Is AI Disrupting Product Development?

We are regularly told that AI has caused a deep shift in product development. The old patterns are no longer relevant, and you have to either adapt or die.

In this article I’ll try to assess if that’s indeed the case. Have we fundamentally changed the way we develop and sell tech products? Is this finally “the year of AI” (as we were told last year and the year before it), or are we just in the land of “inflated expectations” per the Gartner Hype Curve. 

Let’s go over some of the main claims. 

The Cost Of Software Development Is Dropping To Zero

If you’re a non-coder, a solo entrepreneur, or an early stage founder, software development has absolutely become much cheaper for you. While zero is somewhat hyperbolic we’re talking about a massive cut in costs.

But these are relatively niche cases. Most real-world, production software is far more demanding: 

• Many requirements — real-world software has to be functional, usable, efficient, robust, secure, scalable, and maintainable (partial list). 
• Hard constraints — software design has to balance requirements, tech capabilities, dependencies, existing codebase, timelines, costs, and backwards compatibility, among other factors. 
• Maintenance — The software has to be tested, maintained and updated for its lifetime. It’s not a one-time effort.
• Org complexities — Software development also includes project planning and management, resource allocation, dependency resolution, dev rituals, politics… 
• Not just coding — Software development is just one part of the larger product development cycle that includes goal-setting, research, product discovery, product delivery, go-to-market, and post-launch activities (partial list).

Today’s AI coding agents are quite capable, but they’re solving just part of the problem. They’re not yet smart, reliable, or human-like enough to do all, or even the majority of the work. So these expensive human developers are still very much necessary, and while their tasks may be somewhat different, they are still the true agents and AI is just a tool. This is why even AI-native companies are hiring engineers. 

It’s possible that in the future our model of software development will fundamentally change and humans will be taken out of the loop. Cursor is conducting interesting experiments on fully autonomous AI software development. The results are promising, but they show it takes dozens of agents running for weeks and consuming a huge amount of tokens to build something like a web browser. I don’t see companies willing to switch this all-AI model any time soon. 

We Can Build Software 10x-100x Faster 

Many feel that AI is making them more productive, and some tasks do take a tenth of the time used to take. But there’s a tradeoff between time and the quality of the output. What we save in production we may shift to data prep, context sharing, prompting, output reviews, and iteration. Software development is getting faster, but I’ve yet to see any serious research to support the claim of 10x, or even 1.5x acceleration in real-world production software. 

More importantly, the 10x-faster-shipping claim, even if it was true, is missing a key point: the metric you should optimize for is not time-to-bits-in-production but time-to-outcome. The goal is not to ship stuff as fast as possible, but to ship things that create value for users and the business. Research shows that most ideas miss the mark, so just accelerating delivery without accelerating research and discovery will just end up creating more waste, quicker. AI can definitely help with research and discovery (explained below), but currently all the focus and the hype seems to be on faster output. 

The Top 1% Are Using AI “Right” and Gaining Massive Advantages

We are told that some companies and individuals are already “living in the future” and are far less productive and successful because of how they use AI. The examples cited are usually prominent AI companies such as Anthropic or Cursor and certain high-profile individuals. 

As with any new technology, there’s a disparity between early-adopters and the mainstream. The early adopters are indeed doing impressive things with AI, and are using it to learn as well, which may create very fast positive growth loops. Still I suggest taking the “top 1%” claims with a grain of salt:

• Atypical cases — The AI-native startups we all hear about are at the eye of the AI storm. They’re facing massive demand and are developing AI-centric products. It’s not clear that what works for them will work for regular companies. Early-adopter influencers usually live off sharing content and selling courses, so they too are not your typical developer or product manager. 
• Top 1% in what? — You can be top 1% in AI knowledge, AI-workflows, and AI token usage and still underperform compared to someone who understands the role, the context, and the environment better, even if they’re just a casual user of AI. AI is just a tool — a means to an end. Being a heavy user doesn’t automatically make you good at your job.
• No one right way — It’s my impression that the early adopters are not all using AI the same way. Everyone is still exploring and experimenting, and there are few clear best practices yet.

The “top 1%” and similar messages are often meant to induce FOMO and to drive you into action: read my post, sign-up to my newsletter, buy my course. There are genuinely thoughtful people out there that are exploring ways of using AI and sharing their findings, but in my experience these positive actors don’t pretend to know it all and don’t use fear tactics. 

