Automate work that is frequent, repeatable, stable enough to describe, and measurable. Start with a process that has a clear owner, usable inputs, manageable exceptions, and enough volume for the benefit to matter. Keep human judgment where risk, context, or trust matters.
The process everybody complains about most is not automatically the best one to automate.
Pain is a useful signal, but you still need to understand why the process is painful.
Sometimes the problem is repeated manual work. Sometimes it is a broken handoff, unclear ownership, bad information, or a rule nobody agrees on. Automation only helps with some of those problems.
The same complaint can come from two different processes
I worked with two accounting firms that came to me with almost the same issue: during tax season, documents were arriving late, deadlines were getting close, and clients were becoming frustrated.
It would have been easy to give both firms the same solution.
But the first firm's bottleneck was on the client side. It already had a portal, but clients found it confusing and kept emailing documents instead. The useful intervention was a clearer intake experience and a live view of what was still missing.
The second firm's client intake was fine. Its employees had no shared view across all active clients, so files falling behind were discovered too late. That firm needed internal visibility and automatic deadline flags.
Same complaint. Different process.
This is why we map the work before choosing the tool.
Score the process before automating it
| Question | A strong automation signal |
|---|---|
| How often does it happen? | Daily or weekly, with enough volume for saved effort to accumulate |
| Can the normal path be described? | Inputs, decisions, outputs, and ownership are understandable |
| Is the process stable? | The underlying policy is not changing every week |
| What happens when it goes wrong? | Exceptions can be detected and routed to a responsible person |
| Are the inputs usable? | Required information is accessible and sufficiently reliable |
| Can success be measured? | Time, cost, errors, backlog, throughput, or capacity has a baseline |
| How much judgment is involved? | The normal path is rules-based and judgment-heavy cases can be escalated |
A process can look suitable and still be a bad investment. After assessing the workflow, compare the expected benefit with implementation, integration, training, monitoring, and support costs.
Processes that are often worth evaluating
Approvals and internal requests
Purchase requests, access requests, expenses, discounts, and other approvals can often be collected, routed, tracked, and escalated automatically.
If every request becomes a special case because decision rights are unclear, fix the policy first.
Document intake and data entry
Applications, invoices, claims documents, forms, and onboarding files can be received, classified, checked, and entered into the appropriate system.
AI may help interpret variable documents, while normal rules validate required fields and send uncertain cases to a person.
Movement between systems
Repeatedly copying customer, order, project, or financial information between systems is a strong systems integration signal.
First decide which system owns the record. Automating a disagreement between two systems only makes the disagreement move faster.
Recurring reporting and reconciliation
A stable report can gather trusted data, apply agreed calculations, prepare the output, and flag differences for investigation.
Do not automate a report nobody trusts. Resolve the definitions and source ownership first.
Onboarding and follow-up
Employee or customer onboarding often combines documents, approvals, account creation, communications, and handoffs. Routine reminders and status updates can also be automated when the rules are clear.
Keep human interaction where reassurance, judgment, or relationship-building matters.
How to automate a business process
- Define the boundary. State where the process begins, where it ends, who owns it, and what a completed output looks like.
- Follow real cases. Record every system, handoff, wait, re-entry, exception, and workaround—not just the clean process document.
- Remove unnecessary work. Delete redundant steps and clarify rules before automating what remains.
- Design the target process. Define inputs, decisions, systems, permissions, human review, exceptions, outputs, and measures.
- Choose the simplest suitable mechanism. It might be a native feature, integration, workflow platform, script, focused custom system, AI inside one step, or a combination.
- Test difficult cases. Use missing fields, duplicate submissions, conflicting records, failed connections, and requests outside the normal rule.
- Launch with ownership. Train each role, document the workflow, assign an internal owner, and measure the result against the baseline.
Do not automate these processes first
Avoid processes that are rare, poorly understood, changing constantly, full of unmanaged exceptions, or dominated by high-stakes judgment.
Sometimes the answer is to remove a step, clarify a rule, assign an owner, configure software you already own, or connect two systems.
Do that first.
At Losung Consult, we help organizations map, redesign, and implement business processes through workflow automation across the systems they already use.
Start with one process and one measurable outcome. That is enough to find out whether automation makes sense.
Map one process before buying a platform
Losung helps businesses redesign and implement workflows that teams can see, test, and own.
Explore workflow automation consulting