Why Your Users Reject Automation: The Lesson Big Tech Is Ignoring

Simon Willison, a software engineer known for his work with Django and, more recently, AI tools, published a thought that should be required reading for any PM working with automation: “People do not yearn for automation.”

His thesis is simple and uncomfortable. The tech industry operates on the premise that users want to do less. That any automated task is an improved task. That removing steps from a process is always a gain.

The adoption data tells a different story.

The problem isn’t the technology

Think about how many automation features you’ve seen launch with great fanfare and, months later, get abandoned or turned off by default.

Gmail has had auto-reply suggestions for years. Most users ignore them. Google Photos offers automatic album and collage creation. How many people do you know who use that regularly? AI writing assistants show up in almost every text editor now. And often, the first thing power users do is figure out how to disable them.

It’s not that the technology doesn’t work. It works. The problem is something else.

Control is perceived value

There’s a fundamental difference between “automation that helps me” and “automation that decides for me.”

When you type in a search field and the system suggests how to complete your phrase, that’s help. You stay in command. The suggestion appears, you accept or ignore it, the decision is yours.

When the system rewrites your entire email without asking, or reorganizes your photos into categories you didn’t choose, or publishes content on your behalf based on “detected patterns,” the feeling shifts. You’re not being helped. You’re being replaced.

Automation as assistance

• Suggests, doesn't impose
• User confirms the action
• Easy to undo or ignore
• Amplifies capability without removing control

Automation as replacement

• Decides without asking
• Action already happened when user sees it
• Undoing requires effort
• Removes steps that gave a sense of mastery

The distinction seems subtle, but it directly affects adoption and satisfaction metrics. Products that automate too much too soon often see drops in engagement, even when the feature “works perfectly” from a technical standpoint.

The mistake of assuming fewer clicks is always better

In my view, much of this problem comes from a misinterpreted metric.

In UX, there’s a legitimate principle that reducing friction improves conversion. If a user needs 7 clicks to complete a purchase and you reduce it to 3, you’ll probably see improvement. That’s well documented.

But “reduce friction” got translated, at some point, into “eliminate any effort.” And those two things aren’t the same.

Friction is what gets in the way without adding value. Effort can be part of the value.

Think of an image editor. Part of the satisfaction comes from making adjustments, seeing the result, refining. If a button for “improve everything automatically” solves it in one click, the technical result might be good. But the user learned nothing, didn’t feel like they created something, has no story to tell.

For professional or creative tools, this matters. For products where users build identity through use, this matters a lot.

The specific case of digital marketing

For anyone working in digital marketing, this point has a direct application.

It’s common in the industry to get tempted into automating more and more of the operation: auto posts, auto replies, automatic segmentation, AI-driven campaign optimization with no human intervention.

The promise is clear. Less operational work, more scale, lower cost.

The risk, less discussed, also exists. When you automate interaction with your audience, you lose the ability to pick up on nuance. When AI writes your posts, your voice gets diluted. When the algorithm decides what to show to whom, you stop understanding your own audience.

Structurally, it makes more sense to automate repetitive, low-cognitive tasks (scheduling, reporting, alerts) and keep humans in the decisions that define positioning, tone, and strategy.

How to evaluate if your automation will work

Before implementing an automation feature, or before adopting a tool that promises to “do everything by itself,” it’s worth running through some questions.

• Did the user ask for this, or am I assuming they want it?
• Does the automation suggest or decide?
• Is it easy to undo or adjust the result?
• Does the user still feel they’re in control of the process?
• Am I removing friction or removing engagement?
• Does this feature work equally well for new and experienced users?

If most answers lean toward “decides on its own, hard to undo, user becomes a bystander,” there’s a real risk of rejection. Not because the technology is bad, but because the feature design ignores how people relate to tools.

The blind spot in the AI rush

The current race to add AI to every product has a structural blind spot. Product teams are pushed to show “innovation,” and AI became synonymous with innovation. So features get launched not because users asked for them, but because the market expects them.

This creates a dangerous cycle. Automation features launch, adoption is low, the conclusion is “we need to improve the AI,” more investment goes into the technology, the feature keeps missing the real problem.

The real problem, often, is that the user didn’t want to be replaced. They wanted to be empowered.

Implication for sellers

For anyone on the sales or marketing side of a product with automation features, this has a direct implication for messaging.

Selling automation as “you don’t have to do anything” can sound like a benefit to whoever’s writing the copy. To the user, it can sound like “you’re about to lose control.”

My read is that it works better to position automation as a superpower, not as a replacement. “Do in 5 minutes what used to take 2 hours” is different from “the AI does it for you.” The first keeps the user as the protagonist. The second turns them into a spectator.

The balance that works

None of this means automation is bad or that AI shouldn’t be used. It means the way you implement it matters as much as the technical capability.

Products getting this right tend to follow some patterns:

Automation is opt-in, not opt-out. Users choose to activate it, not discover how to turn it off.

Transparency about what was automated. If the AI did something, the user knows what and can adjust.

Granular controls. It’s not “AI on or off,” it’s levels of assistance that the user calibrates.

Focus on tasks nobody likes doing. Automation for organizing files, categorizing emails, generating reports tends to be better received than automation for creating content or making strategic decisions.

Wrapping up

Big Tech is learning, at great cost, that execution speed isn’t the same as perceived value. That doing something for a user isn’t the same as doing something to a user.

For PMs, marketers, and product owners, the lesson is clear. Before you automate, ask whether the user actually wants to be taken out of that loop. Often, the answer is no.