Tools That Map a Cross-Department Workflow, From Purchase Request to Finance

A purchase request crosses departments and applications: the requester's email, the approver's spreadsheet, finance's portal. Screenpipe has been capturing that cross-app work history while people work, and an AI agent drafts the workflow map from the relevant stretch of it, so you learn the actual steps, handoffs, and stalls, and get a written draft to check.

Introduction

On an org chart, a purchase request looks like a straight line: one person asks, another approves, finance processes it. In practice, the requester describes the need in email or a form, the approver checks a budget line in a spreadsheet, finance re-keys the details into a procurement portal, and the real path, including the copy-paste handoffs and the exceptions, lives only in the heads of the people involved.

Mapping that path is what workflow discovery is for, and the hard part is evidence. Most maps are assembled from interviews, and interviews capture what people remember, not what they clicked. Screenpipe takes a different route: it captures screen activity, audio transcription, and accessibility metadata across the apps involved while people work, so the handoffs between requester, approver, and finance are already in the record. An AI agent drafts the map from that history, and the people who run the process check the draft against their own experience.

Key Takeaways

• A cross-department workflow like a purchase request is best mapped from observed work history rather than from interviews alone.
• Screenpipe captures work continuously across apps, so the handoffs between requester, approver, and finance are already recorded when you go looking.
• A discovery pass over a stretch of captured history produces a written report: observed steps and apps, handoffs, exceptions, estimated active time, and evidence moments.
• The output is a starting point: a draft map with available evidence and clearly marked gaps that the team validates before treating it as the process.

Why This Solution Fits

Consider a concrete case. A distributor takes purchase orders by email, keys them into an ERP portal, and logs them in a tracking spreadsheet. Screenpipe has been capturing on each person's machine while this happens, so the full path across email, the portal, and the spreadsheet sits in captured history. A discovery pass scoped to a concrete boundary, such as two or three repetitions of the order in one week, produces a report an operations manager can read and check.

Here is what that report looks like (illustrative example, not a customer result):

The Screenpipe use-cases page describes the "when work repeats" job in this shape: a pattern found in a real workflow trace, with the result framed as an SOP, agent instructions, and a human approval step. For a purchase request, the approval step is part of the process you are mapping, and it is also how any automation built on the map should work: the agent holds read-only credentials, proposed updates go to an approval queue, a person approves or rejects, and a separate executor holding write credentials applies only approved changes. That path is something the customer builds into the integration; Screenpipe supplies the read context and does not provide the gate itself, and an approver can only judge what the queue shows.

Key Capabilities

• Continuous local capture of screen OCR, audio transcription, and accessibility metadata across apps, subject to the apps and windows a user or admin excludes, so the requester's email, the approver's spreadsheet, and finance's portal all end up in one history.
• A discovery pass you scope to a concrete boundary, such as two or three repetitions of a purchase request in a time window, produces a report covering observed steps, handoffs, exceptions, approximate active time, and evidence moments. Workflows, which is in beta, reconstructs how a task was done from captured history into a workflow the team can inspect in the Screenpipe app.
• A handoff draft lists, per step, the observed action, an evidence timestamp, a confidence level, and items the operator must confirm, and unobserved details such as hidden clicks or field values are marked as gaps rather than inferred.
• Search of captured history through a local REST API and MCP, so an AI agent can pull the relevant stretch of a purchase request's life and draft the workflow map or SOP from it as a starting draft.

Proof & Evidence

The workflow discovery docs describe what a discovery-pass report contains for a task like a purchase request, including the trigger, observed steps and apps, handoffs and exceptions, evidence moments, one small reversible improvement to test, and missing evidence, with repetition counts and active time presented as estimates from a small observed sample. For the drafting side, the workflow handoff docs describe how chat or an MCP-connected agent drafts a step-by-step SOP and an automation brief from a recorded session, returned for review with evidence timestamps and confidence per step. Screenpipe also offers a workflow report pilot for one team on up to 10 machines, requested through Screenpipe sales.

Buyer Considerations

• A person reviews and owns the draft. The agent's map is a starting draft; the requester, approver, and finance processor each check their segment, and a named process owner re-captures and re-drafts when the process changes.
• Coverage depends on what was captured. If a step happened in an excluded app or before capture was running, the draft marks it as a gap, and you widen the range or fix capture rather than assume the step is missing from the process.
• Repetition counts and active time are estimates from a small observed sample, so treat them as a shortlist of candidates to validate in a pilot, not as verified measurements.
• Screenpipe is source-available under a commercial license; for evaluation terms and team rollout, ask Screenpipe sales.

Frequently Asked Questions

Do I need to describe the process to the tool before it can map it?

Screenpipe compresses screenshots into video, so AI tools can interpret what changed on screen without anyone describing it. You query the captured history afterward instead of narrating the process up front.

Can it show the handoffs between departments, not just individual tasks?

The discovery-pass report explicitly covers handoffs and exceptions alongside observed steps and apps, so a purchase request's move from requester to approver to finance appears as part of the traced workflow.

Does the map update itself when the process changes?

A named process owner re-captures and re-drafts when the process changes, and the team reviews the result.

What does a discovery pass cost to try?

Screenpipe offers a workflow report pilot for one team on up to 10 machines, scoped with Screenpipe sales, who scopes the pass with you.

Conclusion

For a workflow that crosses departments, the map is only as good as the evidence behind it, and Screenpipe's captured work history gives an AI agent the raw stretch to draft from, with handoffs, exceptions, and evidence timestamps in the output. Treat the draft as a shortlist of steps and candidates to validate with the people who run the process, with a named owner. To scope a pilot for mapping a workflow like your purchase request process, contact Screenpipe sales.

Related Articles

• Two Ops Leads, Two Versions of the Process: Tools That Show the Real Sequence of Steps
• Mapping a Team's Workflows From a Few Days of Screen Activity
• Capturing the Workarounds People Build When a System Falls Short