Table of Contents
Do not index
Do not index
Hide CTA
Hide CTA
Hide cover
Hide cover
It’s almost hard to imagine, today, the struggles of file formats from the ’90’s and early aughts. Microsoft was the dominant player in productivity software, locking a generation of businesses into using Office for everything. Good luck creating a new word processor or spreadsheet app if it couldn’t open and save Microsoft Office files.
Today’s business software tends to be better at data portability. Thanks partly to API-centric designs, partly to data portability regulations, most software used at work today offers CSV or JSON data exports.
If you decide, tomorrow, to move your email newsletter to a new service, it’s possible to take your subscriber data with you. Importing it into your new service might not be so easy, though.
You’ll have to clean up the data and manually align old headers to new ones before importing. One app’s exports store names as a single column where another expects separate first and last names. One writes dates in human-readable formats, while another uses UNIX timestamps. And that’s before you try to normalize tags between the two apps or handle file attachments.
You’re not locked into the original app, per se, but it’s not easy to switch, either.
That was, until AI came along. Hand ChatGPT an export file from one app, and seconds later you’ll have an import file ready for another app.
Switching costs shrunk to zero overnight.
Lock-in by tedium is dead
File formats were never a perfect lock-in. At worst, you could open two apps side-by-side, copy data from one, and paste it into the other. Tedious, but possible.
Incumbents were incentivized to make switching costs as high as possible. Why simplify switching to the competition? Thus not only program-specific .doc and .psd files for Word and Photoshop, respectively, but version-specific files that weren’t fully compatible even with older versions of the same software. Open a .doc file from Word today in Word ’97 or a modern .psd file in Photoshop CS1, and odds are something will break.
We weren’t just locked into a single application—we were locked into upgrading as well.
AI imports change the equation for startups
Startups enter the playing field with the opposite motivation. They want to break in, want people to switch in and start using their product as soon as possible. They’re incentivized to build around more open standards—and to invest in building import tools so people can jump ship from competitors and try their product as easily as possible.
But building and maintaining these migration tools is consumes time, the most precious resource for any startup.
Luckily, generative AI showed up and broke the equation on both sides. Hand it an export file, ask it to reformat it for a new app, and moments later you’ll have turned migration costs into a rounding error. AI to SVG, JSON to CSV, all convertible and cleaned up in a few seconds of compute time. Can’t export the data? Your computer’s AI-powered OCR can extract the text, or a GPT 4 crawler could read the page and extract data—something we’re experimenting with as a universal way to import forms into Fillout.
We’ve seen the benefits first-hand with Fillout’s AI form importer.
It’s partly doing something users could have done manually, all along. Tedious, sure, but you could open a Constant Contact exported spreadsheet, and change its Email field to an Email Address field before importing your subscribers into MailChimp. You could split a Name field by space to get separate first and last name columns in any spreadsheet, or normalize tags by hand. You could, if you really wanted to, convert a JSON file to a CSV with a bunch of find-and-replace queries.
Behind the scenes, GPT’s running code to convert the file—something a user could have done, perhaps, with a bit of terminal skills and code literacy. Most people wouldn’t go to the trouble, though. It’s easier to just keep paying and putting up with an app that’s good enough. The switching costs were just too high.
Even templates get better and more customizable with AI. We built templates for Fillout (such as this RSVP form template), with a new twist: We baked in AI customizability. Describe how you want to use this form and what data it’ll gather, and it’ll tweak the template for your needs. The new form is derived from a template, yet personalized and ready to use, in less time than it takes to even select a traditional template.
Suddenly switching costs aren’t just lower than before. They’ve disappeared entirely. No need to trial software with dummy data; you can jump right in, ready to work in a new tool, with AI making it perhaps better than your older app was before the switch.
Building a long-term, AI-proof moat
Which leaves the million dollar question for developers. Generative AI can code basic software, and make it easier for people to jump from one app to another. What’s left for developers to do?
Building better software. Everyone’s building a database today, seemingly, to try to amass as much data in one place as possible to power internal AIs. And yet, it’s hard to imagine AIs not making it increasingly easy to migrate away from those apps, too. The only sustainable moat in software today is doing a better job at serving customers and solving their problems.
GPT can iterate on existing ideas. It’s incredibly good at writing text and code that’s a remixed version of what came before. But it doesn’t know your customers, can’t sit down with them and observe them at work, can’t turn those observations into novel insights.
What AI can do is remove busywork. Imagine moving from Gmail to Outlook, or HubSpot to Salesforce, with a prompt. Imagine the new app’s AI spinning up templates for your business and cleaning up your data, making the switch feel like a fresh install. You could afford to move every time one of them released a new feature! Which, inevitably, means that investments in UX and new features driven by R&D will become more important than dumping money into sales and marketing in the hopes that people are locked in after you nab them.
Locking people in only makes them resent your business—and, increasingly, it won’t even work as AI lowers the switching costs. The way to move forward is to build more human-focused software, outsourcing the drudgery of imports/exports and more to AI, freeing your dev team up to build better software that people love.
That’s how you win against the bots.
Image Credit: Moat photo by Kees Streefkerk via Unsplash
Written by
Dominic is a co-founder at Fillout. He previously worked in engineering & product at Retool. Prior to Retool, he started Cheer (backed by Sequoia and acquired by Retool in 2020).
