At first glance, it looks like a well-oiled machine. In reality, it’s the opposite. Workers have the latest equipment, but ironically, they still spend most of their hours on repetitive, manual tasks that should be automated.

As production slows to a crawl, burned-out employees make more mistakes. Products roll off the assembly line late, often with defects needing repair. Although staff members work evenings and weekends to catch up, they keep falling behind.

I’ve just described the typical software development team.

What needs to change

Software underpins so much of modern industry, and our lives in general. The software development lifecycle (designing, building, and deploying code) is often thought to be orderly and streamlined, but it’s actually chaotic and inefficient.

As the founder of three companies that serve software developers, I’ve seen businesses large and small struggle with this train wreck. A big part of the problem is that while writing code is supposed to be their main job, developers spend up to 70% of their time doing all of the things that must take place after—testing, deploying, fixing bugs, managing changes, and so on.

AI has the potential to dramatically reduce manual toil in many areas. But alone, it’s not enough.

That situation has dire consequences for any company with a software team. Businesses are wasting time and money, but they also risk losing talented people who will leave to work elsewhere.

It doesn’t have to be this way.

A recipe for developer burnout

As software developers will tell you, the odds are stacked against them.

My company recently surveyed 500 engineers about their experience at work, and the responses were a cry for help.

A quarter of developers work at least 10 days of overtime a month, and more than half cited burnout as a reason for peers quitting. Behind these numbers lies a disorganized, inefficient process—one that remains alarmingly manual and error-prone despite the high-tech backdrop and high-stakes product.

As consumers, we might expect that code updates are rolled out continuously—new features and fixes implemented as soon as they’re available. But six out of 10 organizations are still releasing code updates monthly or quarterly, well below the ideal daily pace for an efficient development process. That can mean holdups in delivering new features, dealing with critical issues, and responding to customer feedback.

Even getting internal sign-off is problematic. Code reviews—the peer review process that helps developers improve quality before shipping—are a cornerstone of software development. Yet almost 70% of developers wait more than a week to finish them. This translates to costly delays and missed opportunities to identify and fix problems.

This shaky development process leaves engineers lacking confidence in their work. Roughly four out of 10 developers surveyed said they can’t release code to production without risking failures, while a similar number experience failures at least half the time.

On top of that, developers are still slogging away on low-value tasks. For example, when a software deployment fails, almost 70% roll it back manually. That means spending hours returning the code to its previous version.

Where is AI?

AI is billed as a solution to developer toil and has the potential to positively transform how developers work. By itself, though, it doesn’t necessarily help.

For example, developers can have AI write small chunks of code using copilots, but the technology is still prone to mistakes. So time saved from using AI must go into quality assurance—having a human check the code.

No wonder developers are conflicted about AI. In one recent survey, 70% said they believe it will help shrink their workload. But almost 30% said they don’t trust AI, while one quarter called it unreliable or inefficient.

How to build a developer-friendly culture

So, how can software teams get out of this mess?

Part of the solution is changing the culture around development. Unsurprisingly, businesses can look to early-stage Silicon Valley startups as a model. At those small firms, fewer processes and bottlenecks stand in developers’ way. Approvals are streamlined, and developers are able to draw a direct line from their work to the customer’s experience.

Bigger companies can follow that example by cutting down on interruptions and distractions that overload engineers and make it tough for them to achieve a state of flow, whether that’s needless approvals or excessive meetings. It’s also crucial to seek and incorporate quick feedback from end users, which developers crave so they can speed up and improve software updates.

Competitive pay is also table stakes for sustaining a strong dev culture. But above all, developers want a compelling, clear company mission, challenging and stimulating work, and colleagues whose skills they respect.

Rethink technology

Better software tools can drive much-needed change—if they’re deployed the right way.

Developers are already awash in tools, which only adds to the burden of toil. On average they manage 14, many from separate vendors. More than half say it takes longer than a week to learn new DevOps tools.

With all those siloed tools in play, it would take 100 days to onboard a new recruit. And the pain doesn’t end there. Because nearly all developers juggle multiple tools with different interfaces, workflows, and licensing, the result can be confusion, cognitive overload, and lack of consistency in the development process.

The answer here is also hardly a surprise: one stop where the latest tools live and all the components just work with one another.

But this productivity suite concept, so familiar to most consumers, remains surprisingly elusive in software development. Most developers still don’t have a central hub where they can easily access what they need without navigating multiple platforms.

AI’s role

Meanwhile, AI does have a central role to play, but the key is to treat it like yet another tool—one that augments human intelligence rather than replaces it.

Its utility extends well beyond coding; quality assurance is where AI may be most valuable. When a software deployment fails, AI can quickly scan millions of lines of code to help engineers determine what went wrong, then explain the problem and a fix in plain language.

It can use its encyclopedic knowledge of security flaws to spot and repair vulnerabilities. And developers will no longer need to spend hours testing software updates, a task AI can crush in minutes by pinpointing the right tests.

The payoff of integrating the right platform is well worth it. We estimate that for every 1,000 developers, consolidating tools would improve productivity by 53%, saving more than a million hours a year.

But even the right tools aren’t enough without something equally important: buy-in from developers on the ground. When introducing tools that can boost efficiency, I’ve found it’s crucial not to make sweeping changes right away. Developers are deeply attached to their toolkit, and they need to see proof before making a switch. For a company with thousands of developers, start with a few hundred. Then build on their success with new tools and processes. This is the best way to usher in a broader culture shift.

Ultimately, it’s in companies’ best interests to ensure that their software teams are happy and productive. By taking steps to make the development process more efficient, they can solve the toil problem and save time and money. That frees up developers to focus on what they do best—creating products that make life easier for customers, too.

Jyoti Bansal is CEO and founder of Harness.