How to automate desktop work without an API

A lot of the software that operations teams use every day has no API. Not a bad API. No API at all. That creates a real problem when you want to automate work, because most automation tools are built around the assumption that your software will talk to them.

This piece explains why that assumption fails so often, what the alternatives are, and how teams are getting around it.

Which software has no API

More than people expect. Here are common examples:

• State insurance carrier portals (almost none of them have APIs)
• Government filing systems, including many EDGAR and state licensing portals
• Legacy CRMs like Redtail in older deployment configurations
• Workers' compensation and benefits administration platforms from smaller vendors
• Custom-built internal tools that were written ten or fifteen years ago
• Desktop-only software that was never designed to connect with anything

These are not fringe edge cases. For many insurance, financial services, and compliance-heavy firms, these systems are the core of daily operations. Agents spend hours every week clicking through them manually.

Option 1: Screen scraping

Screen scraping reads the HTML or pixels of a page and extracts data from them. It was one of the first approaches people used to automate web-based software with no API.

The limitation is fragility. Scraping tools are tightly coupled to the structure of a specific page. When that page changes, the scraper breaks. Maintenance is constant, and scraping only works for reading data, not for filling out forms or navigating workflows.

Option 2: RPA

Robotic Process Automation is a step up. Instead of reading HTML, it records your mouse clicks and keystrokes and replays them. Because it operates at the level of the screen rather than the page code, it works across web apps, desktop applications, and even terminal software.

RPA handles the "click through a portal" problem reasonably well, and it does not need an API. But it is still brittle in a different way. It records a fixed path. If a vendor moves a button, adds a step to their flow, or changes a field label, the bot fails. Someone has to go back in and re-record.

For firms with small ops teams, that maintenance overhead adds up. You trade hours of manual clicking for hours of bot maintenance, and it is not always a good trade.

Option 3: AI desktop agents

This is where the approach changes. An AI desktop agent does not follow a recorded script. It reads the screen the way a person does: it sees the elements, understands the context, and figures out what to click or type next.

Because it reads rather than replays, it handles variation. If a carrier portal redesigns their form, the agent looks at the new form, identifies the fields it needs to fill, and completes them. No re-recording required.

Take a concrete example. An insurance ops team needs to check policy status on a carrier portal that has no API. With an AI agent, the workflow looks like this: the agent opens the portal, logs in, navigates to the policy search, enters the policy number, reads the status from the screen, and logs it to the internal system. All of that happens without any integration work on the carrier's side.

The same approach works for government portals, legacy CRMs, and any other desktop software a person can navigate. If a person can click through it, an AI agent can too.

How Zo handles this

Zo runs on your desktop and watches how you work. You show it a workflow once or twice. It learns the steps, the apps involved, and how to handle variations. After that it handles the workflow on its own.

This works with any application. No integration needed, no API key, no waiting for a vendor to build something. Zo sees whatever is on the screen and operates through the same interface you do.

For firms where large parts of daily operations run through software that has no API, this is not a workaround. It is the right tool for the job. Most firms are fully set up within four weeks, with no dedicated IT resources required.

And because Zo builds a record of every workflow it runs, the institutional knowledge does not disappear when people leave. The process is documented and repeatable from day one.

If your team is spending hours a week on work that runs through systems with no API, it is worth seeing what that looks like when it is handled automatically.