→

Back

SUCCESS STORY · ASTORE (ACCOR GROUP)

Astore｜Product benchmarking industrialized on Snowflake

KPC industrialized Astore's product benchmarking on Snowflake. A scanned invoice becomes a costed recommendation in minutes, on real data rather than estimates.

Data

Snowflake

Gen AI

Data Products

Data Platform

Data Plan

AI Offer

Hospitality & Tourism

The impact in numbers

Days → min.

Time to run a product benchmark

750k

References in the Astore catalog

~20

Sales reps using the tool

−15%

Savings measured on a quote compared to the Astore catalog based on real cases.

The impact in numbers

Days → min.

Time to run a product benchmark

750k

References in the Astore catalog

~20

Sales reps using the tool

−15%

Savings measured on a quote compared to the Astore catalog based on real cases.

The impact in numbers

Days → min.

Time to run a product benchmark

750k

References in the Astore catalog

~20

Sales reps using the tool

−15%

Savings measured on a quote compared to the Astore catalog based on real cases.

RESULTS

The results for Astore

From days to minutes: the sales decision, accelerated

An invoice uploaded, a costed recommendation in minutes. The sales rep focuses on the client conversation. The Excel analysis drops out of the process.

The results for Astore

Recommendations on real data

The comparisons rest on real invoices and the Astore catalog, not on estimates. On a real case, the measured gap reached a 15% saving.

The results for Astore

AI in production, invisible

An application used daily by sales representatives, without technical skills. AI works behind the scenes.

RESULTS

The results for Astore

From days to minutes: the sales decision, accelerated

An invoice uploaded, a costed recommendation in minutes. The sales rep focuses on the client conversation. The Excel analysis drops out of the process.

The results for Astore

Recommendations on real data

The comparisons rest on real invoices and the Astore catalog, not on estimates. On a real case, the measured gap reached a 15% saving.

The results for Astore

AI in production, invisible

An application used daily by sales representatives, without technical skills. AI works behind the scenes.

RESULTS

The results for Astore

From days to minutes: the sales decision, accelerated

An invoice uploaded, a costed recommendation in minutes. The sales rep focuses on the client conversation. The Excel analysis drops out of the process.

The results for Astore

Recommendations on real data

The comparisons rest on real invoices and the Astore catalog, not on estimates. On a real case, the measured gap reached a 15% saving.

The results for Astore

AI in production, invisible

An application used daily by sales representatives, without technical skills. AI works behind the scenes.

Background & core challenges

A benchmark that is reliable, repeatable and fast

A purchasing group that has to prove its savings, invoice in hand

Astore, the Accor group's purchasing arm, pools volumes to offer hoteliers prices below those of their usual suppliers. To win over a hotelier who is not yet a client, its sales reps compare that hotelier's invoice with the Astore catalog, 750,000 references, and cost out the savings available.

That comparison was done by hand, in Excel, line by line: several hours, sometimes several days per request, with a result that depended on each person's experience.

Go from several days to a near-instant result per request

Make it usable by sales reps with no technical background, built into the sales cycle

Break the dependency on individual experience for consistent results

Base every recommendation on real data, not on estimates

Background & core challenges

A benchmark that is reliable, repeatable and fast

A purchasing group that has to prove its savings, invoice in hand

That comparison was done by hand, in Excel, line by line: several hours, sometimes several days per request, with a result that depended on each person's experience.

Go from several days to a near-instant result per request

Make it usable by sales reps with no technical background, built into the sales cycle

Break the dependency on individual experience for consistent results

Base every recommendation on real data, not on estimates

Background & core challenges

A benchmark that is reliable, repeatable and fast

A purchasing group that has to prove its savings, invoice in hand

That comparison was done by hand, in Excel, line by line: several hours, sometimes several days per request, with a result that depended on each person's experience.

Go from several days to a near-instant result per request

Make it usable by sales reps with no technical background, built into the sales cycle

Break the dependency on individual experience for consistent results

Base every recommendation on real data, not on estimates

Our approach

The whole chain inside Snowflake, with no data movement

From the raw document to the business deliverable, with no extraction off the platform. AI is placed only where it creates value: extraction, normalization, matching — and stays invisible to the end user.

AI extraction & reading

From the PDF invoice to usable data

The invoice is uploaded and read automatically by AI extraction. Product labels are normalized by an LLM so they can be compared with the catalog.

Cortex AI

AI Parse Document

LLM Complete

Matching & scoring

Catalog matching and prioritization

Each product is matched against the catalog by semantic search (Cortex Search), combined with SQL fuzzy matching and an LLM. The recommendations are then scored and prioritized automatically.

Cortex Search

SQL Fuzzy Matching

LLM Complete

Application & report

Business validation and PDF report

The sales rep validates the recommendations in a Streamlit application hosted inside Snowflake. They then produce a PDF report with price gaps and savings.

Streamlit (SPCS)

Snowflake

PDF report

Our approach

The whole chain inside Snowflake, with no data movement

AI extraction & reading

From the PDF invoice to usable data

The invoice is uploaded and read automatically by AI extraction. Product labels are normalized by an LLM so they can be compared with the catalog.

Cortex AI

AI Parse Document

LLM Complete

Matching & scoring

Catalog matching and prioritization

Each product is matched against the catalog by semantic search (Cortex Search), combined with SQL fuzzy matching and an LLM. The recommendations are then scored and prioritized automatically.

Cortex Search

SQL Fuzzy Matching

LLM Complete

Application & report

Business validation and PDF report

The sales rep validates the recommendations in a Streamlit application hosted inside Snowflake. They then produce a PDF report with price gaps and savings.

Streamlit (SPCS)

Snowflake

PDF report

Our approach

The whole chain inside Snowflake, with no data movement

AI extraction & reading

From the PDF invoice to usable data

The invoice is uploaded and read automatically by AI extraction. Product labels are normalized by an LLM so they can be compared with the catalog.

Cortex AI

AI Parse Document

LLM Complete

Matching & scoring

Catalog matching and prioritization

Each product is matched against the catalog by semantic search (Cortex Search), combined with SQL fuzzy matching and an LLM. The recommendations are then scored and prioritized automatically.

Cortex Search

SQL Fuzzy Matching

LLM Complete

Application & report

Business validation and PDF report

The sales rep validates the recommendations in a Streamlit application hosted inside Snowflake. They then produce a PDF report with price gaps and savings.

Streamlit (SPCS)

Snowflake

PDF report