AI Automation Document Intake Power Automate Salesforce

Syte Link: supplier
paperwork, run by AI.

Supplier waste transfer notes and invoices were found in inboxes, renamed, checked and uploaded to Salesforce by hand. We built the intake on the systems Syte Link already had. In October, AI uploaded 3,826 of them.

3,826 transfer notes uploaded by AI in October
90% with no manual upload
Jul → Sep kick-off to live
Live on Power Automate and Salesforce
Engagement Snapshot
ClientSyte Link
SectorSite services for construction: waste management, concrete, aggregates and welfare hire
TrackAI Automation
VolumeAround 500 documents at a time, three times a week
Outcome3,826 waste transfer notes uploaded by AI in October; 422 by hand
StatusLive since September 2026

Key Highlights

→ 90% of October’s waste transfer notes uploaded with no manual step
→ Built on the Power Automate and Salesforce that Syte Link already paid for, with no new platform
→ A dedicated inbox for supplier documents, with duplicate detection
→ The options costed in a report for the board before anything was built
→ From kick-off in July to live in September
Context & Challenge

Syte Link supplies construction companies with waste management, ready-mix concrete, aggregates and welfare hire, delivered through a network of supplier partners. Every job those suppliers do comes back as paperwork: a waste transfer note, an invoice, or both.

That paperwork was handled by hand. Documents were found in inboxes, renamed, checked and uploaded to Salesforce, around 500 at a time, three times a week.

The work followed clear steps and happened in volume, on systems that were staying. That made it a strong candidate for automation, provided it was built around those systems and kept a person on anything unusual.

Approach & Solution

From inbox to Salesforce, without the re-keying.

Hypothesis
Stage 01 — Hypothesis

Map the admin process and cost the options.

We mapped the admin process as the team really ran it, in Process Flow, then costed the options in a report for the board. The decision to build was made on that report.

Experiment
Stage 02 — Experiment

Prove it on real documents, beside the team.

The intake was run on Syte Link’s own supplier documents in timed batches, and each run was reviewed with the people who did the work by hand, so the unusual cases surfaced before anything depended on it.

Formulation
Stage 03 — Formulation

Build the intake on the systems already in place.

The production intake runs on Syte Link’s existing Microsoft Power Automate and Salesforce. Supplier documents arrive at a dedicated inbox, duplicates are detected, and the documents are uploaded to the right record.

✓Built on Power Automate and Salesforce, with no new platform to buy or learn
✓A dedicated inbox for supplier documents
✓Duplicate detection before upload
Execution
Stage 04 — Execution

Live, with the exceptions left to a person.

The intake went live in September. Documents that do not go through it are still uploaded by the admin team, so nothing is lost, and the share handled automatically is counted each month. Weekly check-ins ran for the first weeks after launch.

Results & Impact

What changed.

Before Ferrous Labs After Ferrous Labs
Supplier documents found in inboxes and renamed by hand Supplier documents arrive at one dedicated inbox
Each document checked and uploaded to Salesforce manually 3,826 waste transfer notes uploaded by AI in October
Around 500 documents at a time, three times a week 422 left for manual upload in the same month
Duplicates caught only if someone noticed Duplicates detected before upload
In October I had to upload 422 WTNs that didn’t go through the AI process. AI uploaded 3,826. That’s amazing, well done team.
Philippa Etheridge, Administration Assistant, Syte Link
Technology

Stack

Microsoft Power Automate Salesforce Process Flow
Paperwork your team re-keys every week?

Talk to a co-founder.

Scoping is free: about two hours with a co-founder, then a proposal with a price for each stage.

Ferrous Labs free scoping call. Scoping is free: a call of about two hours with a co-founder to understand the work, followed by a proposal with a price and timeline for each stage. There is no charge for either, and no commitment to go further. How to take it up: book the call, or start with the five-minute AI Opportunity Finder.