There have never been more automation tools. AI agents that can read your inbox, summarize your meetings, draft your replies, write your code, organize your week. Every month, another wave.
And yet most people we talk to are still doing the same manual, repetitive, draining work they were doing two years ago.
Why?
We spent the last few months trying to answer that question. We sat down with 22 people across roles, industries, and skill levels. About three-quarters had a technical background. The rest didn't. Roughly half worked in tech. The other half didn't.
We asked them what their work was like. What was hard. What they wished they could change.
What we found was four barriers that show up over and over, regardless of what someone does for a living or how technical they are. Three of them are personal. The fourth is structural. Together they explain why automation, despite being the most-promised category in software for two decades, is still mostly aspirational for most people.
Here they are.
Barrier 1: You can't see what's broken
The first thing that surprised us: when we asked people what was hard or frustrating about their work, most of them gave us vague headline complaints. "Juggling priorities." "Communication." "I'm just busy." Nothing specific. Nothing actionable.
So we tried a different question: "If you could wave a magic wand and fix one thing about your current workflow, what would it be?"
Suddenly, vivid pain. Specific tasks. Things they hated. The same people who'd shrugged five minutes earlier now had three answers each.
One person we interviewed told us:
"Yeah, I'm aligned. I've had like a ton of discussions. But just keeping everybody in the loop is such a tedious part of the job."
Another, describing a quarterly process:
"It would be like to eliminate this manual tracking for sure. Just takes too much of our time. It's done almost every quarter. And it's like the time, the week that it's being done, across the board, everyone hates it."
A third interviewee describing a recurring task:
"Sometimes I receive an in-store request asking me to align with the inventory online. The request is so... it's not standardized. Sometimes they provide correct store IDs. Sometimes the store ID has to be uppercase, but they provide lowercase. The format is not correct. I have to manually calibrate all the formats to make sure it's correct, put it into the machine, and click the button."
None of these people had thought of these tasks as automatable until we asked the right question. They'd been doing them for months or years. They were just the job.
This isn't a personal failing. It's how attention works. Repetitive tasks fade into the background. Your brain stops flagging them as worth attention because they're predictable. The discomfort of doing them becomes "Monday."
You can't automate what you can't name. And most people can't name what's broken in their work without help.
That's barrier one.
Barrier 2: You don't know where to start
Let's say you do see what's broken. You know exactly which task drains your week. You even have access to AI tools that could probably handle it.
Most people freeze anyway.
The problem: too many possibilities, no specific starting point. Decision paralysis applied to automation. When you can automate anything, you often automate nothing.
One person told us:
"How do you, like, direct AI to do it? How do you make the message to AI to do it? I think it takes the time to learn how to be more specific about it."
They knew the AI was capable. They just didn't know how to point it at their work.
Another interviewee described it more directly:
"I don't think it's very clear from the UI what else you can do. Because when you just say like a prompt, I'm like, 'oh, I have to think of what I need.' But I don't really have an option, right? I don't know what else it can do. So how far can I imagine my problem? In our previous conversation I realized I could ask it for a standup report on my email, or maybe create a Linear issue or something. So I think that was, like, not clear."
They elaborated:
"Maybe if you have, like, a dropdown or sample prompts or templates that I could probably use, I feel like that might help, rather than me having to completely think from scratch."
Another wanted something even more specific:
"There are so many. The tech stack is huge at this point. There are so many possibilities. So if someone could tell me which tool should I be using, which would be more efficient for me for each task that I'm doing. Like, somehow look at my Linear and tell me that, hey, for this task you should be using this instead of Claude Code or something. That would be really nice."
What people need isn't more capability. It's a specific starting point. A real example. A workflow someone like them built, working end-to-end, that they can copy and adapt.
This is one reason the same five workflows keep getting automated everywhere: morning briefs, weekly reports, inbox triage, standup notes, customer summaries. They're not the best automations. They're just the ones with public templates and obvious starting points.
Everything else stays manual. Not because it can't be automated, but because nobody's shown what "starting" looks like.
That's barrier two.
Barrier 3: The cost of starting is higher than the cost of not
OK, you've identified the work. You have a concrete starting point. Now you actually have to set it up.
This is where most people quit.
One interviewee told us about an AI tool their coworkers used regularly:
"I know OpenClaw, but I never tried it myself. I know my coworkers are using it, but I've never tried it."
When we asked why, they said the setup was too complex and time-consuming. They'd never even attempted it. The fact that their teammates were using it successfully wasn't enough. The perceived activation cost was high enough that they just... didn't.
