AI, robotics and crypto, in context Deutsch

Automation

Buying Automation Software: What to Check Before You Sign

Most automation tools look alike in a demo. The differences show up later, in licensing, maintenance and the first broken integration. A buyer's guide to the questions that matter.

Buying Automation Software: What to Check Before You Sign

Few software purchases look as easy as automation. The demo shows a form that fills itself, an invoice that routes itself and a report that lands in an inbox at seven every morning. Within an hour, the sales engineer has built a workflow that would have taken the team a week by hand.

The trouble starts months later. A connected system changes its interface and three workflows stop. The person who built them has moved on. Licensing turns out to be priced per run, per bot or per user, and growth makes the bill climb faster than expected. None of this was visible in the demo.

This guide is for the people who have to sign the contract: operations leads, finance teams and IT managers at small and midsize organizations. It does not rank products. It explains which questions separate a tool that will carry real work from one that will end up as shelfware.

Know which kind of tool you are buying

Automation software is not one category. Vendors blur the lines in their marketing, but the underlying designs differ. Each family suits different jobs and fails in different ways.

  • Workflow and integration platforms connect applications through their official interfaces (APIs). They move data between systems when something happens, such as a new order or a signed contract.
  • Robotic process automation (RPA) drives existing screens the way a person would, by clicking and typing. It helps when an old system offers no interface, but it breaks when the screen layout changes.
  • Low-code app builders let business teams create small forms, approvals and databases without much programming.
  • Built-in automation inside ERP, CRM or accounting systems handles rules within one product. It is often already paid for.
  • Scripts and scheduled jobs written by developers remain the cheapest option for narrow technical tasks, as long as someone maintains them.

A common and expensive mistake is to buy a screen robot for a job that a simple API connection could do more reliably. The reverse also happens. A team buys an integration platform and then discovers that its most important system has no usable interface at all. Before comparing vendors, decide which family fits the work.

Read the pricing model as carefully as the feature list

Automation tools are priced in very different ways. Some charge per user, some per workflow run or task, some per bot or per connected application. Each model rewards a different usage pattern. Per-run pricing looks cheap during a pilot and can grow quickly once a high-volume process goes live.

Ask the vendor to price three scenarios: today, double the volume and ten times the volume. Ask which features sit in higher tiers. Single sign-on, audit logs and version control are often reserved for enterprise plans, and those are exactly the features a growing deployment needs.

Then add the costs that never appear on the quote. Someone has to design, test and document each workflow. Someone has to watch it, fix it and update it when connected systems change. In many organizations, this internal time ends up larger than the license itself.

Test integrations against your real systems

Every vendor claims hundreds of connectors. What matters is the depth of the few you need. A connector that can create a customer record but cannot update one, or that ignores custom fields, will force workarounds within weeks.

Insist on a pilot with your own systems and your own data, not a sandbox prepared by the vendor. Pick a process that is moderately complex and slightly messy, for example orders that sometimes arrive with missing fields. Watch how the tool handles those exceptions. Clean demos tell you little. The awkward cases tell you almost everything.

During the pilot, also check how errors surface. Does a failed run send an alert to a named person? Can you rerun it safely without creating duplicate invoices? Can you see what data passed through each step? These questions sound dull, yet they decide whether the tool can carry work that matters.

Settle ownership and governance before go-live

Automation tends to spread. Once one team has a working tool, others start building their own workflows, often with personal accounts and copied credentials. After a year, nobody knows how many automations exist or which ones touch customer data.

Good tools make governance possible, but they do not enforce it on their own. During evaluation, look for role-based access, central credential storage, audit logs and a clean way to move workflows from test to production. Then decide who may build, who must review and who gets called when something fails outside office hours.

An automation without a named owner is not a saved cost. It is a future incident with an unknown date.

Security and privacy reviews belong here too. An automation account often has broad access to several systems at once, which makes it an attractive target. Treat it like a privileged user, with limited rights, regular reviews and credentials stored in a vault rather than inside the workflow.

Plan the exit on day one

Workflows built in a proprietary designer rarely move cleanly to another product. If the vendor raises prices, gets acquired or drops a connector, you may face a rebuild. That risk does not rule out a platform, but it should shape the choice.

Ask whether workflows can be exported in a readable format, whether logs and run histories can be downloaded and how long data is kept after cancellation. Keep your own record of each workflow's purpose, inputs and outputs outside the tool. That record makes a later migration far less painful.

What decides in the end

When the shortlist is down to two or three candidates, the deciding factors are rarely features. They are fit, cost over several years and how much the setup depends on a few people. Organizations that buy well usually do three things. They pilot on real work, they price growth honestly and they name an owner for every workflow before it goes live.

The best automation software is often the least exciting in the demo. It handles errors visibly, scales its price predictably and lets someone other than its original builder understand what it does. That sounds like a low bar. In practice, it is a surprisingly high one.

Questions and answers

Do small businesses need dedicated automation software?

Not always. Many accounting, CRM and shop systems already include rules and scheduled tasks that cover simple needs. A separate platform makes sense once several systems have to exchange data or the same manual steps repeat across teams.

What is the difference between RPA and an integration platform?

An integration platform talks to applications through their official interfaces, which is usually stable and fast. RPA imitates a person using the screen, which works with older systems that offer no interface. It is more fragile, because small changes to a screen can stop a robot.

How long should a pilot run?

Long enough to see real exceptions, which often means several weeks of normal operations rather than a few days. The pilot should include at least one failure and its recovery. If nothing goes wrong, the test was probably too clean.

Who should be involved in choosing the tool?

The people who own the process, someone from IT or security and whoever will maintain the workflows afterward. Finance should check the pricing model against expected growth. Leaving out the future maintainers is the most common gap.

Is open-source automation software a good alternative?

It can be, especially for teams with technical staff who can host, update and secure it. The license is free, but running the platform is not. A fair comparison includes the internal time needed to operate it reliably.