, year built, building materials, room dimensions, and property details. I asked specifically for it to generate a thorough renovation plan, outlining budget constraints, language requirements, and explicitly stating that the document would be reviewed by a bank.
- Calculations via prompting: This step involved iterative prompting for detailed calculations. For example, I had ChatGPT calculate how many liters of paint I'd need by first determining the total square meters of wall surfaces. Similarly, it calculated the required lengths of water distribution pipes based on provided dimensions.
- Personal customization of prompts: I included personal preferences in my prompts, asking specifically for features such as a fully redesigned kitchen, a proper washroom, installation of a fireplace, and an energy-efficient exhaust air heating system.
- Finalizing and validating the plan: After receiving the detailed AI-generated outline, I did the following:
- Cross-checked pricing against real merchant sources
- Listed reputable companies and experts to carry out the labor
- Clearly separated costs into materials/products and labor categories.
Truth be told, creating the final renovation plan still required several hours of my own input, but AI significantly shortened the process and made it easier for me. By effective prompting, the AI could identify critical renovation components, remind me of details I forgot, and suggest essential calculations for cost estimations. While not every AI-generated detail was flawless, verifying information and refining calculations was easy enough.
The result was a renovation report I felt genuinely proud of. A loan specialist from a major Nordic bank even remarked it was the most thorough renovation plan she'd ever reviewed!
genai with more customization and control
If you want to learn from professionals how we control AI for more customization and control, my colleagues Teppo Hudsson and Jyri Lintunen are hosting a webinar on the topic on Wednesday 21 May at 13.00-14.00. You can sign up here.
