Skip to content

Documented case study · Construction · Mexico

From 2 weeks
to 5 minutes.

How we took supplier document validation at a construction company from a manual process that could not be sustained to an operation run by AI, rules and people — without hiring anyone.

Savings equivalent to 10 full-time employees in a single operation.

PRESENTED AT AI SUMMIT 2026 WITHClaro EmpresasHITSS

Before · manual validation

79 suppliers. 750 documents a week. A team at its limit.

Just 13 active suppliers generated 750 documents a week. Each one had to be opened and checked: was it the right document, was it current, did its data match the supplier’s onboarding record? Then someone had to check the official lists. All by hand.

REPSESIROCSAT tax compliance opinionLiability insurance policyTax status certificatePayroll CFDI
Illustration: an overwhelmed analyst reviewing pending and expired documents

325 h

per week just to validate documents by hand

25 hours per active supplier

10.8

full-time employees required

Impossible to sustain

15 days

to respond to a supplier

It held up onboarding, contracts and construction work

See it in action

The same week, on the platform.

Every document goes through five validations before a person touches it. What complies is approved; what does not is rejected or requested again from the supplier; anything doubtful goes to human review.

ONE WEEK OF OPERATIONS · IN 25 SECONDS

13 active suppliers · 750 documents

ARRIVING THROUGH THE PORTAL

0 / 750

Waiting for documents…

▪ AI READING▪ CORRECT TYPE▪ VALID PERIOD▪ INTERNAL CROSS-CHECKS▪ OFFICIAL LISTS

By hand

0 h

≈ 2 documents per hour · 10.8 full-time employees

With Appropia

0 h

50 documents per hour · half of one person’s time

Simulation using the volumes and times measured in the case. The documents shown are examples of the kinds of decisions the engine makes.

The validation engine

It does not just receive files. It reads and validates them.

STEP 0

AI reading

Extracts the key fields from the PDF or image: type, folio number, dates, amounts, period.

STEP 1

Correct type

Detects whether it is a REPSE, a SIROC, a tax status certificate, a liability policy…

STEP 2

Valid period

Checks that it covers the right period and has not expired.

STEP 3

Internal cross-checks

Tax data from supplier onboarding against the document, and across documents.

STEP 4

Official lists

Checks SAT, OFAC and PEP lists in real time.

It understands meaning, not just format. A negative tax compliance opinion with outstanding tax debts is rejected automatically, without ever reaching the administrator’s inbox.

Who decides, document by document.

MODE 3 · HYBRID

Automatic when there is certainty; doubtful cases go to human review.

The organization chooses the mode for each document type, and each document is required at the right stage of the contract.

EACH DOCUMENT AT ITS STAGE

  1. 01

    Onboarding

  2. 02

    Contracting

  3. 03

    Execution

  4. 04

    Closeout and payments

Illustration: requirements configured by document and by stage

What we built

Three connected pieces. Five layers underneath.

Illustration: supplier portal
01

Supplier portal

Suppliers upload their information and see what is required, the status of each document and their compliance level — without calling anyone.

Illustration: document repository
02

Document repository

The core, built on Alfresco: it stores, indexes and versions every document, with end-to-end traceability.

Illustration: administrator platform
03

Administrator platform

Configure rules, manage projects and sub-projects, approve or reject, and see the full compliance matrix.

01 · DOCUMENT REPOSITORY

A single repository for all supplier documentation

02 · PROCESSES AND WORKFLOWS

Rules by document and by contract stage

03 · IDP · CORTEX

Field reading and extraction with Cortex

04 · INTEGRATED AI AGENTS

Automatic validation and checks against official lists

05 · AGENTICOS · CROSS-CUTTING

Only doubtful cases reach the human inbox

The images are conceptual illustrations of the solution, not screenshots of client data.

Measured results

From impossible to a part-time job.

95.4%

Less time

From 325 to 15 hours per week. Operations stopped being the bottleneck.

25×

Faster per document

From about 2 to 50 documents per hour, without lowering the quality of the validation.

10.3

People freed up

From 10.8 full-time employees required to half of one person’s time (0.5) with the platform.

Same day

Response time

From 15 days of backlog to same-day answers. Projects no longer stall.

79 REGISTERED SUPPLIERS · MEASURED ON A SAMPLE OF 13 ACTIVE SUPPLIERS (750 DOCUMENTS PER WEEK)

The case behind an alliance

What worked for a construction company works for any case file.

We presented this case at AI Summit 2026 with Claro Empresas and HITSS. It is the basis of our commercial alliance: bringing this model to large enterprises and the public sector across the region.

Suppliers, contracts, employee files, customer claims or regulatory filings: the pattern is the same. Documents come in, rules must be met, and today people validate them by hand.

Appropia at AI Summit 2026 with Claro Empresas and HITSS

DOES YOUR OPERATION LOOK LIKE THIS?

  • Hundreds or thousands of documents a month
  • People validating by hand against known rules
  • Answers that take days and slow operations down
WhatsApp
WhatsAppTalk to our team