vectorfy

WORK / TOOLING IN DEVELOPMENT

Autoflow. Evidence from real browser work, turned into an automation plan.

Autoflow records authorised work in selected browser tabs, matches each click and keystroke with the network calls and screenshots behind it, and produces a plan an engineer can review before anything is automated.

Autoflow: capture the process, map the steps and review an automation plan.
WHAT IT IS
Workflow mining for portal work
WHAT IT IS NOT
Not staff monitoring, not a screen recorder, not an autonomous browser agent
OUTPUT
A reviewed automation plan

WHY IT EXISTS

Portal work is hard to automate because nobody has written it down.

Many organisations still run on dozens of third-party portals. Staff open an email, log in, copy details into an internal system, upload documents and mark the source item complete, all day. The logic lives in habit, hidden API calls and undocumented forms. Automating it has meant an engineer sitting beside the operator, guessing from browser developer tools and writing fragile scripts.

WHAT IT CAPTURES

Evidence, not just a recording.

Screenshots alone cannot prove how a workflow works. Autoflow records the structured evidence around each action so mappings can be proven rather than guessed.

Actions and fields

Clicks, focus, input and submits, with the labels, form context and stable selectors around them.

Network and API calls

The requests and responses behind each action, with timing and the values that mattered, redacted by policy.

Copy and paste lineage

Where a value came from and where it went, so the movement of data between systems is recorded, not inferred.

Screenshots in context

Before and after important steps, linked to the action and the request they belong to.

Illustrated product view. Captured workflow showing actions, source fields and a reviewed automation plan. Example data.

Illustrated product view. Example data.

FROM EVIDENCE TO AUTOMATION

Deterministic first. AI second. Review before anything runs.

  1. 01

    Capture

    An authorised operator does the work normally in selected tabs, with recording clearly visible.

  2. 02

    Correlate

    Actions, requests, clipboard events and screenshots are matched into one timeline and a graph of how values move between systems.

  3. 03

    Mine the workflow

    Repeated steps, field mappings, transformations and success markers are found by exact matching before any AI model is involved.

  4. 04

    Propose the plan

    A model receives a compact evidence bundle and drafts a neutral specification: sources, targets, steps, validations, confidence per mapping and open questions.

  5. 05

    Review and build

    An engineer reviews the plan. It is then compiled into the safest option, direct API calls, browser automation or a hybrid, with monitoring and a human fallback.

BUILT FOR SAFE CAPTURE

Consent, minimal capture and audit are part of the product.

Authorised tabs only

Capture runs only in selected tabs under an agreed policy. It never records general browsing.

Visible to the operator

Recording state is always shown, with pause and stop in reach.

Redaction and retention

Secrets and sensitive values are redacted or hashed by policy, with retention rules set per project.

Audit trail

Every event carries its source, policy and redaction state, so the evidence can be trusted later.

CURRENT STATUS

In development, used on Vectorfy projects first.

Autoflow is being built as tooling for Vectorfy integration work: a browser extension, an ingestion service, an evidence viewer and the mining pipeline. If your team copies work between browser-based systems, contact us about an early pilot.

Ask about an Autoflow pilot

Illustrated product view. Example data.

Next project: Conversation Insights