Today we showed the latest AI Workforce release, built for the part of sourcing nobody likes to talk about: the cleanup before the work begins.
Every sourcing project starts with a file. A BOM with merged cells, a stack of supplier PDFs, an export from a system nobody has opened in years. Before a single RFQ goes out, someone on the team spends days fixing part numbers, matching specs and removing duplicates. That work is necessary, but it doesn’t help anyone source better. It only delays the start.
Watch the session: How AI Workforce Cleans Up Your Messy Data (Youtube)
What Changed
AI Workforce runs across five stages of the sourcing cycle, each one built to execute work, not just suggest it.
Any file becomes items you can source. Hand AI Workforce a PDF, a spreadsheet or a legacy export. It reads the parts, specs and quantities and structures them into clean items. No manual mapping, no prep week.
Duplicates get caught on the way in. The same part entered three ways, conflicting specs, missing fields. AI Workforce flags them as the data arrives, so cleanup happens inside the project instead of before it.
Your data keeps getting better. Item data doesn’t stay clean after one pass. New quotes, revisions and supplier updates keep arriving. AI Workforce keeps cleaning and enriching item data over time, so each project starts from better information than the last.
What We Showed Live
In the session, Peter showed how AI Workforce turns messy files into data a sourcing team can use, live on screen.
It starts with an AI agent that extracts item data from whatever files you have, such as spreadsheets or supplier PDFs or old system exports. What makes that possible is the procurement intelligence built into the agent. It reads a file the way an experienced buyer would, so it recognizes part numbers, specs, quantities and units even when every file is formatted differently. It also understands each customer’s own categories, so it knows which properties matter for each type of part and captures them in real time as the data comes in.
From there, Peter walked through three steps:
Cleanup. Inconsistent descriptions, missing fields and mismatched formats become clean, structured records ready for sourcing.
Deduplication. The same part entered more than once under different names or numbers gets flagged, so the team works from a single, accurate list.
Enrichment. Missing attributes and details are filled in, so each item carries what buyers need before an RFQ goes out.
A human stays in the loop throughout, reviewing and approving the changes, so nothing reaches a sourcing event that the team hasn’t checked.
Idan explained why this only works when the AI understands your item data the way your team does, drawing on your own supplier history and part data instead of starting from scratch.
Why It Matters
Messy data is where sourcing time disappears. When item data is inconsistent, every step after it slows down. RFQs go out late, quotes come back in formats that don’t match, and comparing suppliers becomes another spreadsheet project.
Teams on LightSource already see what happens when that friction goes away. Infinite Machine reports getting to market 3x faster, with a 25% lower BOM cost and $10M saved in nine months. Aramsco has seen seven figures in savings. Arxada now runs quarterly RFQs across tail spend with many vendors, a process that used to be impractical.
What Gets Freed Up
This isn’t about replacing anyone. When nobody spends the first week of a project reformatting files, buyers get that time back for supplier relationships, negotiations and the decisions that move cost and schedule.
Watch the Full Demo
Want to see it working on real program files? The full session recording is available now. Idan Mintz, our CTO and cofounder, covers how the ingestion engine reasons about part information. Peter DiFalco, our Lead Solution Engineer, runs the live walkthrough and answers questions from the audience. Unscripted, start to finish.
Next Step
Have a file nobody wants to clean up? Send it to us. We’ll run it through AI Workforce and walk you through the result on your own data, not an example.
Request a demo or explore the platform
If you lead sourcing for automotive, electronics, space, or data infrastructure, the window to make an OS decision instead of accumulating point tools is now.
Today we showed the latest AI Workforce release, built for the part of sourcing nobody likes to talk about: the cleanup before the work begins.
Every sourcing project starts with a file. A BOM with merged cells, a stack of supplier PDFs, an export from a system nobody has opened in years. Before a single RFQ goes out, someone on the team spends days fixing part numbers, matching specs and removing duplicates. That work is necessary, but it doesn’t help anyone source better. It only delays the start.
Watch the session: How AI Workforce Cleans Up Your Messy Data (Youtube)
What Changed
AI Workforce runs across five stages of the sourcing cycle, each one built to execute work, not just suggest it.
Any file becomes items you can source. Hand AI Workforce a PDF, a spreadsheet or a legacy export. It reads the parts, specs and quantities and structures them into clean items. No manual mapping, no prep week.
Duplicates get caught on the way in. The same part entered three ways, conflicting specs, missing fields. AI Workforce flags them as the data arrives, so cleanup happens inside the project instead of before it.
Your data keeps getting better. Item data doesn’t stay clean after one pass. New quotes, revisions and supplier updates keep arriving. AI Workforce keeps cleaning and enriching item data over time, so each project starts from better information than the last.
What We Showed Live
In the session, Peter showed how AI Workforce turns messy files into data a sourcing team can use, live on screen.
It starts with an AI agent that extracts item data from whatever files you have, such as spreadsheets or supplier PDFs or old system exports. What makes that possible is the procurement intelligence built into the agent. It reads a file the way an experienced buyer would, so it recognizes part numbers, specs, quantities and units even when every file is formatted differently. It also understands each customer’s own categories, so it knows which properties matter for each type of part and captures them in real time as the data comes in.
From there, Peter walked through three steps:
Cleanup. Inconsistent descriptions, missing fields and mismatched formats become clean, structured records ready for sourcing.
Deduplication. The same part entered more than once under different names or numbers gets flagged, so the team works from a single, accurate list.
Enrichment. Missing attributes and details are filled in, so each item carries what buyers need before an RFQ goes out.
A human stays in the loop throughout, reviewing and approving the changes, so nothing reaches a sourcing event that the team hasn’t checked.
