Case Studies

Automating procurement from request to approved vendor

Replacing fragmented forms, manual checks and repeated data entry with one guided workflow

Sector

FMCG and enterprise procurement

Client

Global consumer goods company

Challenge

Fragmented vendor onboarding and procurement requests created errors, delays and repeated manual work

Solution

Guided digital procurement workflow with conversational assistance, automated validation and multi-channel orchestration

/

/

Automating procurement from request to approved vendor

/

/

Automating procurement from request to approved vendor

01

The challenge

A simple vendor request had become a complicated process.

Creating a new approved vendor required employees, procurement teams, suppliers and controllers to complete a chain of steps.

The process depended heavily on complex Excel forms.

Those forms had limited validation.

Requestors often did not know which steps were required.

Supplier information was collected offline.

The same data could be entered more than once.

Errors were discovered late.

That created a low percentage of first-time-right requests and longer procurement lead times.

The result was more than administrative friction.

Delays in vendor setup could ultimately delay the products and services the business was trying to purchase.

02

What needed to change

The organisation did not simply need a digital version of the Excel form.

It needed the process itself to become easier to follow.

The system needed to know:

  • What stage the request was in

  • Who needed to act next

  • What information was still missing

  • Whether submitted data was valid

  • Which participant needed guidance

  • When a human needed to review something

Instead of making the requestor responsible for driving the whole process, the workflow itself needed to coordinate the participants.

03

What we built

A digital procurement process that guides every participant.

WeAreBrain developed an enterprise virtual assistant and workflow system that guided employees, procurement teams, vendors and controllers through the full vendor-request process.

The solution combined:

  • conversational guidance

  • structured online forms

  • workflow orchestration

  • data validation

  • supplier interaction

  • secure portal experiences

  • email

  • eSignature

  • existing enterprise identity

The interface changed depending on which interaction made the most sense.

Conversation was useful for guidance.

Forms were better for structured data.

Email was useful for communication.

A secure portal handled information that needed stronger access control.

1. Start the request

Make it obvious what the user needs to do.

Previously, the employee initiating a vendor request needed to understand the procurement process themselves.

The new system guided them through the required steps.

Instead of asking:

“Which form do I need, and what happens next?”

the user could begin the process through one guided experience.

The system then helped move the request forward.

2. Orchestrate the workflow

The process should remember what comes next.

The system tracked individual requests and coordinated each participant through the relevant stage of the process.

That meant the workflow itself could keep track of:

  • Requestor actions

  • Procurement tasks

  • Vendor input

  • Validation steps

  • Controller review

  • Outstanding information

The user no longer needed to carry the entire process in their head.

3. Guide suppliers through data collection

Supplier onboarding often creates friction because external vendors do not know the buyer's internal process.

The system helped vendors provide the information in the required format.

This reduced the reliance on offline forms and helped address problems earlier rather than waiting for procurement teams to identify incomplete information later.

4. Validate information

Catch problems before they move downstream.

The workflow included validation against external information sources such as:

  • DUNS

  • IBAN

This allowed submitted information to be checked as part of the process.

That matters because bad supplier data creates work for every team further down the chain.

A stronger process aims to make the request correct earlier.

5. Give controllers the right information

By the time information reached the control stage, it could be presented in a more consistent structure.

Controllers received the relevant data in the required format rather than having to reconstruct it from incomplete forms and email threads.

The objective was to increase the chance that a request could move through the process correctly the first time.

6. Add eSignature

The workflow also introduced electronic signatures.

This removed additional paper-based steps and helped vendors complete the process digitally.

It is a small example of an important automation principle:

Improving a workflow often means removing several small handovers rather than introducing one dramatic AI feature.

04

One process, several interfaces

Conversation was one part of the experience.

The original solution used a custom web-based chat interface running through the company's existing SharePoint environment. Active Directory was used to identify internal users.

But not every task belonged in chat.

Other steps used:

  • Online forms

  • Email

  • Secure web portals

  • eSignature

The system selected the appropriate interaction for each part of the workflow.

That design principle remains highly relevant today.

A useful AI workflow is rarely just a chatbot.

05

The workflow

Employee starts vendor request

System guides the required information

Procurement reviews and processes the request

Vendor provides structured data

Information is validated

Missing or incorrect information is resolved

Controllers receive the required data

Signatures are completed

Vendor request progresses towards approval

The important shift was from a user-driven process to a system-orchestrated process.

06

Human control remains part of the process

The system did not attempt to automate procurement decisions.

It automated coordination.

People remained responsible for:

  • Approval

  • Exceptions

  • Commercial judgement

  • Final checks

  • Supplier relationships

Automation handled:

  • Guidance

  • Status tracking

  • Routing

  • Validation

  • Reminders

  • Data collection

That separation is still a useful model for enterprise automation today.

07

The result

Faster processing and fewer avoidable mistakes.

The system launched in 2019.

Testing showed:

Faster processing

Requests moved through the process more efficiently.

Fewer errors

Better guidance and validation reduced avoidable mistakes.

Better participant experience

Employees, vendors and internal teams had a clearer path through the process.

The original case did not publish final realised cost savings, so those should not be claimed retrospectively.

08

Why this matters

Procurement automation is really workflow design.

It is easy to focus on the interface.

The more important problem is what happens behind it.

A procurement workflow can involve:

  • Employees

  • Suppliers

  • Procurement teams

  • Finance

  • Controllers

  • External data sources

  • Identity systems

  • Documents

  • Approvals

The value comes from coordinating those pieces.

That is why this project remains relevant even though the underlying AI technology has changed.

09

From workflow assistant to modern agentic automation

The original system was built before today's large language models and modern agent frameworks.

We should keep that distinction clear.

What the project already demonstrated was a pattern that today's agentic systems build on:

  • Understand where the process is

  • ↓

  • Identify the next required action

  • ↓

  • Guide or contact the right participant

  • ↓

  • Use tools and external systems

  • ↓

  • Validate the result

  • ↓

  • Escalate to a person where needed

Modern AI can make parts of that process more flexible.

The core workflow-design challenge remains the same.

10

How we might build it today

A current implementation could extend the same pattern with capabilities such as:

  • Unstructured document understanding

  • Natural-language supplier communication

  • Automatic extraction from certificates and forms

  • Policy-aware validation

  • ERP tool use

  • Exception summarisation

  • Supplier risk research

  • Procurement knowledge retrieval

  • Dynamic workflow routing

These are current possibilities, not claims about functionality delivered in the 2019 system.

AI services used

AI Agents

Designing a workflow where software coordinates tasks, participants and validation from beginning to end.

Software Engineering

Building the custom application and connecting it with enterprise identity, data and communication systems.

Product Design

Choosing the right interaction for each stage rather than forcing the entire workflow into one interface.

Avatar-1
Avatar-2
Avatar-4
Avatar-4
Avatar-3

Still coordinating procurement through forms, inboxes and manual checks?

The opportunity is often bigger than automating one task. We design workflows where systems, AI and people each handle the part they are best suited for.