Here’s How I Plan on Protecting My SaaS From AI

AI is making software easier to build than ever. But building software and operating a dependable software business are two very different things.

I recently listened to an episode of Startups for the Rest of Us, where Rob Walling and Einar Vollset walked through 10 ways AI is attacking SaaS companies, along with eight competitive moats that might help them survive.

As someone building a bootstrapped SaaS company, I found the conversation particularly interesting.

AI is making it ridiculously easy to build software. Features that previously took weeks to develop can now be built in days or even hours. Customers can vibe-code their own internal tools. Competitors can replicate features much faster. And some software products might eventually become completely unnecessary.

So naturally, I started thinking about my own business.

How vulnerable is Fillvisa to AI? And more importantly, how do I plan to defend it?

Here are my thoughts.

First, a little about what I'm building

I'm the founder of Fillvisa, a software platform that helps individuals and immigration professionals prepare U.S. immigration forms.

The business has three products:

1. Fillvisa Free

A completely free, browser-based immigration form filler.

Users can prepare immigration forms such as I-130, I-485, I-765, and DS-160 without purchasing expensive PDF software or signing up for an account.

Everything runs inside the browser, with data stored locally on the user's device. No accounts, no cloud database, and no AI.

2. Fillvisa OS

The open-source version of Fillvisa.

Anyone can download, self-host, modify, or build on it under its open-source license. It's intended for individuals, developers, nonprofits, and civic technologists who want to make immigration software more accessible.

3. Fillvisa Plus

This is the commercial SaaS product.

Plus is a cloud-hosted workspace for immigration professionals, including document preparers, consultants, and small immigration law firms.

Beyond filling forms, it provides client intake, reusable immigration profiles, document management, team collaboration, access controls, e-signatures, and other workflows.

Our solo plan starts at $40/month, and we also offer white-label team workspaces.

The business model is straightforward: Give the form-filling software away for free, and charge professionals who need a managed, collaborative workspace.

This distinction becomes particularly important when thinking about AI.

The 10 attacks: How vulnerable is Fillvisa?

In the podcast, Rob and Einar discuss ten ways AI could undermine existing SaaS businesses.

Here's my initial assessment on how I think each one applies to Fillvisa.

These are my subjective assessments, not predictions. At first glance, Fillvisa appears quite vulnerable. After all, we're building software that helps people fill out forms. That's exactly the kind of repetitive, structured work AI is getting increasingly good at. But I think there's another side to this.

AI makes it easier to build software. It doesn't necessarily make it easier to operate that software as a dependable service.

And that's where I believe Fillvisa can develop some meaningful competitive advantages.

Moat #1: Continuously maintained immigration forms

This is perhaps the most overlooked aspect of immigration software.

The U.S. Citizenship and Immigration Services (USCIS) regularly publishes new editions of immigration forms.

Sometimes the changes are minor. Other times, questions are added or removed, sections are reorganized, or field structures change entirely. For someone using a paper form, this might just mean downloading the latest PDF. For a company like Fillvisa, it's more complicated.

Our software needs to understand the underlying form structure. When USCIS publishes a new edition, we may need to update:

• Form field mappings and data structures

• Conditional questions and validation rules

• Calculations and field dependencies

• PDF generation and autofill logic

• Automated and manual tests

And we need to make sure the software continues working correctly.

This is recurring, specialized maintenance work.

Now, imagine a small immigration law firm decides to use AI to build its own I-485 form filler. With modern coding agents, it might be able to get a working prototype relatively quickly. But what happens six months later when USCIS publishes a new edition?

Someone needs to identify the changes, update the mappings, verify that the data is being populated correctly, test the output, and deploy the update. And then do it again when another form changes.

The firm could certainly use AI to help with this process. But somebody still needs to own the responsibility for ensuring the software works. Most small immigration firms aren't interested in becoming software maintenance companies. They want to prepare cases and serve their clients.

The economics of centralized maintenance

There's also an interesting economy of scale here.

Let's use a hypothetical example.

