Workflow per pilot
No platform project and no unnecessary migration.
Process automationGet time back
We do not set out to automate an entire back office in one go. We select one recurring workflow, make every handover visible and automate only the steps governed by clear rules.
One workflow.
One accountable owner.
One clear boundary.
Phases to a decision
Understand, build, run in parallel, then decide.
Owner for exceptions
Unclear cases do not disappear into an invisible queue.
Turning a bottleneck into a pilot you can evaluate
The first test starts with a real case from your day-to-day operations, not with a tool. Each phase produces a tangible result.
- 1
You show us the workflow
We review real cases, variations, waiting times and manual handovers. No slide decks and no invented ideal process.
- 2
We build that exact workflow
Existing systems stay in place if they do their job. We connect only what the selected next step requires.
- 3
Run both in parallel for two weeks
Manual handling and automation run side by side for a direct comparison. Errors, queries and rework are reviewed every day.
- 4
You decide using evidence
Adopt, refine or stop. The test remains small enough to avoid unnecessary follow-on costs.
Automation should keep responsibility visible
A reliable workflow separates preparation from decision-making. The system handles repeatable work, while your team retains control of prices, approvals and difficult exceptions.
Automation
May
- Identify and route incoming items
- Structure approved fields
- Check duplicates and required information
- Prepare a draft for approval
Team
Decides
- Resolve conflicting prices
- Clarify sensitive or incomplete cases
- Approve binding content
- Change the rules and own ongoing operations
A sample process, visible step by step
This example follows an incoming enquiry through to approval. It contains no customer or production data and does not imply automatic sending.
Intake
A document or email arrives and receives a unique case number.
System: intake loggedCapture
Customer data and required fields are structured using the approved sources.
Check: customer number missing?Reconcile
Rules, duplicates and conflicting information are checked before the case proceeds.
Stop: two prices conflictPrepare
The system provides a draft with its sources and open points for review.
Status: awaiting approvalApprove
An accountable person reviews the content and decides on the next step.
Owner: Sales OperationsExceptions show whether the process works in practice
Happy paths are straightforward. Live operations depend on how missing data, duplicate cases and expired access are handled, and who owns the decision.
TimeEventResponseOwner
09:31:06Document detectedFields extractedSystem
09:31:07Price conflictWorkflow pausedSales Ops
09:31:08Notification sentNo automatic sendingTeam
Done means: The issue is visible, contained and assigned to an accountable person. Retry limits, timeouts and escalation are defined before go-live.
Clarify the workflow first, then connect the systems
Which source is authoritative?
CRM, ERP, inbox or spreadsheet: every field needs one current, authoritative data source.
What may the system do?
Reading, writing, preparing and sending require different permissions. These are deliberately limited during the pilot.
Who resolves conflicts?
An exception path needs a named owner, a deadline and a status. Otherwise, the automation simply creates another queue.
How does the process remain traceable?
Sources, changes and approvals are logged so the team can explain the case later.
What financial difference can a clearly defined process make?
Move the sliders using conservative assumptions. The open formula deducts setup and ongoing costs. It is a planning aid, not a forecast.
Illustrative net labour value in the first year
A negative result is valid. Under these assumptions, the selected workflow does not pay for itself.
When process automation makes sense
A strong candidate occurs frequently, follows repeatable rules and visibly consumes time today through searching, rekeying or chasing. It also needs a clear human review point for exceptions.
If the objectives, data sources or responsibilities are still unclear, we do not start with technology. A prioritised AI consultation is the better first step. If your main concern is availability and call handling, the AI phone assistant for Swiss SMEs is a better fit.
Questions before the first connection
Clear answers about software, outages and how value is calculated.
Does this require new software?
Not necessarily. We first check which systems are already in place, which interfaces are available and where a clear manual review point would work better.
What happens if an interface fails?
The exception path is part of the process design: log the status, limit retries and notify an accountable person. A silent failure is not an acceptable operating state.
How is the value calculated?
We use your volume, actual handling time, automatable share, setup costs and operating costs. The formula remains visible and can be adjusted conservatively.
From the AlpenAgent blog
Two useful deeper reads
Next step
Check a real case before it becomes a project
Bring one typical process and one difficult exception. In the call, we clarify which data are missing, where a person must decide and whether a pilot makes sense at all.