Are AI Coding Tools Devaluing Software Products?
#development#generativeai#codeassist#genai
Jason St-Cyr

Jason St-Cyr @jasonstcyr

About: Dad, Writer, DevRel guy at work. Opinions expressed are my own, of course.

Joined:
Jun 20, 2019

Are AI Coding Tools Devaluing Software Products?

Publish Date: Jul 8
0 0

A recent LinkedIn post triggered me to thinking again about the devaluing of software engineering:

A screenshot of a LinkedIn post by Josh Sorenson. The text reads: My new toxic trait? “Oh I could build that for free with Cursor and n8n.” Yes. Technically true. But now I’ve fired the $29/mo SaaS tool and hired myself as the dev, support team, product manager, QA, and customer success rep… all for zero extra hours in the week. Genius, truly. It’s giving woodworker energy: Sees a $150 table at Target and says, “I could make that for like… $40.” Six months later they’re still watching joinery tutorials..... I launched one iPhone app and now every app in the App Store feels like a personal dare. “Oh I could build that in a weekend with Supabase.” No, Josh. No you couldn’t. But hey, if you’re gonna overbuild, at least make it fun. Better to be creating than consuming. Just maybe… schedule some maintenance time too. 😅

People severely underestimate the last 20% of getting something to being launched, and also generally forget about all the work it takes after you think you are finished building a thing. Not to mention getting people to use your thing, if you are trying to make a few dollars off of it to pay for that time you are putting in.

I worry sometimes that the AI democratization of building is leading to an undervaluing of the very real cost of making real software products that real customers can use. The prototype that you get from these tools feels like you have it almost built but it is really just the tip of the iceberg.

Everyone Is a Builder

The good side of this ease of building is that these new tools allow you to go and build something for yourself (if you want to) for fun. You won't be saving money if you are trying to make it solid and great for others, but you might save yourself some dollars to get yourself a thing that is "good enough" for your usage. As the original LinkedIn poster, Josh Sorenson stated, "Better to be creating than consuming". In a lot of senses, building software is just like a lot of maker hobbies. Along the way, you might even learn something! (Assuming you don't just let AI build everything and you are actually looking at what it is doing and absorbing the why of it's choices.)

Side note on learning: Harvard University's Jal Mehta was quoted as saying "...if you aren't the one piloting the vehicle, the AI is the one learning, and you are just sitting in the passenger's seat." Please make sure to be piloting!

It can be incredibly rewarding, in and of itself, to build something new. The tools out today allow most people to be builders, which is a great democratization of the hobby. Build something cool! It's fun! I promise you!

The old role of development is shifting. Patrick Dubois recently posted about four patterns of AI-native development where the focus of technical practitioner time is moving to a more abstract level with the usage of AI assistants.

Diagram showing the transition of development time towards 4 other focuses: Product, Operations, QA & Architects, and Data. The arrows in the diagram imply that developers will need to spend more time in different areas, depending on a different abstraction capability typically done by other teams or roles.

  1. Pattern #1 -- Producer to Manager: Human in the loop that accepts, rejects, refines the work of the agents. You are managing the AI developer.
  2. Pattern #2 -- Implementation to Intent: Focus less on how to accomplish the goal, and more on what we are building. Requirements are key.
  3. Pattern #3 -- Delivery to Discovery: Experimentation cost is lower so we can get rich feedback better on potential opportunities and alternatives.
  4. Pattern #4 -- Content to Knowledge: Stop trying to document every individual's implicit knowledge, but that never gets kept up to date. Using AI drives a bigger need to have the knowledge, and AI can generate that knowledge for you as you go. It knows why it built a thing, and what it does.

This ultimately feels like a lot of 'specialized' roles that we've held in the past are going to start blurring together. The tools are taking away the syntax/specialization knowledge and replacing it with higher-level problem solving needs. This opens up a lot of folks to be able to contribute in different ways to the delivery of technical solutions.

But if everybody thinks they can build software, how will this impact product companies?

Are Software Products a Commodity Now?

Right now, the tech industry is entirely reliant on selling software to people who understand that it would cost them more to build than to buy. That understanding of their cost to build is incredibly important! This might be a cost in salaries, risk, infrastructure, operational costs... whatever an organization believes would be their cost to doing it themselves goes into the decision on a buy vs build.