Another was even more direct about the dropoff point:
"I think personally it would be more so for new users. Because there's going to be a lot of people that are seeing this stuff and they want to try it out. And when they go ahead and they throw in that NPM install, and then they get hit with 'you need a Claude install as well' on top of that... that's where I think a lot of people might just give up, not even bother continuing."
And underneath both quotes is the same rational calculation. Another interviewee put it directly:
"When we want to automate the workflow, we just want to save time and increase productivity. We don't want to spend so much time on the tools that are supposed to improve our efficiency."
This isn't laziness. It's math.
The whole point of automation is to save time. But for most workflows, the setup takes longer than just doing the task manually. You spend hours learning the interface, hours wiring up triggers, hours testing edge cases, and only after all that do you start saving time. For tasks that happen weekly, the time-to-payoff is months. For tasks that happen monthly or quarterly, the math literally never works out.
So people skip it.
Past experience compounds the problem. People have tried tools that promised the world and delivered a 40-step setup. They've spent time learning interfaces that didn't pay off. So even when a new tool actually works, the past experience makes them hesitant to invest.
Most automation tools assume you'll spend an hour learning before you get any value. For most people, especially the non-technical ones, that's where the relationship ends.
That's barrier three.
Barrier 4: Your organization won't let you
Here's the barrier that's hardest to fix, and the one we hear about most often from people working inside larger organizations.
Even if you see what's broken, know exactly where to start, and are willing to invest the setup time, your company may not let you actually do it.
One person we interviewed, who works at a large enterprise, described what this looks like in practice:
"Every tool we use either has to be client approved or internal approved. Once that is approved and the budget is given for that, then we would go ahead and purchase that tool person by person."
Approval. Budget. Per-person license. For every new tool. By the time the process is done, the moment of motivation that made you want to try it is months in the past.
But the deeper version of this problem isn't just "I can't get the tool approved." It's that even when you can use AI tools personally, you can't connect them to the systems where your actual work lives. The same person:
"I cannot connect my work environment with this. If I use this for personal purpose, I cannot connect my Salesforce. I cannot connect my Teams. Because it's sitting in my office laptop."
They have access to ChatGPT. They have tasks they'd love to automate in Salesforce. The two cannot be wired together. The automation they want requires bridging their personal tools and their work data, and that bridge is forbidden.
This isn't a big-company problem only. Another interviewee told us:
"I think big companies have some degrees of concern on privacy. We are allowed to use ChatGPT for sure. We're not allowed to use Gemini. I don't know why. I don't know why. But it's kind of like company choice."
"I don't know why." That phrase comes up a lot when you talk to people about their org's tool restrictions. The choices feel arbitrary from inside. Some tools approved, some not, no clear logic. You're left working around the rules without understanding them.
For a lot of people, this is the real wall. Not "I can't see what's broken." Not "I don't know how to automate it." But "I'm not allowed to."
This barrier is harder than the others because it's not about you. It's about your environment. And environments change slowly.
That's barrier four.
What this means
These four barriers are layered. Even if you get past the first one, you hit the second. Past the second, you hit the third. And even if you push through all three, the organization around you may stop you anyway.
Together they form a wall most people never get past, even when they know exactly what they want to fix. That's why workflow automation, despite being the most-promised category in software for two decades, is still mostly aspirational.
We think the wall is breakable. Not all of it, not for everyone, but more of it than the current tools assume. We've spent the last few months building things specifically to address the first three barriers, and thinking hard about how to make automation safe enough that the fourth becomes less restrictive over time. We have opinions about what works. But we'd rather show you than tell you.
So here's what we're doing instead.
On Saturday, May 16, we're hosting a free hands-on lab in San Jose. We'll walk through how to spot what's automatable in your own work, where to start when you have a hundred options, and how to set up your first AI teammate in under 15 minutes. No coding.
Cap is 30 people. RSVP: https://luma.com/yqpulmgq
On Tuesday, May 20, we're running a similar event in San Francisco. Details and RSVP coming soon. Follow @vm0_ai to be the first to know.
If you can't make either event, follow us anyway. We'll write up what people actually build at the lab. The workflows, the surprises, the things that didn't work. We'll publish them in the next post. So you'll get the practical answers either way.
The four barriers above are real. We hear about them from almost every person we talk to. The first three are solvable today. The fourth takes longer, but it's solvable too. We're going to spend the next few weeks proving it, at the events, and in what we publish next.
Quotes are anonymized to role. Interview sample: 22 people across software/AI, healthcare, performing arts, manufacturing, consulting, retail, biomechanics, and nonprofits. Roughly 73% with a technical background and 27% without.