Companies and Consumers Will Develop Their Own Software

“I coded a clone of [known product] in 2 days, so I no longer need to pay for it” is a story you hear regularly on social media. The assumption is that because development is now “10-100x faster” and “costs near zero to develop”, people and companies will start making their own software instead of buying it — a major disruption to the industry. 

In certain cases this is already happening. Company employees, including ones with no technical knowledge, are creating their own utilities for internal use. Some consumers are indeed vibe-coding casual apps.

However, that does not mean the business of software is done. Customers are not just paying for the bits. They’re paying for someone to take care of their needs so they won’t have to think about it. If every enterprise started developing their own Salesforce, Google Docs, or Slack they’ll have less time and resources to focus on their core business and the cost savings will quickly dissipate. Consumers are similarly busy with their lives, and very few will even bother to try to create their own software, as was the case with no-code apps, VB-script, 3D printing, and other do-it-yourself methods of product development. 

I feel that the more realistic expectation is for AI to lower the barrier of entry, which may increase competition and drive prices down. Indeed 2025 saw a surge in the number of apps submitted to app stores (but of course there’s more to business success than just shipping). A report by PricingSaaS found that SaaS prices grew by 14.7% in 2025, partly because of bundling of AI and usage-based pricing. So it looks like AI hasn’t killed the software market quite yet, in fact it seems to be invigorating it.

AI Is Taking Over Tech Jobs

Since the introduction of ChatGPT people have been talking about a cooling in the job market. There are occasional mass layoffs and hiring freezes, experienced workers take longer to land a role, and junior roles seem to be hit the hardest. 

However the cause is not necessarily AI — this analysis suggests that the slowdown started before the launch of ChatGPT and is better explained by economical reasons: rise in interest rates, post-covid effects, and a generally unstable economy. Other research found that so far (as of the publishing of the report in October 2025) there’s no discernable change both in employment and unemployment rates in jobs exposed to AI automation.

Even so, I believe AI is likely to impact the tech job market:

• There’s plenty of anecdotal evidence that jobs in copyrighting, design, editing, support, and other professions are being automated. 
• AI companies are spending much money and effort training their models to become cheaper replacements for some professionals: from divorce lawyer, to interior decorator. I suspect that product owner roles that are mostly about producing deliverables: PRDs, backlogs, user stories, are susceptible to automation as well. 

• The AI productivity hype may cause executives to freeze hiring, as has already happened in some companies. AI is also a comfortable excuse for layoffs. 
• The AI bubble might burst, causing massive job loss.

AI Is Redefining the Role of Product Managers

The claim here is that the PM role is rapidly changing due to AI, and that hiring companies are now preferring PMs with deep AI knowledge. Those who don’t have it are at a major disadvantage. 

A casual sample of open PM positions on LinkedIn shows that currently AI is not often a major requirement. Still, there’s no denying that as a product person, you should learn to use AI.

The question is “use how”? Several, somewhat overlapping answers are being proposed:

• AI-PM — this is a PM who has a deep understanding of the inner workings of AI, and has mastered prompt engineering, fine tuning, GAR, evals, Skills, and other topics. This is the sort of PM companies who develop AI-powered products are hiring, reportedly landing bigger compensation. 
• Builder PM — This PM vibe-codes prototypes, designs UI or even helps in product development. 
• Dev-like PM — This PM uses dev tools like Claude Code and Cursor to become a better and more productive product manager. 

I see value in all of these, although less so in the Builder PM — PMs already have enough on their plate; making them coders adds very little. Still I feel that we may be again conflating the technology with the needs and goals. As I wrote in a previous article, PMs are there to ensure the company is creating high-impact products. This goal breaks into many activities (far beyond writing PRDs and stories). AI can help with many of those.

Final Thoughts

Will AI disrupt the tech industry? Possibly, but so far it hadn’t changed any of the basics: 

• We still need to build things that are valuable, feasible, usable, and business-viable (Kagan’s four risks)
• Customer needs, expectations and behaviors haven’t materially changed 
• No new business or economical models were introduced (perhaps with the exception of usage-base pricing / token passthrough) 
• Company org structures and processes are mostly the same 

In other words, we’re mostly doing the same things, just with new tools (that are far from perfect). But that could be just a delayed reaction of the market. When new technologies emerge, we tend to overestimate the short-term impact and underplay the long-term change (Amara’s law). This happened with the Web and with mobile, and I’m assuming we’re seeing the same with AI. So I advise staying attuned to what’s happening and experimenting with AI, but also taking big pronouncements of transformation or disruption with a large grain of salt. At the end of the day AI is just a tool and we are the intelligent actors.