This content originally appeared on HackerNoon and was authored by Kirill Baranov
Over the past decade, dozens of home services marketplaces reached early traction – and then quietly stalled. They reached a few million in revenue, raised some capital, built custom software, and still couldn’t scale sustainably.
For a long time, it seemed more likely to be a problem of timing, funding, or market size. However, after building and operating a dog walking and pet care marketplace myself, I realised the truth was much simpler.
Most of these businesses failed not because of weak demand, but because operational complexity was too expensive to automate.
That used to be the case. Before the age of AI – and now it’s changing the equation.
Hi, I’m Kirill. I’m a co-founder of a dog walking and pet sitting marketplace and a product-led operator who has spent the past decade building two-sided platforms. I’ve worked across product, operations, fundraising, and supply growth – and most of my hardest lessons came from scaling real-world services (where trust and reliability are everything!).
Inside a “Hard” Marketplace
When we were building our platform, our service looked simple from the outside.
A customer needs someone to walk their dog => a provider shows up => the job gets done.
In reality, it was nothing like that. Our average customer ordered around nine services per month. Regular users booked close to twenty. Many needed the same provider on a fixed weekday schedule, similar to a personal tutor or babysitter.
At the same time, most providers were part-time. They chose which jobs to take. Which meant we couldn’t force schedules or easily assign shifts.
On top of that:
- Owners wanted consistency and trust
- Dogs had different temperaments
- Not every provider fit every client
- People got sick
- Plans changed
- Walks were rescheduled
- Services were bundled
- Multiple pets were involved
\
Basically, every seemingly “simple” order had twenty edge cases. It was nothing like a glamourised image of a “dog-walking Uber” – because the service was relationship-driven, recurring, and emotionally sensitive. And that made it fundamentally harder to scale.
Why These Marketplaces Keep Breaking
Looking back, most home services marketplaces struggle for the very same reason, in one way or another. They try to run high-frequency, relationship-based services on infrastructure designed for one-off transactions.
Traditional marketplaces are good at matching once.
Home services require:
- Continuous scheduling
- Long-term pairing
- Constant coordination
- Trust management
- Quality control
Each of these adds operational cost – and when accumulated, they absolutely blow up complexity.
Historically, this meant that companies could grow to $1–3M ARR and then hit a wall. Every new customer required more support, more coordinators, more manual work. Unit economics stopped improving, margins flattened, and growth slowed.
Why We Couldn’t Fully Automate It Back Then
In all fairness, we tried.
In the early days, orders were distributed in Slack, then Telegram. Providers replied with “+1” if they were available. Someone from the team assigned the job manually.
Later, we built our own CRM and mobile apps. We spent years automating workflows for:
- Booking
- Cancellations
- Rescheduling
- Messaging
- Reporting
- Payments
- Add-ons
- Insurance
- Substitutions
Each feature solved one problem – and then created two more. Ultimately, because the issue wasn’t vision – it was cost. Building end-to-end automation for dozens of edge cases required more engineering resources than most early-stage companies could afford.
So we compensated with human labour – customer success teams, coordinators, and operations managers. Which sort of worked – until it didn’t.
The Structural Shift AI Enables
What AI in its current state changes is not just efficiency – it elevates what is economically possible. For the first time, marketplaces can automate high-complexity workflows at marginal cost.
Three areas matter most👇
1. Intelligent Matching
In our case with the dog-walking marketplace, matching wasn’t about distance.
It was about:
- Dog behaviour
- Owner preferences
- Provider experience
- Past interactions
- Schedule compatibility
Previously, this lived in spreadsheets and people’s heads. Now, machine learning systems can learn these patterns continuously and improve with every transaction.
Matching becomes dynamic, not static.
2. Automated Coordination
Most marketplace friction happens after the match.
Who confirms? \n Who reschedules? \n Who follows up? \n Who reminds?
In our system, this required constant human involvement. Today, AI agents can manage most of this communication. They don’t forget, don’t get tired, and don’t scale linearly in cost. They simply turn coordination into software.
