Tools/March 24, 2026

Best desktop automation tools that work with legacy software

Legacy software is not going away. Insurance portals built in 2008. ERPs that only run on Windows XP virtual machines. Municipal permitting systems with no documentation. Trading platforms that vendors stopped updating a decade ago. These tools power real workflows at real companies, and nobody is replacing them anytime soon.

The problem is they were built before automation was part of the conversation. No API. No webhooks. Sometimes no keyboard shortcuts at all. So when you want to automate work that runs through them, your options are limited and the tradeoffs are real.

Here is a practical look at what actually works.

Where legacy software shows up

Before getting into tools, it helps to know which industries are most affected. Legacy software is concentrated in:

Insurance. Carrier portals for quoting, binding, and claims are notoriously old. Many are web-based but use non-standard form structures that break standard automation.
Financial services. Fixed-income trading platforms, custodian portals, and older order management systems are often thick desktop clients with no export function worth using.
Healthcare. EHR systems like older versions of Epic, Meditech, and Allscripts all have APIs now, but many smaller practices run on-premise deployments from years ago that never get updated.
Government and municipal work. Permitting software, utility systems, court filing portals. State and local government IT budgets are tight and replacement cycles are slow.

If your team works in any of these areas, you already know the feeling: someone sits there copying data from one screen to another, all day, every day.

Traditional RPA: UiPath and Automation Anywhere

UiPath and Automation Anywhere were built for exactly this problem. They use screen capture, image recognition, and UI element detection to interact with software the same way a person would. No API required.

They work. For stable, high-volume processes inside large IT environments, they are the industry standard for a reason.

The issues:

Cost. UiPath licensing starts around $10,000 to $15,000 per year for a single attended bot. Automation Anywhere is comparable. Add implementation costs from a consultant or internal developer and year-one spend often hits $50,000 to $150,000.
Brittleness. Traditional RPA records exact pixel coordinates or UI element IDs. When the software updates, or someone changes screen resolution, or a portal changes a field label, the bot breaks. Someone has to fix it.
Setup time. A typical RPA project takes 4 to 12 months from scoping to production. That is before maintenance begins.

For a 500-person firm processing thousands of transactions a day, this math works. For a 30-person team, it usually does not.

Lightweight tools: AutoHotkey and Power Automate Desktop

AutoHotkey is a free Windows scripting tool that lets you automate keystrokes, mouse clicks, and window interactions. Power Automate Desktop is Microsoft's free desktop automation tool, included with Windows 10 and 11.

Both can handle basic legacy software automation. If you need to fill out the same form 20 times a day with predictable inputs, these tools can do it.

The limitations are significant:

Someone needs to write and maintain the scripts. AutoHotkey requires scripting knowledge. Power Automate Desktop has a visual interface but still needs a technically capable person to configure and debug it.
They break on variation. If the form changes, if an error dialog appears, if data is in an unexpected format, the script stops and a person has to intervene.
They are designed for simple, linear processes. Anything that requires reading data from the screen and making decisions based on it gets complicated fast.

Good for: one-off automations on stable software where someone on your team is comfortable maintaining scripts. Not great for anything that needs to handle variation or run without oversight.

AI desktop agents

A newer category of tools takes a different approach. Instead of recording exact coordinates or writing scripts, they read the screen the same way a person does, understand what they are looking at, and figure out the right actions to take.

This matters for legacy software because the agent does not need to know in advance exactly where a button will be or what a field will be called. It reads the screen, understands the context, and acts accordingly.

Zo works this way. You describe the workflow in plain language. Zo learns it by watching a person do it a few times, then runs it on its own. It works on any desktop software, including apps that traditional RPA struggles with, because it is not tied to specific UI element IDs or pixel positions.

The practical differences:

Setup is faster. Most workflows go live in 4 weeks rather than 4 months.
It handles variation better. If a form field is in a slightly different position, or a popup appears, the agent can adapt.
Cost is closer to $2,000 to $5,000 per month depending on workflow volume, which is significantly less than enterprise RPA.
No developer required. The workflow description happens in natural language, not code.

How to choose

The right tool depends on a few factors:

Volume and complexity. High-volume, complex workflows at a large firm: traditional RPA is worth the investment. Low to mid-volume at a smaller firm: the overhead is hard to justify.
How stable is the software? If the legacy system never changes, script-based tools hold up fine. If it updates periodically or has variation in inputs, you want something more adaptive.
Who will maintain it? RPA and AutoHotkey need ongoing technical maintenance. AI desktop agents require less, but still need someone paying attention when workflows change.
Budget and timeline. If you need something running in weeks, not months, and do not have a six-figure automation budget, the traditional RPA path is probably not the right starting point.

Legacy software is a real constraint. But it does not have to mean someone on your team is stuck copying data by hand. The tools have gotten better, and the right one for your situation probably costs less and takes less time to set up than you expect. If you want to see what this looks like on your specific workflows, it is worth having that conversation.