AI alone doesn't solve problems, understanding problems does
When I think about integrating AI into a product, I don't start with the tech, I start with the person on the other side of the screen, what they are trying to achieve, why, and where they struggle.
Every time I speak to someone using a product, an Uber driver, a parent using ChatGPT, a host using Airbnb, or a business owner using Shopify, I’m interested in how they experience it. Trying to find where things don’t work as expected is something I do all the time, in and outside of work, and over time I’ve realised it has helped me spot pain points faster and understand what really matters to users.
It's always better to start with the problem, not AI
Two things I’ve learned from asking so many questions, users do not care about AI, they care about the experience and how fast they get results, and the AI feature that matters most is the one that moves them closer to what they actually want.
So, the best way to shift to problem-first thinking is to separate the how from the what. For example:
The how: “I need an AI tool to summarise meetings."
The what: “I’m losing four hours a week going through transcripts trying to find action items.”
The “how” is the solution, and the “what” is the problem, it highlights the issue and shows the consequence.
You can apply this to any product. Pick an app, speak to someone who uses it often, ask a few questions, understand their problem, and suggest a solution. Then test that idea with people you trust. Repeating this helps you get better at spotting real problems.
For this example, I’ll use Checkatrade. I use it when I need work done at home, and those jobs take time, so I’ve had enough conversations with tradespeople to understand how they use it.
A quick example using Checkatrade
A bit of context first: Checkatrade is a UK platform that helps homeowners find tradespeople. In 2025, they launched TradeMore, a job management platform for trades, and later added AI features, including an AI Receptionist that answers calls and captures job details. This context matters because it shows a move from a services directory to a platform for tradespeople.
When I was getting work done at home, I used Checkatrade to hire a plumber, an electrician, and a floor specialist, and we spoke about the apps they use to get work. As a user of the app, I already knew what worked and what didn’t for me, but I wanted to understand their experience.
I avoid asking what they like or which features are missing, because those answers often lead to false/weak signals. Still, they help start the conversation. A simple way to find the core value of an app is to ask: “If you didn’t have this app, what would be the hardest thing to do?”
What I found was interesting. No one mentioned replying to customers, checking WhatsApp, or managing emails. They didn’t care which platform they used. They cared about getting work. From what I saw, the problem for trades is not managing jobs, it’s finding them. Most see software as a cost unless it brings in more work.
That’s why Checkatrade launched TradeMore, but it still focuses on admin and time saving. The long term direction looks like a shift towards a platform that helps trades run their business. If that’s the case, the real product question is not what the next AI feature should be. It’s how AI makes the platform the place where trades run their business.
The missing feature is more demand
Conversations about home services happen all the time across social platforms like Nextdoor, Facebook, Reddit, TikTok, YouTube, and Instagram. They are spread out and hard to track. One way to help trades connect with that demand is to let them tap into those conversations through ads and sponsored messages.
They could partner with a service like Sprinklr to pick up these conversations, filter them by urgency, location, and intent, and show a prioritised feed inside TradeMore. This helps trades spot opportunities, respond faster, and turn them into jobs. But most are busy during the day, on site, answering calls, quoting, and driving. If they still need to chase leads, they won’t use it.
Here’s the problem: adding a new AI tool is only half the job, the other half is making sure it fits how people already work.
Turning availability into booked jobs
Tradespeople run a traditional business and are wired around the 4 Ps, and “Promotion” is one of them. So, that’s a good place to start. For example, if a plumber has availability next week, the system could suggest promoting their services locally, prepare a campaign based on their profile, and help reach the right audience. They only need to review and approve it. The more TradeMore understands how plumbers work, the more it can automate, including ad campaigns across social and search. Meta allows ads by postcode or radius, targets recent movers or new homeowners, and supports instant forms so customers can request a quote without leaving the app.
Tradespeople don’t want to spend hours managing ads, and many don’t have the time. If you say to them, “Here’s a tool to manage campaigns”, they won’t use it. If you say, “Our platform helps you find work when you’re available”, they will use it, and may even upgrade their subscription to unlock that feature.
Another option is to focus on sponsored posts. Meta’s Advantage+ can use images of local issues to reach the right audience. TradeMore’s dashboard could show where problems are happening, where posts were published, and which forms bring enquiries or jobs. This becomes an AI marketing assistant that promotes services based on real demand, profile, reputation, availability, and budget.
Personally, I think TradeMore’s AI direction makes sense. The job of the AI Receptionist is not only to answer calls, take bookings, and handle payments. It also collects useful data about the business and the demand around it, such as common problems, pricing, location, and types of jobs. This can help understand each tradesperson’s business and in the future focus on promoting businesses outside the platform.
From product to platform
The AI Receptionist is a good starting point for building a platform, because automation needs data. Over time, AI will likely take a bigger role in solving the demand problem. The goal, at least how I see it, should be to make the platform more proactive in helping trades find work, so a plumber who relied on directories can now see where demand is and approve campaigns from their phone while on a break.
What’s interesting about the Checkatrade case is that instead of adding more features to their existing site, they decided to create a new product with more advanced tools, some powered by AI and separate pricing. Right now, using TradeMore is optional, but as the platform improves and helps trades manage and find work, more trades will use it.
This post shows how problem discovery works. Of course, the more you know about the industry, the product, and the users, the better your analysis will be. If you do this outside work, pick a service you use, talk to users, find one problem, and see how it fits into a longer term AI strategy.
Until next time.