Suppose USCIS publishes a major update to Form I-485, and properly updating the software requires 20 developer-hours. If 100 independent immigration firms maintain their own software, that's potentially:

100 firms × 20 hours = 2,000 developer-hours

Now compare that with Fillvisa.

We implement the update once, test it, and deploy it across our hosted platform.

1 platform × 20 hours = 20 developer-hours

Obviously, these are illustrative numbers. Actual development effort varies, and centrally maintaining a platform introduces its own responsibilities. But the underlying economics are compelling. The cost of maintaining a form implementation can be spread across hundreds or thousands of customers. And as our customer base grows, the maintenance cost per customer can decrease. AI might reduce the cost of making these updates, but it can reduce our costs too.

This is why I believe the ability to reliably maintain software matters more than the ability to generate its initial code.

Of course, this only becomes a real competitive advantage if we consistently deliver accurate, timely updates. Merely claiming to support the latest forms isn't enough.

Moat #2: Managed infrastructure

The second distinction is between software that runs locally and software delivered as a managed service.

Fillvisa Free is intentionally browser-based. Users don't need accounts. Their form information is stored locally, and the application doesn't require a backend for storing their data. Fillvisa OS takes this further by making the software available for people to self-host and customize. AI-generated software can serve similar needs. Someone can build a form filler, run it on their computer, and generate a completed immigration PDF. That's perfectly fine.

In fact, I'm actively encouraging this kind of accessibility through Fillvisa Free and OS.

But things change considerably when a professional firm needs to use the software across multiple employees and clients. Now we're talking about authentication, user permissions, secure document storage, backups, availability, deployments, and ongoing infrastructure maintenance.

AI can help build all these things. What it doesn't eliminate is the responsibility of operating them. If a law firm develops its own internal application, it becomes responsible for keeping that application secure, reliable, and up to date. With Fillvisa Plus, the firm pays us to manage the software and its infrastructure. This is a different value proposition. I'm not just selling access to code or features.

I'm selling the convenience of having somebody else operate and maintain a critical business tool.

For a small immigration practice, $40/month can be a much more attractive proposition than maintaining an internal application. And as the software handles more sensitive client information, the operational responsibilities become even more important.

Moat #3: Collaboration is a different product category

Here's another important distinction. Filling out an immigration form is primarily an individual task. A person enters information, generates the PDF, and saves it.

That's exactly the kind of workflow that local software, open-source tools, and AI-generated applications can handle well.

But running an immigration practice involves much more than filling out forms. Consider a typical family-based immigration case. The preparer needs to collect information from the client, request documents, prepare several related forms, review the information, coordinate signatures, and keep track of the case.

Sometimes multiple employees are involved. Sometimes a client needs to provide additional information or correct something. There may be several iterations before a completed application package is ready. This is where Fillvisa Plus comes in.

We've been building features around these collaborative workflows:

Client Immigration Profiles: Information entered once can be reused across multiple forms in the same case.

Client intake: Professionals can collect information from clients instead of manually transcribing it into every form.

Team collaboration: Multiple staff members can work within the same workspace, with roles and page-level access controls.

Document management: Supporting documents can be organized alongside the relevant cases.

E-signatures: Applicants, interpreters, and preparers can coordinate signatures.

An AI coding agent could absolutely help someone build these capabilities.

But once again, there's a difference between creating a working prototype and operating a dependable multi-user application. More importantly, there are switching costs.

Imagine a small immigration firm with hundreds of client profiles, active cases, uploaded documents, and employees collaborating through Fillvisa Plus. Switching to another platform is no longer as simple as replacing a form filler. The firm needs to migrate its data, move ongoing workflows, retrain employees, and potentially change how it interacts with clients.

The more deeply Fillvisa becomes embedded in a firm's daily operations, the more inconvenient it becomes to replace. I think this is one of our strongest potential long-term moats.

Moat #4: Distribution through Free and Open Source

This one is particularly interesting because, on the surface, it might seem counterintuitive. If AI is making software easier to build, why would I also give away our software for free and release the code publicly? Aren't we making it easier for competitors? To some extent, yes.