What happens when everyone thinks they can build it themselves with AI tools cheaply (for now) and over the weekend? Regardless of the reality of the situation, regardless of the reality of the cost to the system, the fact of the matter will be that the perception by those making the decisions will be that this cost is low.

Why pay half-a-million dollars a year for a SaaS content product that you could build in a weekend with only the features you need?

It doesn't matter that:

  • That's not how long it will take to get launched in your organization
  • It isn't just a one-off build, you actually will have a cost to maintain it long term
  • You won't be able to get all the features and will need to invest to keep having features over time as the industry changes
  • You assume all the risk for all aspects of the software, including the legal ones

If you don't value the process of engineering software anymore, you won't be willing to pay as much for somebody to build software. We've seen this historically with other trades. Remember people who got paid just to type things up onto paper? Now it's so easy for anybody to learn and do that we just don't pay people to be skilled at that anymore.

I believe we are going to see a massive drive-down of the willingness to pay for software features or services from companies that offer engineering services. These vendors are already racing to massively increase velocity or decrease costs with AI themselves to try to get ahead of this. Based on last August's GitHub report, over 97% of devs were using AI tools in some manner, even if their organizations hadn't put out policies on usage yet, and that around 30-40% of organizations are actively supporting the usage of the tools, and this is only growing. This differs from the Stack Overflow developer survey which had the numbers a little lower for developers currently using AI tools in 2024 (62%) but still with definite signs of growth of usage in the report.

Regardless of the numbers, the perception of value is going to shift and software vendors that aren't ready to shift with that are going to be doing massive layoffs or closing altogether as they can't make enough revenue to cover the costs of the business.

So what are product vendors going to offer instead? When engineering is no longer valued by the customer, what do we pivot to?

Delivering Humanity as a Feature

I loved an article I read about the "Who cares?" era. The defining piece that stands out is the human behind the piece. Nobody will value the code you wrote or the feature there, but if you can make somebody care about something? That is going to have an increasing value. Humans are going to become starved of actual real creations and things that were cared about, so if we can find a way to bring that into the software we deliver, then could become the real product. Who connects with you best?

As somebody who works in developer communities and developer relations, I can't think of a better time to be part of the humanity of a software company. This is the time when real people are going to be at their most sought after by customers. If you can build something where it seems like you actually care about your customers and want to help them succeed or help their community thrive, that is something people will value. That is something a person could sign a check for, which appeases the almighty revenue generation targets!

Most importantly though: that is something I believe is worth doing for people.

However, I suspect we are in a race condition on this track. Choosing to go a route like this will only happen if the devaluing stage is achieved fully. Right now, this 'devaluing' is predicated on cheap access to AI tooling. A lot of these AI companies are operating at a loss and are charging far less than they need to in order to grow their customer base. It's a common pattern of penetration pricing, as seen with Netflix, and Uber, and Mailchimp (among others):

  1. Make a solution that is easier. With AI code assistants, the human is in the loop to validate with their expertise, but more humans are now able to be in the loop.
  2. Go from free or freemium to a cheap option at something that makes it a no-brainer to choose to buy vs build.
  3. Remove free options and increase prices to the point where you are profiting as a business.
  4. Increased costs to the consumer cause the buy vs. build discussion again.

With AI tools, we are already on the second phase. I think we could see a massive impact to the industry when the third phase rolls around. Right now we are seeing a lot of folks entirely dependent on AI to do their work. I no longer publish without having an AI look over my writing for issues, and I use it daily for building code. In fact, I would say I rarely go to the keyboard first to type code, I am always starting with the AI prompt, even for simple tasks. When the prices go up, will I abandon my crutch? How much can they pull on my monthly subscription before I say the productivity value isn't there?

So is AI devaluing software engineering?

Back to the original question I had, before I started my rant... yes, I believe AI code generation and application building tools have made the average individual see less value in off-the-shelf software. I believe the average person has no idea how much effort it is to build real software, launch it, and operate it, but that doesn't matter. The perception of value is all that matters when it comes to whether a business can run. If people won't pay for it, it doesn't matter what it cost you to build it.

And this genie is not going back in its bottle.

References:

Comments 0 total

    Add comment