From Paper to AI, Building Software That Frontline Workers Actually Want
MaintainX | Nick Haase
Breaking the Bottleneck is a weekly manufacturing technology newsletter with startup perspectives, interviews, news, funding announcements, manufacturing market maps, 2025 predictions, and more!
💥 If you want to chat, feel free to reach out at aditya@machinafactory.org.
🏭 If you were forwarded this and found it interesting, please sign up!
From Paper to AI, Building Software That Frontline Workers Actually Want 🎙️💬
Nick Haase is the co‑founder of MaintainX, the leading maintenance and frontline work execution platform. He has spent thousands of hours on the shop floor helping businesses transform their operations with intelligent, frontline‑friendly software. He regularly writes and speaks about digital transformation.
He is the host of #TheWrenchFactor, a LinkedIn Live series exploring the emerging trends and technologies shaping the future of industrial operations and asset management.
Follow @MaintainX to get notified when the next episode goes live.
“The shop floor knows instantly whether you built something for them or for someone above them.”
You came from private equity and consumer startups like Loot! before co-founding MaintainX. What was the "magic moment" when you saw an opportunity to impact factory workers, and how did that shift your understanding of what industrial software needed to be?
I was doing consulting work for Fortune 100 companies on future technologies, and I came into manufacturing with a Silicon Valley mental model of “digital transformation.” I assumed the ecosystem would be much further along.
Then I started spending time inside plants, including best-in-class operations, and I was genuinely surprised by how much critical work was still done on pen and paper. That was the moment. It made something obvious feel urgent: you cannot become data-driven if you are not capturing the data digitally.
The more time we spent with teams, the clearer the stakes got. Labor constraints. Tribal knowledge aging out. And this wave of AI coming fast that will only be as good as the foundation beneath it. If your foundation is still paper, you end up with a huge chasm and a missed opportunity.
It also reshaped my view of industrial software. It cannot start as a corporate initiative and hope it trickles down. It has to start with the people doing the work, and it has to be simple, practical, and helpful in the flow of a shift.
You've said digital transformation is a "sales job" where shop floor workers are the customers, not just stakeholders. How do you handle resistance on the factory floor? What specific product features or behavioral tactics have actually worked to win over skeptics?
Most resistance is rational. People have seen tools come and go, and they have been asked to do extra work “for visibility” that does not help them.
So I start with one question: “What’s in it for me?” The frontline does not wake up excited to make leadership’s dashboards prettier. If the product does not make their day easier, it will not stick.
What works is reducing friction in ways technicians can feel immediately. Less chasing information. Less rework. Faster shift handoffs. Clearer context attached to the job. Fewer interruptions because the history lives with the asset, not in someone’s memory.
And the behavior matters too. You have to listen first, improve based on what you hear, and make wins visible in a way that feels supportive, not surveillant. The shop floor knows instantly whether you built something for them or for someone above them.
MaintainX scaled to 10,000+ customers initially using product-led growth in an industry notorious for multi-year sales cycles. How did you architect a bottom-up adoption model early on that forced corporate buy-in from the floor up, and what's the one friction point that still kills PLG motion in conservative manufacturing environments?
We built for the frontline worker first. The person with no budget authority, but who determines whether these initiatives succeed or fail.
Early on, the goal was to make the tool useful for a technician as an individual and as a team, and then make the value compound as more people participate. When that is true, adoption naturally moves upstream because the improvement shows up in the work itself.
Some of our strongest enterprise deployments started with a single champion at a single plant. Someone frustrated with the status quo who just wanted to make life easier for their team. The crew adopts it. The plant manager notices things are running better. Then regional leaders see the lift and want to scale it.
The friction point that can still kill PLG is the moment IT gets involved. And they are right to care. Nobody wants unvetted software inside an enterprise. So bottom-up only works at scale if you can quickly earn IT’s trust with security, controls, and real partnership.
You advocate for "Standardize to Scale" and digitizing tribal knowledge before veterans retire. But capturing undocumented intuition is incredibly hard. How can AI extract the "quirks" and edge cases that a 40-year veteran just knows? What's being lost in translation when we digitize expertise?
You cannot capture a veteran’s entire brain by asking them to list everything they know. The real context shows up at the point of work, when they are diagnosing a failure, making a repair, or handling an edge case that never appears in a manual.
So step one is making it easy to capture that context in the moment, in whatever format is natural. A quick note. A photo. A voice message. A few lines added to a checklist. Lightweight, immediate capture is how the nuance survives.
There is also a clear benefit for the veteran. It creates a record they can reference later, and it helps colleagues on other shifts solve issues without waking someone up at 2 a.m.
Where digitization can go wrong is when it flattens nuance into rigid structure. Some knowledge is procedural and should be standardized. But a lot of expertise is contextual judgment. The goal is to preserve that context and make it usable, not reduce it to a dropdown.
With recent notes on multi-agent failure detection and LLM-generated data models, you’re clearly moving toward “agentic” maintenance. What does the technician’s role look like in five years? What guardrails need to be put in place around the agent-human workflow?
There is incredible innovation happening in AI right now, but manufacturing is not an environment where you can tolerate risky failures. People can get hurt. Downtime is expensive. So the next five years look like trust but verify.
I think technicians will have faster access to the right information, clearer recommendations, and more proactive support, but still remain the decision-makers. AI should reduce search time, surface relevant history, and suggest next steps, not pretend certainty where it does not have it.
Guardrails start with grounding and transparency. Where did this recommendation come from? What evidence supports it? How confident is it? And can the system say “I don’t know” when it cannot answer safely?
That is how we think about it. Our AI tools should not roam the open internet or pull from random forums. They should rely on trusted, context-specific sources like manuals, work history, and asset data. If the answer is not grounded, the system should be explicit and route the user to the right human.
And in five years, technicians are still central. Even in more automated plants, humans remain the best mechanics and problem-solvers for keeping complex systems running. AI should amplify that craft, not replace it.
To contact Nick, reach out to him on LinkedIn here. He’s always happy to help and share more about MaintainX.