ASSESSMENT AND IMPLEMENTATION

Business process automation, from the first step to the last.

Copying information between tools, chasing approvals and checking the same fields takes work. Stolen Orbit designs and implements connected workflows with explicit rules and exception handling your team can understand.

Let’s discuss your process

Start with a handoff that repeats.

An automated process connects an event to a sequence of actions: a request arrives, data is checked, a draft is prepared, a person approves and the agreed tools are updated. Evaluate the whole workflow, including the work people still need to do.

Scroll the table to compare all columns.

Examples to assess against your own process
ProcessPossible automationControl to retain
QuotesPrepare a reminder when a response is missing.Exclude closed opportunities, unsuitable contacts and requests not to receive messages.
OrdersCollect fields and prepare a draft in the business system.Approve quantities, product codes, prices and ambiguities.
SupportClassify a request and suggest the next step.Route complaints, missing information and out-of-scope cases to a person.
EcommerceTrigger sequences from agreed events and rules.Check sending conditions, content and recipient data.

Rules, integrations or AI: choose for the task.

Start with native features in existing tools. If they do not cover the process, assess connectors or a dedicated integration. AI may help interpret free text; deterministic checks remain useful for product codes, required fields or sending conditions.

Not every workflow needs an AI agent. When the steps are known, explicit rules make the flow easier to inspect. Compare volume, exceptions, available access and operating cost.

There is more to delivery than connecting two apps.

  1. A process map and explicit rules

    Define the trigger, fields, states, owners and permitted actions. Include waiting, rejection and missing information in the map.

  2. Integrations with limited access

    Use the access required for the flow and separate test environments where supported. Agree on ownership of credentials and accounts.

  3. Duplicate and interruption handling

    Decide how to recognise an event already processed, handle a failed attempt and resume without creating a second order or message. State any limitations imposed by the external system.

  4. Operational visibility

    The team should see what completed, what stopped and why. Agree on alerts, outcome records, a manual procedure and responsibility for updates.

Evaluate the workflow before expanding it.

A possible pilot covers one request type and one operational connection. Before introduction, test missing data, duplicates, revoked access and service unavailability. Early runs can produce drafts for review rather than final actions.

  • Measure total work per case before and after, including checking and correction.
  • Count completions, exceptions, errors and duplicates prevented or missed.
  • Distinguish system delay, approval waiting time and human effort.
  • Calculate recurring cost and the work needed to maintain the flow.
  • Expand only when the agreed criteria are met.

When the process needs work first.

Automating a step without an owner or shared rule can accelerate confusion. First decide who can approve, which data is authoritative and what happens when two systems disagree.

If volumes do not justify the project cost, a clearer procedure, a shared template or an existing product feature may be the right outcome.

FROM IDEAS TO A BRIEF

A template to work from.

First AI pilot brief

An editable Markdown template for your process, data, owners and go/no-go criteria.

Download the Markdown template

Practical questions

Are business process automation and AI the same thing?

No. Automation can follow predefined rules without AI. AI may add text interpretation or content preparation to part of a workflow. We choose where it belongs and how to verify its output.

What happens when an integration stops working?

The project should define detection, notification and recovery. Do not blindly repeat a failed action: first check whether the destination system executed it.

Which process should we start with?

Choose frequent work with accessible inputs, reasonably stable rules, an owner and a measurable result. The first-project guide helps compare candidates before choosing tools.

Which process should improve first?

Bring one specific task, the tools you use and the point where work gets stuck. We can define what is needed to assess a first project.

Let’s discuss your process