3. AI-Driven Onboarding
Onboarding part-time providers is hard. They need to learn rules, standards, tools, and expectations – for a job that may only generate limited income. Historically, this required trainers and managers.
AI can now handle large parts of this:
- Explaining workflows
- Testing knowledge
- Answering questions
- Monitoring early performance
This lowers CAC and improves quality at the same time.
Evidence from the Market
We’re already seeing this playbook work in other verticals.
In healthcare, Counsel Health uses medical AI to handle intake and basic consultations, allowing doctors to focus on high-value work.
In recruiting, Paraform gives independent recruiters AI tools that let them run more searches with fewer resources.
In construction, Remi uses AI to manage permitting, inspections, and scheduling, enabling fixed-price offers.
In legal services, Lawhive combines AI with human lawyers to deliver flat-fee representation.
In talent marketplaces, platforms like Jobhire.ai and MeetDex are automating screening, interviews, and coordination.
Different industries, same pattern: AI replaces the invisible operational layer that used to kill margins.
What a Next-Generation Dog Walking Marketplace Could Look Like
If we were building our platform today, the product would look very different.
Imagine this experience:
A customer sends a message: “Can someone walk Max tomorrow at 8am?”
An AI agent:
- Understands preferences
- Checks history
- Finds the best provider
- Confirms availability
- Schedules the service
- Handles payment
- Sends reminders
No app navigation, no manual coordination, no friction whatsoever. Behind the scenes, matching improves over time. Providers get better job recommendations. Customers get more consistency.
The platform becomes invisible – and that’s the ultimate goal.
Why Now Is Different
We attempted pieces of this years ago, but the technology wasn’t ready back then.
Today, three things have changed:
- AI agents can reason across workflows
- Development cycles are dramatically shorter
- Experimentation is cheap
And we’re still in the very nascent stage of the AI automation era. But even now, founders can already test complex automation in weeks, not years. Iteration loops are faster, mistakes are much cheaper to make, and learning is continuous.
This fundamentally shifts the risk profile of building “hard” marketplaces.
What AI Still Won’t Fix
AI is not magic. Or at least, it’s not quite there yet – so some constraints remain.
Marketplaces still fail when:
- Supply is structurally scarce
- Quality is purely subjective
- Trust can’t be systematised
- Incentives are misaligned
AI amplifies good systems that can and should be streamlined, but it doesn’t replace them. So say, if a marketplace failed at $1M ARR because of poor fundamentals, AI likely won’t save it (though it might prolong its existence). But if it stalled at $30–50M ARR because of operational overhead, AI very much might give it a boost.
From “Do Things That Don’t Scale” to “Build What Finally Can”
If you’re no stranger to the startup world, you’ve certainly heard the “do things that don’t scale” mantra from Paul Graham’s essay, where he argued that early-stage founders should use manual, non-repeatable work to learn deeply about their product, users, and workflows before trying to optimise and automate.
That’s exactly how we survived at the beginning – with manual matching, personal calls, spreadsheets, and endless coordination. We did things that clearly wouldn’t scale long-term, just to understand the problem deeply enough.
For years, that was the only way to build marketplaces in complex, trust-heavy categories. You learned first, patched with people, and automated later – if you had the resources.
What’s different now is that founders don’t have to stay in that phase for so long. AI doesn’t eliminate the need to understand your users or your workflows. But it dramatically shortens the path from “manual chaos” to “systemised operations.”
Many of the processes that used to require large teams and years of engineering can now be prototyped, tested, and automated much earlier. This doesn’t mean skipping the hard work.
It means the hard work finally compounds.
So if I were starting again today, I wouldn’t build a marketplace first. I would build an operating system for trust, powered by AI, and let the marketplace emerge from it.
That’s where the next generation of category leaders will come from.
This content originally appeared on HackerNoon and was authored by Kirill Baranov
Kirill Baranov | Sciencx (2026-04-16T18:24:31+00:00) Why Home Services Marketplaces Were Unscalable – Until AI. Retrieved from https://www.scien.cx/2026/04/16/why-home-services-marketplaces-were-unscalable-until-ai/
Please log in to upload a file.
There are no updates yet.
Click the Upload button above to add an update.