August 10, 2023
The proliferation of high-quality low code tools in recent years has reshaped the first phase of startup launch. The cost of entering the market is dropping as low code tools and AI make product rollout easier - note that this doesn't make success easier, but it does make the first rung of the product ladder easier to access. This also raises an important question - should founders use low code solutions? Or should they try to use AI-based code generation to accelerate development? If low code is the right path, when should an upgrade be considered?
Simple question: Should you use low code or AI-assisted dev to build your MVP?Simpler Answer: Yes.
But which route should you take?
Can you build a product MVP using a lego-style approach, where you snap pieces together rather than writing code from scratch? No code solutions make this promise - that you can drag-and-drop your way to a working product MVP. There are huge number of vendors in the space, with no-code requiring zero knowledge of software development and low-code attempting to simplify development while preserving more flexibility.
https://www.g2.com/categories/low-code-development-platforms/small-business#grid
In my personal experience, the highest productivity no-code platforms can boost productivity 8x versus writing code, enabling founders to get an MVP to market fast and test their idea. It's not pure magic - you'll have to make compromises on UI and capabilities with the quicker no-code tools. More flexible low code tools will enable you to build more custom UIs, but the productivity boost achieved is much lower, perhaps 2-3x traditional approaches.
AI-assisted software development is really just software development! GPT4 will spit out a good starting point for many different use cases, but that's not so different than starting with a good template. GitHub Copilot and its competitors do boost productivity, but the boost is more in the 50-100% range per Microsoft's studies. AI is definitely improving the developer experience, but doesn't provide enough of a boost to compete with no-code when it comes to shipping a quick MVP.
When I started Fraction, I immediately had a choice to make: how were we going to build our developer portal? As a technical founder I could estimate the effort required to write a Vue / nodejs / Postgres app (the most familiar stack to me at the time). I estimated 8 weeks for a proper MVP. We needed to move faster! So I started looking at no-code solutions, and spent a day experimenting with a few. I landed on Softr + Airtable as an easy to use combo, and it took about a week of full-time effort (spread out over a month given my actual availability) to get a usable portal out the door. Since I made that effort, many low code tools have added AI-driven experiences to further accelerate development.
Fast-forward to the present: we shipped that portal, and in the year that passed we built a viable and growing business. Now it's time to upgrade, so that we can control the user experience and add capabilities. We are working with one of our own fractional developers to build a new portal, this time using Nextjs + Vercel. Eight weeks appears to have been a reasonable estimate of total time needed! Adding all the capabilities that we now want will take multiple phases thereafter. But we can make that investment with the confidence that we're building it to grow a viable business, not making a risky bet with zero clients. When you consider moving past low-code, take a stepwise approach. We are finding that it's possible to leave the data infrastructure in place while replacing the front end, to avoid disrupting our existing business.
As we speak with prospective clients, and as I talk to early stage founders, I'm encouraging them to leverage no-code or low-code for their MVP. This enables rapid iteration with prospects and makes it possible to test out your business ideas without committing a ton of capital. If you want our help, we can build a no-code MVP in a month instead of a traditional one in several times that.
For traditional "real" code engagements, we're finding success when clients have a running business, and therefore have a solid understanding of their own needs. When I built my last startup, no-code didn't exist, so I had to spend over a year in the lab! But today it makes sense to spend that year after some initial traction, and not before.
P.S. Some technical founders will read this and cry foul - "I'm building technical innovation, how could I possibly do that in no-code?"
To which I answer, for most startups the technical innovation isn't in the usual boilerplate pieces of the app - it's often in data gathering and analytics, or integrations, or a specific component of the user experience. You can segment your app so that you focus on the core IP that you have, while offloading to low-code the work on random admin screens etc, which I guarantee you aren't making you any money!
