Automation: Which Processes to Tackle First

When businesses decide to automate, the temptation is to tackle everything at once. But automation projects that try to solve all problems simultaneously tend to stall, run over budget, and deliver disappointing results.

The most successful automation initiatives start small and build momentum. In this article, we share the framework we use with clients in Haarlem and across the Netherlands to identify which processes to automate first for the fastest return.

McKinsey reports that organizations scaling automation deliberately outperform those that automate ad hoc. Prioritization is not bureaucracy; it is how you avoid spending budget on workflows that never move the needle.

We evaluate each candidate process on two dimensions: how often it occurs and how much time or money each occurrence costs. A task performed 50 times per day that takes 5 minutes each time is a much better automation candidate than a monthly task that takes an hour.

Plot your processes on a simple grid. The upper-right quadrant, high frequency and high impact per occurrence, is where you start. These are your quick wins. They deliver measurable savings from day one and build organizational confidence in automation.

Add a third lens where useful: error cost. A low-frequency task that triggers expensive rework, such as wrong invoice lines or missed SLA deadlines, can jump the queue. ISO automation standards emphasize documenting the process before you automate it; that discipline alone often reveals steps you can delete before writing code.

In our experience, the highest-ROI first automations are almost always data entry and notifications. Moving data between systems, sending status updates, and generating routine reports are repetitive, error-prone, and easily automated.

For example, automatically creating an invoice when a project milestone is marked complete, or sending a customer a status update when their order ships. These automations are simple to build with the right deployment pipeline, immediately useful, and give your team a taste of what is possible.

They also pair well with custom web applications and portals: once customers can see status themselves, notification automations reduce inbound email rather than adding noise. Start where humans copy-paste between tabs; that is where hours leak away unnoticed.

One critical mistake is automating a process that is fundamentally flawed. If your approval chain has seven unnecessary steps, automating all seven just makes a bad process happen faster. Fix the process first, then automate.

We always start automation projects with a process review. We ask: does this step actually need to exist? Can these three approval stages become one? Is this data being collected because someone needs it, or because it has always been collected? Simplify, then automate.

Involve the people who run the process daily, not only IT. Gartner's hyperautomation research stresses that discovery workshops prevent building elegant software around workflows nobody wants. A half-day mapping session often saves weeks of rework.

After identifying your quick wins, build a roadmap with three horizons. The first horizon covers the next 30 days: simple automations you can deploy immediately. The second horizon spans 3 to 6 months: more complex workflows that require integration work.

The third horizon is 6 to 12 months out and covers strategic automation, things like predictive analytics, AI-powered decision support, or complete end-to-end process automation. For inspiration, browse our collection of business process automation examples. By this point, the early wins have already paid for the investment.

Document owners per horizon. Automations without an owner become "the script only Mark understands". When Mark leaves, the automation breaks quietly. Assign a business owner and a technical owner, even for small scripts, and review quarterly whether each automation still matches how work actually happens.

In 2026 the prioritization framework still holds, but a new option has joined the toolbox: AI agents. Where classic automation follows fixed rules, an agent can handle preparatory work that is too variable for rigid scripts, such as gathering data from multiple systems and drafting a proposal a human then approves. The frequency-times-impact grid still decides what to tackle first; agents simply widen what is automatable in the third horizon.

The caution is unchanged: never automate a broken process, and keep a human in control of irreversible decisions. We cover where agents make sense, and where they do not, in our guide on building an AI agent for your business processes.

Almost every automation on your roadmap depends on systems talking to each other. An invoice cannot generate itself if your project tool and accounting package are not connected. In practice, the majority of automation work we do for clients is integration work: linking a CRM to a planning system, a webshop to a warehouse, a form to a workflow.

This has a practical consequence for prioritization. Before committing to an automation, check whether the systems involved expose an API and what that connection costs to build. A quick win that requires a complex integration with a closed legacy system may belong in horizon two rather than horizon one. Conversely, once a solid integration exists, additional automations on top of it become cheap.

The first connection carries the cost; every automation after it rides for free. Use our project calculator to put rough numbers on both sides of that equation. Where APIs are missing, sometimes a lightweight middleware layer or nightly sync is enough to unlock horizon-one wins without waiting for a full legacy redevelopment.

A consultancy near Haarlem Central spent roughly eight hours each week compiling client status from three tools: time tracking, project management, and spreadsheets. Partners wanted one view before Monday meetings; juniors copied numbers by hand. We did not start with AI or agents. We connected the APIs, normalized the data model, and built a small internal dashboard that refreshed hourly.

That single automation paid back in under two months. Only then did we automate client-facing status emails triggered from the same data. The lesson repeats across industries: internal visibility first, external notifications second, strategic agents third.

If your team maintains a "weekly report ritual", that is usually horizon one in disguise. We describe comparable builds on our web application development page and through software integrations. A two-week discovery usually surfaces one process worth automating before any large roadmap conversation.

Automation is a journey, not a destination. Start with the processes that hurt most, prove value quickly, and expand from there. Prioritize integrations and data quality before exotic AI; the boring automations often fund the ambitious ones.

If you want help identifying your highest-impact automation opportunities, we offer a free process assessment to get you started, or explore integration options if disconnected systems are the blocker.