Idan explained why this only works when the AI understands your item data the way your team does, drawing on your own supplier history and part data instead of starting from scratch.
Why It Matters
Messy data is where sourcing time disappears. When item data is inconsistent, every step after it slows down. RFQs go out late, quotes come back in formats that don’t match, and comparing suppliers becomes another spreadsheet project.
Teams on LightSource already see what happens when that friction goes away. Infinite Machine reports getting to market 3x faster, with a 25% lower BOM cost and $10M saved in nine months. Aramsco has seen seven figures in savings. Arxada now runs quarterly RFQs across tail spend with many vendors, a process that used to be impractical.
What Gets Freed Up
This isn’t about replacing anyone. When nobody spends the first week of a project reformatting files, buyers get that time back for supplier relationships, negotiations and the decisions that move cost and schedule.
Watch the Full Demo
Want to see it working on real program files? The full session recording is available now. Idan Mintz, our CTO and cofounder, covers how the ingestion engine reasons about part information. Peter DiFalco, our Lead Solution Engineer, runs the live walkthrough and answers questions from the audience. Unscripted, start to finish.
Next Step
Have a file nobody wants to clean up? Send it to us. We’ll run it through AI Workforce and walk you through the result on your own data, not an example.
Request a demo or explore the platform
If you lead sourcing for automotive, electronics, space, or data infrastructure, the window to make an OS decision instead of accumulating point tools is now.
Today we showed the latest AI Workforce release, built for the part of sourcing nobody likes to talk about: the cleanup before the work begins.
Every sourcing project starts with a file. A BOM with merged cells, a stack of supplier PDFs, an export from a system nobody has opened in years. Before a single RFQ goes out, someone on the team spends days fixing part numbers, matching specs and removing duplicates. That work is necessary, but it doesn’t help anyone source better. It only delays the start.
Watch the session: How AI Workforce Cleans Up Your Messy Data (Youtube)
What Changed
AI Workforce runs across five stages of the sourcing cycle, each one built to execute work, not just suggest it.
Any file becomes items you can source. Hand AI Workforce a PDF, a spreadsheet or a legacy export. It reads the parts, specs and quantities and structures them into clean items. No manual mapping, no prep week.
Duplicates get caught on the way in. The same part entered three ways, conflicting specs, missing fields. AI Workforce flags them as the data arrives, so cleanup happens inside the project instead of before it.
Your data keeps getting better. Item data doesn’t stay clean after one pass. New quotes, revisions and supplier updates keep arriving. AI Workforce keeps cleaning and enriching item data over time, so each project starts from better information than the last.
What We Showed Live
In the session, Peter showed how AI Workforce turns messy files into data a sourcing team can use, live on screen.
It starts with an AI agent that extracts item data from whatever files you have, such as spreadsheets or supplier PDFs or old system exports. What makes that possible is the procurement intelligence built into the agent. It reads a file the way an experienced buyer would, so it recognizes part numbers, specs, quantities and units even when every file is formatted differently. It also understands each customer’s own categories, so it knows which properties matter for each type of part and captures them in real time as the data comes in.
From there, Peter walked through three steps:
Cleanup. Inconsistent descriptions, missing fields and mismatched formats become clean, structured records ready for sourcing.
Deduplication. The same part entered more than once under different names or numbers gets flagged, so the team works from a single, accurate list.
Enrichment. Missing attributes and details are filled in, so each item carries what buyers need before an RFQ goes out.
A human stays in the loop throughout, reviewing and approving the changes, so nothing reaches a sourcing event that the team hasn’t checked.
Idan explained why this only works when the AI understands your item data the way your team does, drawing on your own supplier history and part data instead of starting from scratch.
Why It Matters
Messy data is where sourcing time disappears. When item data is inconsistent, every step after it slows down. RFQs go out late, quotes come back in formats that don’t match, and comparing suppliers becomes another spreadsheet project.
Teams on LightSource already see what happens when that friction goes away. Infinite Machine reports getting to market 3x faster, with a 25% lower BOM cost and $10M saved in nine months. Aramsco has seen seven figures in savings. Arxada now runs quarterly RFQs across tail spend with many vendors, a process that used to be impractical.
What Gets Freed Up
This isn’t about replacing anyone. When nobody spends the first week of a project reformatting files, buyers get that time back for supplier relationships, negotiations and the decisions that move cost and schedule.
Watch the Full Demo
Want to see it working on real program files? The full session recording is available now. Idan Mintz, our CTO and cofounder, covers how the ingestion engine reasons about part information. Peter DiFalco, our Lead Solution Engineer, runs the live walkthrough and answers questions from the audience. Unscripted, start to finish.
Next Step
Have a file nobody wants to clean up? Send it to us. We’ll run it through AI Workforce and walk you through the result on your own data, not an example.
Request a demo or explore the platform
If you lead sourcing for automotive, electronics, space, or data infrastructure, the window to make an OS decision instead of accumulating point tools is now.
Faster sourcing. Lower cost. Less chaos.
See how LightSource connects engineering, procurement, and suppliers in one operating system to help you launch faster at lower cost.
SOC 2
Kearney #1 2024
Gartner Cool Vendor
Procuretech 100
G2 Top Rated
Faster sourcing. Lower cost. Less chaos.
See how LightSource connects engineering, procurement, and suppliers in one operating system to help you launch faster at lower cost.
SOC 2
Kearney #1 2024
Gartner Cool Vendor
Procuretech 100
G2 Top Rated
Faster sourcing. Lower cost. Less chaos.
See how LightSource connects engineering, procurement, and suppliers in one operating system to help you launch faster at lower cost.
SOC 2
Kearney #1 2024
Gartner Cool Vendor
Procuretech 100
G2 Top Rated
Trusted by:
Trusted by:
Trusted by: