Insights

Insights

Replacing SaaS With Custom Software – AI and the New Build vs. Buy Math

Agentic engineering has cut the cost of custom software by 30–50%. For the subscriptions that run your core business, replacing them can now pay back in one to three years.

Smiling man with light brown hair and glasses wearing a light purple shirt in an outdoor setting.

Shawn Crowley

Two women sitting in front of a Dell computer monitor in an office, one explaining something while the other listens attentively.

Get posts like this delivered to your inbox

We'll send our top posts on a monthly basis.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.

For most of the last two decades, when a company needed software to run a core business process, the build vs. buy analysis pointed to buy. A SaaS product cost less up front, carried less risk, and could be running in months. Building custom software meant a larger budget, a longer timeline, and more ways for the project to go wrong.

That bargain erodes over time. Your team changes how it works to fit the software. The vendor decides what gets built next. Fees climb at every renewal and grow with every user or project you add. A few years in, many companies are paying more each year for a product that fits them less well.

When Replacing SaaS With Custom Software Pays Off

Agentic engineering changes the math. We're seeing it cut the cost of building custom software by roughly 30–50%, and the SaaS product you already use gives agents a detailed reference for what to build. Companies we're talking with now see payback periods of about one to three years for replacing a core SaaS product, after counting the cost of owning and supporting the new system.

Replacement isn't right for every tool. Payroll, email, and accounting software should usually stay SaaS. Below I'll cover where replacement makes sense, what it costs to own the result, and how to tell which situation you're in.

Why Buying Won, and What Tradeoffs You Accepted

Control has always cost more, which is why buying usually won. A SaaS vendor spreads its development costs across thousands of customers, so you get a capable product for a fraction of what it would cost to build one yourself. In exchange, you accept the product as the vendor designs it.

Those concessions tend to fall into four groups.

Fit. You change your processes to match the software's workflows, because the software won't bend to match yours. Your people click through screens full of features they never use when what they need is a short, direct workflow. Where the product falls short, you build extensions, integrations, and workarounds to fill the gaps.

Control. The vendor owns the roadmap. Features you need may never ship, and updates you didn't ask for arrive on the vendor's schedule. The processes that set you apart from competitors are often the hardest to customize, because the product was built for the typical customer.

Cost. Renewal prices rise, and by then the software runs critical operations, so you have little room to negotiate. Per-seat and volume-based pricing means your bill grows as your business grows, even when the price per seat holds steady. Many companies also pay for seats and modules that hardly anyone uses.

Data and risk. Your operational data lives in someone else's system. Exports are limited, API access may be restricted or cost extra, and vendors increasingly sell AI features as add-ons while limiting how your own AI tools can reach your data. You also carry the vendor's risks: an acquisition, a change in product direction, or a product that gets retired.

Taken together, the value you got by buying erodes through rising costs and growing friction.

Why Companies Stay Stuck

Most leaders I talk with already know their core SaaS product isn't serving them well. They stay for understandable reasons.

The software has become an essential dependency, so switching puts critical operations at risk and means retraining people. Years of extensions and integrations make any replacement more complex, whether the replacement is custom software or another SaaS product. Evaluating alternatives takes significant time and money, and there's a reasonable suspicion that a new product would bring the same problems. There's always a contract to finish first. And many companies don't have the internal capacity to own and support software.

Each of those concerns has a better answer now than it did a few years ago.

What Changed

I've written about how agentic engineering changes the economics of legacy modernization: roughly 30–50% lower cost, with the final number depending on how much of the work is rebuilding known behavior versus building something new. Replacing a SaaS product follows the same pattern. The workflows you keep are known behavior, documented and observable in the running product, and that's the work agents speed up most. Redesigned workflows and new features compress less, about as much as any new development.

In practical terms, $2 million of agentic development can now deliver what used to take roughly $3–4 million. That moves the point where building costs less than buying. Systems that weren't worth building a few years ago can now pay for themselves within a few renewal cycles.

Your SaaS Product Is the Spec

Replacing a SaaS product has an advantage over most custom builds: the thing you're replacing is still running, and it's well documented. Agents can read the vendor's documentation, walk the workflows your team actually uses, and turn what they find into functional specs before any budget is committed. Admin reports and audit logs show which users and modules are active. Where a product doesn't report usage in detail, the data itself fills in the picture: a module with no new records in a year isn't part of how your team works. The scope becomes your slice of the vendor's product, and you can redesign the experience along the way.

Agents work from everything you can see through your licensed access: the documentation, the live screens and workflows, the API, your configuration, and your exported data. The specs they produce are checked against the running product, and those specs then drive both the code and a cleaner interface design.

That smaller scope is the biggest difference from modernizing a legacy custom application. In a modernization, you usually need to replace everything the old system does. When you replace a SaaS product, you rebuild the part your business uses, which is often a fraction of a platform designed to serve every customer in its market. Because the mapping happens before you commit a budget, the scope is known up front, and the estimate is more reliable.

The SaaS product also keeps running during the build, so agents can run the same inputs through both systems and compare the results. I explain how this works in more detail in my article on legacy modernization. It's one of the main reasons rebuilding known behavior is the least expensive work in a project.

You copy what the product does for your business and leave behind the parts that slow people down. A workflow that takes six screens in a general-purpose product might take two in software built for one company.

Where the Limits Are

There are limits. Some of what a SaaS product does can't be observed from the outside, such as internal calculations, rules engines, and years of accumulated configuration. Some dependencies are easy to forget until they break, like a scheduled report or an automated email someone relies on. Usage mapping and side-by-side testing shrink this risk without eliminating it, so a good scoping phase looks for these explicitly.

What Owning It Costs

Owning your software comes with ongoing costs. You'll pay for hosting, maintenance, security and compliance, and for a period of overlap while you move off the old system. Your current license fee already covers most of these, since the vendor builds hosting, support, and compliance into its price along with its margin. When you own the software, you take on those costs directly and decide how much to spend on each.

There are one-time costs too: moving your data, reconnecting integrations, and training your people on the new system.

Ongoing Costs

For a core operational system, ongoing costs typically run $100,000–300,000 a year, depending on how much you continue to improve the system. Hosting, storage, and monitoring are a small part of that, often around $18,000 a year. A support contract covering about a week of developer time each month brings the total to roughly $115,000.

That keeps the system patched, secure, and running, with little room for new work. A dedicated developer, or a larger support contract, puts the total closer to $200,000–300,000 and keeps the system improving as your business changes. For a system that runs a core part of the business, I'd plan on the higher range. The amount is a choice about how fast the system improves, and you make that choice instead of a vendor.

How SaaS Support Costs Compare

SaaS has support costs of its own. Most companies running a core platform have one or more internal administrators who configure workflows, manage users, and build reports, and many also pay outside consultants for extensions and integrations. A custom system needs an internal owner too, so those roles roughly balance out. The ongoing cost of custom software covers hosting, maintenance, and steady improvements. Larger new features are separate investments you choose to make, on your timeline instead of the vendor's. The examples below compare subscription fees with build and support costs. They leave out consultant and extension spend on the SaaS side, so the payback estimates lean conservative.

Switching Over in Phases

You may not have to switch over all at once. A phased approach replaces one workflow or module at a time, which spreads out the risk and shortens the stretch where you're running both systems. Depending on how your SaaS contract is priced, you may be able to reduce licensing costs as each piece moves over. Some pricing models make that hard until you leave entirely.

The Math: 3 Examples

The measure I use here is payback period: how long it takes after launch for avoided subscription fees to cover the cost of building and running the replacement. For the companies we're talking with, that's about one to three years.

Replacing a Construction Management Platform

We recently spoke with a construction firm spending about $1 million a year on its construction management platform, with fees that rise as its project volume grows. A focused custom system covering the workflows it actually uses might cost $2–2.5 million with agentic development and $100,000–300,000 a year to host, support, and improve. That system would pay for itself in about three years, sooner as the firm grows. The same build would have cost $3–5 million with traditional development.

Extending Software You Already Own Instead of Buying

The same math applies before you sign. A manufacturer we're starting work with was evaluating a CRM and a project and service management platform that together would have cost about $500,000 a year, a figure that would grow as the company scales. Instead, they're extending custom software they already own. We expect to deliver the functionality they need for $400,000–500,000, which pays back in about a year.

Weighing a Per-Seat Legal AI Platform

The same question is now reaching AI tools themselves. We're talking with a major regional law firm that expects to spend nearly $1 million a year on a legal AI platform, priced per seat. Platforms like this are built on the same foundation models any company can license directly. That makes them a candidate for the same math, especially when only some of those seats are heavily used, or when the firm's real advantage lies in its own precedents, templates, and practice workflows.

The economics differ from replacing a traditional SaaS product. Model usage is a real ongoing cost, and accuracy and security requirements in legal work are high. But for a firm paying per seat for capabilities it can build on directly and tailor to how it practices, the question is worth asking.

When to Replace, and When to Keep the SaaS

Replacing a SaaS product is worth considering when most of these are true:

  • It runs a core or differentiating process, not a commodity one.
  • Your usage is large or growing, with per-seat or usage-based fees that rise as you grow.
  • Your team changes how it works to fit the software.
  • Features you need aren't on the vendor's roadmap, and updates arrive whether you want them or not.
  • You've built extensions and workarounds to make it fit your process.
  • Fees keep rising, and you have little room to negotiate.
  • Your people work around feature bloat and a cluttered interface.

Keep the SaaS when it handles a commodity function like payroll, accounting, email, or basic HR. Every competitor uses the same tools there, and owning your own brings little benefit.

Here's where we see the strongest candidates by industry:

Industry Commonly replaced SaaS category
Healthcare Patient scheduling and intake, care coordination, referral management, and other tools built around the electronic health record (EHR)
Financial services Client onboarding, loan origination workflows, advisor CRM and reporting portals
Education Admissions and enrollment CRM, scheduling, and add-ons to student information and learning systems
Insurance Claims management, underwriting workbenches, agency and broker management
Construction Construction and project management platforms
Manufacturing CRM, quoting and configure-price-quote tools, quality and shop-floor systems
Legal Practice management, per-seat AI platforms

In regulated industries, the core system of record, such as an EHR or a core banking platform, usually stays. The opportunity is in the per-seat SaaS products built around it, and compliance work such as HIPAA counts for more of the ownership cost.

Even if you decide to keep the SaaS, a credible estimate for a custom build gives you bargaining power at renewal time.

What You Get When You Own It

The benefits line up with the problems that prompted you to look for alternatives.

Fit. The software is shaped around your process and the things that set you apart, so you're not using the same two or three platforms as every competitor. Workflows can be designed for how your people work, and the gains are measurable in time per task and steps removed.

Control. You decide what gets built and when. Updates happen when you want them, and you can respond to a change in your business without waiting on a vendor's roadmap. There's no acquisition or product retirement to plan around, and security and compliance follow your requirements, with your data in your environment.

Cost. Costs are predictable and don't climb as you grow. You can add users and usage for far less than per-seat pricing would charge. You also own a software asset, and in some cases development costs for internal-use software can be capitalized, while subscription fees are an operating expense. Talk to your finance team about how that applies to you.

Data and AI. You own your data and your integrations, so you can collect and connect operational data in the way you want. That's what makes AI automation across the company practical, for individuals and across departments. For example, an agent could check a project's schedule, its open purchase orders, and its crew assignments, then flag a conflict before it delays the job, because all three systems share data you control.

Getting Started

Start with an agent-accelerated scoping phase. Map the workflows your team actually uses, measure usage, and come out with a credible cost, a defensible range, and a phased plan.

Plan the replacement in phases if possible, and start well before your next renewal date so the contract timeline works in your favor. Bring in a team with experience in both product development and supporting software after launch.

If you're paying a significant amount each year for SaaS that runs a core part of your business, we can help you map what you use against what a custom system would cost to build and own.

‍

Replace SaaS With Custom Software – When AI Makes It Pay Off

Agentic engineering has cut the cost of custom software by 30–50%. For the subscriptions that run your core business, replacing them can now pay back in one to three years.

Tell Us About Your Project

We'd love to talk with you about your next great software project. Fill out this form and we'll get back to you within two business days.

Share Your Project
Want to see what we can do?

Check out some of our work.