Customer Challenges
Homeowners waste an average of $500+ annually on minor repairs that could easily be completed as DIY projects.
Unfortunately, 70% of DIY attempts end in frustration due to lack of clear guidance.
Contradicting Advice Overload
"Have you felt overwhelmed by conflicting advice when searching repair forums, leaving you more confused than when you started?"
Costly Professional Solutions
"Have you ever regretted paying hundreds of dollars for a professional to solve a problem in just minutes that you wish you could do it yourself?"
Endless Search for Relevant Help
"Have you spent countless hours sifting through Youtube tutorials, only to discover none address your specific situation?"
Holiday Repair Emergencies
"Have you experienced the frustration of a clogged toilet on Christmas Eve when no professionals are available until after the holidays?"
Customer Research
Together with the team, we designed an interview script with open-ended questions aimed at understanding users’ journeys, uncovering past experiences, and identifying pain points. Over the course of two weeks, we conducted interviews with more than 30 participants—ranging from first-time homeowners with no DIY experience to seasoned DIYers and professional handymen.
From these interviews, we synthesized recurring themes and distilled the findings into three distinct user personas, each representing a key segment of our target audience. These personas became the foundation for shaping product requirements and prioritizing features.
Users Persona
The Eager DIY Learner (Prioritized Persona)
This is the prioritized personal because:
The Eager DIY Learner, typically a millennial or Gen Z new homeowner, is a digital native who readily seeks online solutions. Frustrated by unclear instructions, they are highly receptive to MyHandyAI's personalized, AI-guided repair solutions. Targeting this segment first provides invaluable product validation and taps into powerful growth through their network.
The Experienced DIYer
This persona is deprioritized for our initial MVP because:
- They aren't actively seeking DIY solutions; their experience enables them to handle most of the work.
- Their motivation is finding trusted professionals for the jobs they don't want to do by themselves.
- However, we recognize the value in connecting users with reliable professionals. This essential feature is strategically planned for a future phase of MyHandyAI, broadening our appeal once our DIY foundation is solid.
The Service Provider
Integrating service professionals is a vital part of our long-term vision. However, for the MVP, the focus is on building a strong DIY community first, as the market already has many platforms connecting customers with handymen. This feature is strategically planned for a future phase to intelligently bridge the gap between homeowners and trusted professionals.
Customer Journey Map
This journey map illustrates the end-to-end experience of our prioritized persona, Liam, as he navigates home repair and improvement tasks.
Insights show that while Liam values confidence, efficiency, and minimal disruption, he often faces significant friction—from unclear problem definitions and an overload of conflicting online advice to difficulties executing tasks and sourcing reliable tradespeople. His emotions range from anxiety and frustration to cautious optimism, and ultimately to either relief or regret, depending on the outcome.
These observations directly informed the design of key MyHandyAI features—such as personalized, step-by-step guidance—aimed at reducing friction, building user confidence, and empowering successful DIY experiences.
Customer Insights
Following in-depth customer interviews and persona identification, a quantitative survey was conducted on Reddit, yielding over 200 responses.
A compelling 79% of homeowners expressed a desire to complete home tasks themselves, provided they receive clear, step-by-step guidance.
These findings revealed two critical insights, guiding our feature prioritization:
Would you prefer to complete more home tasks yourself if you had clear, step-by-step guidance?
Insight 1: Users desire customized solutions.
Users' top-rated feature, with an average score of 7.6/10, is "Smart home repair suggestions based on my home situation, tools, and my skills." This highlights a strong desire for personalized guidance that considers their unique setup, available tools, and current skill level, empowering DIYers with tailor-made advice.
How would you like to interact with the app when diagnosing your home repairs or providing guidance?
Insight 2: Visual guidance is essential.
Over 65% of users prefer visual methods for explaining home issues, with 34% specifically favoring photos and 31% preferring a combination of methods. This strong preference led "Customized repair guidance with images" to be the second highest-rated feature, achieving an average score of 7.3/10.
This foundational information directly informs the prioritization of our MVP features.
AI Model Selection and Input/Output
Choosing the Right AI Model: Our Selection Criteria
Selecting the optimal AI model is crucial for MyHandyAI's success. We carefully evaluate potential models based on several key criteria to ensure we deliver the most accurate, reliable, and user-friendly experience possible. Here's what matters most:
Accuracy & Domain Relevance
- Handle complex, multi-step reasoning for troubleshooting home repairs.
- Possess deep general DIY and home repair knowledge, with the ability to integrate with RAG (Retrieval Augmented Generation) for up-to-date content.
- Maintain an extremely low hallucination rate—less than 1%—because incorrect advice could lead to frustration or even damage.
Context Length & Memory
- A sufficient context window size to parse detailed user inputs, including photos, descriptions, and past interactions.
- Robust multi-turn memory to sustain long and complex conversations, ensuring continuity as users work through a problem.
Multimodality Support
- Image input, which is critical for users to upload photos of their specific problems for visual diagnosis.
- Generating text and structured output, including accurate step-by-step guidance, tool/materials lists (with affiliate links), and more.
- Voice input/output for hands-free guidance while users are working on a repair.
Latency & Cost
- Fast response speeds for conversational problem troubleshooting, ensuring a smooth and responsive user experience.
- Optimized cost per million tokens, crucial for the long-term scalability of our platform. We plan to use a light LLM for quick clarifying questions and a heavier LLM for detailed, in-depth guidance.
Model comparison table
Our Multi-Model AI Strategy for MyHandyAI
At MyHandyAI, we leverage a dynamic multi-model strategy to deliver the best possible experience for every user interaction. This approach allows us to optimize for both speed and accuracy, ensuring you get quick answers when needed and deep, reliable guidance for complex tasks.
Quick Diagnostics & Chat Support
For initial problem identification, quick clarifying questions, and responsive chat-based interactions, we utilize lightweight models like gpt-5-mini and nano.
Their low latency ensures a smooth, fast, and user-friendly experience when you need answers on the fly, making initial problem-solving feel seamless.
In-Depth Solution Generation
When it comes to generating comprehensive, accurate, and detailed step-by-step guidance for your home repairs, we rely on the robust capabilities of gpt-5.
This powerful model ensures you receive precise, reliable solutions, giving you the confidence to tackle even complex DIY projects successfully.
By intelligently routing tasks to the most suitable AI model, MyHandyAI provides a tailored and highly efficient support system for every step of your home repair journey.
Input/Output
AI Proof of Concept
The AI Proof of Concept for MyHandyAI successfully demonstrates the integration of our core hypothesis, rigorous LLM evaluation, and self-reinforcing data strategy into a cohesive system. Our initial tests validated that personalized, step-by-step guidance significantly boosts beginner DIYers’ confidence and task completion, achieving 90% accuracy. We rigorously benchmarked multiple LLMs across accuracy, latency, multimodality, and cost to identify the optimal model, while our data flywheel ensures each successful repair strengthens the system. This comprehensive validation noticastates only confirms the viability of our AI-driven approach but also establishes a robust foundation for future scaling, subscription growth, and delivering meaningful user impact.
AI Architecture and Tech Stack
Roadmap
Business Model
MVP and Product Demo
