"We could handle ten invoices. Now it's thousands": iFood connected legacy systems with Make to keep supplier payments from stalling

iFood is a Brazilian technology company and a reference in delivery, running on a vast network of suppliers that provide the resources the business needs to operate. The majority of iFood's supplier invoices flow through a standardized portal process without issues.

But when an invoice falls outside that standard process (roughly 10% of the total volume), it gets blocked and requires manual intervention to resolve. That's where the Procurement Team has to find out who is responsible for the fix and contact them.

Until recently, that meant manually gathering data from four legacy systems (Databricks, Gmail, Jira, and Slack) just to identify the invoice requester. Then the team had to contact the requester manually to resolve the issue before it resulted in a late payment.

iFood already had an internal AI agent platform, Toqan, that could automate part of this process. On its own, Toqan handled a handful of cases well. As the volume grew, it made sense to pair Toqan with a dedicated automation platform to orchestrate the full workflow reliably.

Leandro Alvarenga, Procurement Analyst at iFood, solved this by combining Toqan with Make. He used Toqan for the parts where the AI excels – searching and extracting data – and Make to orchestrate the rest. Connecting systems, routing information, and handling the workflow end-to-end.

Together, the two platforms handle an unlimited number of invoices reliably.

As a result, the Procurement Team saves two and a half hours a day, removing what used to be a tedious, fully manual process. Keeping supplier payments moving on time.

"I'm genuinely excited about it. It's the first platform I've seen that can connect different systems and modules together to build a workflow like this. I think it's really exciting to work with."

The challenge: hours each day of Invoice troubleshooting

Manual search stalling invoice payments

Whenever an invoice ran into a problem, someone on the procurement team had to track it down by hand – searching iFood's data lake for the details, digging through Gmail for the PDF, then opening a Slack thread to tag whoever needed to fix it.

Each invoice issue took about 15 minutes to resolve in this way. An average of 10-15 invoice issues a day meant two and a half hours per day spent purely on manual, repetitive search. Resulting in stalled payments and unhappy suppliers.

Scaling AI beyond individual cases

At first, Leandro tried connecting iFood's AI agent platform Toqan directly to some of the systems he's using for invoice troubleshooting.

It worked for a handful of cases, but scaling up to thousands of invoices required a more structured approach. An AI agent on its own isn't designed to orchestrate multi-step workflows across multiple systems. That's where Make comes in.

"Toqan handles the intelligence part. Make handles the orchestration. Together they handle far more cases than either could alone, because combining the company's AI with a reliable automation platform gives you the best of both worlds."

The solution: processes made predictable by Make

Toqan feeds the data, Make does the rest

Leandro built a workflow where the Toqan AI agent searches the company's data lake for blocked invoice details and sends the results to Make over a webhook.

Make handles everything else.

It opens a Jira ticket, finds the invoice's PDF attachment in Gmail, and posts a Slack thread tagging the person who needs to make the fix, with the attachment already included.

There's no need to open the invoice, search for the attachment, or start the conversation manually anymore. Leandro simply joins a thread that's already been built for him, and does the final part of the job – specifying the required fix.

"I use Make to connect all the systems. I connect my AI agents and other systems to do this whole workflow, to help us handle invoice blockages and purchase order problems, and to notify the right person."

Consistent, structured messages at scale

By keeping Toqan focused on what it does best – searching and extracting data from the data lake it was already connected to – and letting Make orchestrate the rest, the workflow stays reliable at any scale.

Make also builds the Slack messages from a fixed structure every time, rather than asking the AI to generate them from scratch. This ensures consistency – nothing gets invented, and nothing changes no matter how many invoices come through.

It also keeps token usage low – since Make handles the message construction and orchestration, the AI only processes the data it needs to search and analyze. Not the entire workflow output. Less prompt volume per invoice means lower cost and faster execution, even at scale.

This enables the procurement team to run this workflow at scale.

The result: One clear process, no ceiling on volume

The team saves two and a half hours per day that were previously spent on manual, repetitive search.

But the most telling change is scale. What used to be a slow manual process now runs quickly and at scale. iFood can handle as many invoices as they need, with Toqan and Make working together to keep every step reliable.

What's more, every invoice that gets picked up and resolved faster means a supplier who isn't left waiting on a payment that stalled in someone else's inbox. Keeping business-critical relationships on solid ground.

Part of a broader shift: "What I've built shows what's possible"

Leandro's workflow didn't happen in isolation. AI experimentation runs close to the center of how iFood operates. Both Toqan and Claude are available to the whole company, and Leandro's workflow is turning into a reference point for how to use them together with automation platforms like Make.

The automation raised so much excitement that he's getting questions and interest from the wider company. Leandro is currently testing a second AI workflow – automatic responses to supplier queries.

"People across the company often ask me how to use AI with Make after seeing my workflows. I think what I've built is helping show others what's possible."

And so the connections spark momentum. One Procurement Analyst's automation became a spark for further change. And one more connected system inspires others at iFood to innovate.