But I've increasingly come to believe that protecting the source code of a basic form filler isn't where our long-term advantage lies. Instead, I want Fillvisa Free and Fillvisa OS to help us reach people who would otherwise never discover Fillvisa.

Someone searching for a free way to complete an I-485 might discover Fillvisa Free.

A developer looking for open-source immigration software might discover Fillvisa OS.

An immigration nonprofit might use our open-source tools to help low-income families.

A document preparer might start using our free forms and eventually realize that they need client intake, reusable profiles, or team collaboration.

That's when Fillvisa Plus becomes relevant.

The model is:

Free software → Adoption → Trust → Professional usage → Paid workspace

Not everyone who uses Fillvisa Free will become a paying customer. And that's okay.

One of the reasons I'm building Fillvisa is to make immigration paperwork more accessible, particularly for individuals and families who cannot afford expensive software. The free and open-source products serve that mission independently of whether users upgrade. At the same time, they create opportunities for distribution, referrals, and brand recognition. A competitor might use AI to replicate our form-filling features.

But replicating an established community, customer relationships, organic search presence, educational resources, and accumulated trust is a different challenge. Of course, distribution doesn't automatically become a moat. The real test is whether people continue discovering, using, recommending, and eventually paying for Fillvisa.

That's what I want to build.

The combination is more important than any single moat

None of these advantages is particularly strong in isolation. A competitor could maintain USCIS forms. Another company could provide managed hosting. Someone else could build client collaboration features. And plenty of companies have free or open-source products. But I think the combination is more interesting. Fillvisa can offer:

Continuously maintained forms so customers don't need to monitor every USCIS update and maintain form-filling code themselves.

Managed infrastructure so firms don't need to operate their own software systems.

Embedded professional workflows so the product becomes part of how firms manage clients and prepare cases.

Distribution through Free and OS so individuals, developers, and professionals can discover and build trust in Fillvisa before paying anything.

These advantages reinforce one another. Our free products introduce people to the software. Some of those users eventually become professional customers. Those customers adopt Plus for convenience, collaboration, and managed infrastructure.

Over time, their workflows become more deeply integrated with the platform.

And our recurring maintenance work becomes more economical as the customer base grows. That's the business I'm trying to build.

The bigger risk: What if immigration forms themselves become obsolete?

There's one risk that none of these moats fully addresses. What if the immigration filing process changes fundamentally? Today, USCIS relies heavily on standardized forms, many of which are distributed as PDFs. But that doesn't mean the process will always work this way.

Government agencies could improve their online filing systems. More forms could move entirely online. AI agents might eventually help applicants complete much of the process directly. If that happens, a product whose entire value proposition is generating immigration PDFs could become much less useful.

That's why I don't want Fillvisa Plus to remain just a form-filling application. I want it to become a workspace for the broader immigration preparation process. Forms are one part of that process.

Collecting client information, managing documents, coordinating with clients, maintaining records, and preparing complete application packages are other parts. Even if the method of submitting applications changes, professionals may still need software to manage these workflows.

That said, I don't think any SaaS founder can completely protect their business against fundamental changes in customer behavior. The best I can do is stay close to customers and keep adapting the product as their needs evolve.

Building software is getting cheaper. Operating software still has a cost.

My biggest takeaway from the podcast is that the traditional SaaS playbook is changing. For a long time, one of the biggest barriers to competing with a software company was the cost of building the product. AI is rapidly reducing that barrier.

But software development was never the only thing customers were paying for. They also pay for reliability, maintenance, infrastructure, support, security, integrations, and the convenience of not having to manage everything themselves. In the case of Fillvisa, I believe those operational responsibilities can become an increasingly important part of our value proposition. I'm not assuming that AI won't disrupt immigration software. It almost certainly will. And I don't think any of these moats makes Fillvisa immune to competition.

But I'm increasingly convinced that the defensible part of Fillvisa isn't the code that fills out immigration forms. It's the dependable service, professional workflows, and customer trust we're building around that code.

That's the direction I'm taking the company.

And if AI makes building software nearly free, I want Fillvisa to be the service immigration professionals still find worth paying for.