How to automate a manual business approval workflow.

Start by making the approval path visible

Many approval processes begin simply. Someone asks for something, a manager says yes, and the work proceeds. As volume grows, the request may travel through email, WhatsApp, spreadsheets and verbal conversations before anyone acts on it.

A useful workflow system does not begin with software screens. It begins with a clear path. A common pattern is:

request → review → approval → production/action → notification → reporting

The purpose of automation is to make that path easier to follow, not to remove every human decision.

Why manual approval chains become difficult

Email chains lose context

The original request is in one message, an attachment is updated in another, and the final approval is a two-word reply several messages later. Finding the current state requires reading history rather than looking at a defined record.

WhatsApp is fast but hard to govern

Messages are useful for conversation. They are less useful as the authoritative record of who approved which version of a request, particularly when the conversation includes unrelated messages.

Spreadsheets show status, but not always the evidence

A row can say “approved” without capturing who approved it, when it happened, what information they saw or whether the request changed afterward.

Verbal approvals depend on memory

A quick conversation may be entirely appropriate for a low-risk decision. It becomes a problem when the business later needs a repeatable record or when the next person cannot tell whether the approval actually happened.

What a workflow system should provide

Defined stages

Every request has a current state. The state should reflect real work: submitted, under review, awaiting approval, approved, in progress, completed or rejected, for example.

Permissions

Not everyone should be able to perform every action. The requester may edit before submission, an approver may approve or reject, and the production team may only act once approval is complete.

Status tracking and timestamps

The system should make the current position obvious and record important transitions automatically.

Audit trail

A useful audit trail captures actions that matter: submission, changes, approval, rejection, reassignment and completion. It should support the process rather than bury users in technical logs.

Notifications and escalation

The right person can be notified when action is required. Escalation can remind or redirect work when a request sits too long, but it should be used carefully so the system does not become another noisy inbox.

Map the workflow before automating it

1. Define the trigger. What creates a request and what information is mandatory at that point?
2. Define the reviewers. Who checks completeness or correctness before approval?
3. Define approval authority. Does one person approve everything, or do value, department or request type change the approver?
4. Define the action after approval. What work starts, and what information must be handed over?
5. Define the finish. Who is notified and what counts as completed?
6. Define reporting. What does management genuinely need to know about volume, turnaround, bottlenecks or outstanding work?

A simple practical example

Imagine a marketing material request. A branch submits the item, quantity, deadline and artwork details. The brand team reviews the request for completeness. A budget owner approves it. Production receives an automatically prepared job record. The requester is notified when the job moves forward and again when it is complete. Management can see outstanding requests and turnaround without asking each person for an update.

The important part is not that this process has six screens. It may need only a few. The value is that the request, decision and handover all refer to the same record.

Keep human judgement where it belongs

An approval is often a judgement call. Automation should present the right information, enforce required steps and record the outcome. It should not pretend that every decision can be reduced to a rule.

Automate incrementally

You do not need to replace the entire process at once. A sensible first version might centralise requests and approvals while leaving downstream production unchanged. Once that works reliably, notifications, reporting or integrations can be added.

If your approval process currently lives across spreadsheets and messages, these operational warning signs may feel familiar. For the wider approach, see business process automation in practical terms.

For an example of UNKNOT’s task-specific approach, see MergeForge and the broader workflow automation services.