Can you automate an old system with no API?
Usually yes, but it is the one thing that genuinely changes both the price and the risk, so it belongs on the first call rather than discovered in week three. There is nearly always a route in: a scheduled export, a direct database connection, an email or file drop, or (last) driving the interface the way a person would.
The order matters. Each option down that list is slower to build, more fragile in service and more expensive to maintain. The job is finding the highest one on the list that your system supports, not reaching for the bottom one because it always works.
The honest warning: if the only viable route is interface automation, expect it to need attention whenever the vendor changes a screen. That is an ongoing cost, not a one-off, and any proposal that does not say so is understating what you are buying.
The five routes in, best to worst
1. A documented API
The straightforward case, and more systems have one than their users realize, including software that looks two decades old. Check before assuming. Also check whether API access is included in your current license tier rather than one above it, because that discovery mid-build is an avoidable and irritating surprise.
2. Scheduled export or import
Many older systems cannot answer a live question but will happily produce a file on a schedule, or swallow one. This is a perfectly respectable integration. It is not real time. You inherit the export's cadence, so an hourly file means an hourly automation, but it is stable, it survives vendor updates, and it rarely breaks in surprising ways.
Most people dismiss this too quickly because it feels unsophisticated. For a nightly reconciliation or a daily sync it is often the correct answer and the cheapest one.
3. Direct database access
Where the system runs on a database you control, reading from it directly is fast and reliable. Two cautions. Reading is far safer than writing: writing around the application skips whatever validation and business logic the application applies, which is an excellent way to create records the software itself considers invalid. And the schema is undocumented and unstable. A vendor update can rename a column without telling anyone, because officially nobody was looking.
Our default is read-only from the database, and writes through whatever legitimate route exists, even if it is slower.
4. Email or file drop
If the system can email a report or drop a file somewhere, that is a trigger. Genuinely useful for legacy software whose only outbound capability is a print-to-email. Combine with extraction if what arrives is a PDF rather than data, and accept that you are now dependent on the format of that report not changing.
5. Driving the interface (RPA)
Software clicking buttons and reading the screen as a person would. It works, it is the reason RPA exists, and it is the last resort for a reason: it breaks whenever the interface changes, and interfaces change without warning or announcement.
If this is the route, three things must be in the proposal. Explicit failure detection, because a screen robot that mis-clicks can do real damage silently. A maintenance expectation in writing. And a plan for what happens during a vendor update. A proposal that leads with interface automation and does not mention maintenance is underpricing the work.
Ask your vendor directly: "Do you have an API, and is it included in our plan?" Then ask "Can the system export on a schedule?" Those two answers place you on the ladder in about a day, and they change the estimate more than anything else you could tell us.
How to find out what you have
- Search the vendor's site for "API" or "integrations". Thirty seconds, and it settles it more often than people expect.
- Check the admin settings for API keys, webhooks, or an integrations section. Presence of an API key field is presence of an API.
- Ask support the two questions above. Get the tier answer in writing.
- Look at what it can already export. Any scheduled report is a potential integration.
- Find out who administers it and whether anyone still has the vendor relationship. On genuinely old systems this is often the real blocker, not technology, but that nobody knows who the contract is with.
That last point is worth stressing. The most common cause of delay on these projects is not engineering difficulty; it is spending eleven days finding out who can authorize access. Sorting it during scoping removes it entirely.
What it does to the price
A system with no usable API moves an automation up the price list rather than off it. Working out whether an export, a database connection or interface automation is the least-bad route can take a day or two on its own, and each option below a real API is slower to build and more fragile once running.
Two practical consequences. First, it is worth telling us on the first call. It changes both the timeline and the price, and a quote that has not accounted for it will move later, which serves nobody. Second, where interface automation is the only route, budget for maintenance the way you would for any moving part. That is where a Growth or Scale plan, with monitoring and fixes included, earns its keep; on pay as you go, each fix is billed as a $175 change.
When the answer is genuinely no
Three cases where we would tell you not to.
- The system is being replaced within a year. Build the automation against the new one. Anything built against the outgoing system is written off on the migration date.
- Interface automation against something that changes constantly. A cloud product shipping weekly UI updates is a poor target for screen automation. You will spend more on repairs than the automation returns.
- The value does not clear the extra cost. A process worth $200 a month does not justify the engineering that a no-API route requires. Frequency and value still decide it; see what should not be automated.
There is also a legitimate middle path: automate everything around the legacy system, and leave the one manual step where it is. A person copying one figure into an old system once a day, with everything before and after handled automatically, is often a much better deal than a fragile robot doing the same keystroke.
Not sure what your system supports?
Tell us what it is on the call and we will tell you which route in is available, before anyone quotes. Thirty minutes, no charge.
You will talk to Fayez or Kareem. Nobody else.