How one of the world's most iconic beverage companies transformed its Order-to-Cash process — freeing working capital and giving its finance team back time for the work that matters.
The Challenge: Where Cash Slows Down
Campari Group's Order-to-Cash (O2C) process spans order capture, credit checks, fulfillment, invoicing, collections, and cash application. Each handoff is a chance for an exception — a mismatched PO, a short payment, a customer dispute, a credit hold.
At enterprise scale, exception volume becomes the bottleneck. Even a small exception rate across tens of thousands of monthly orders means hundreds of transactions in flight at any moment — each one quietly delaying cash.
The Problem: Human Experts Chasing Exceptions
An audit of Campari's finance operations revealed something familiar to most CFOs: a significant portion of the finance team's time was spent on pattern-matching work — reading emails, checking SAP notes, cross-referencing CRM records, and deciding what to do next.
The context needed to resolve each exception was scattered: a customer email explaining a delayed payment, a handwritten note in SAP from a credit analyst, a CRM record showing a recent contract renegotiation, a shipping confirmation buried in a logistics system.
Context was visible to humans, but invisible to systems. That's why automation projects had stalled — the rules-based tools couldn't see what the analysts saw.
The Solution: AI Agents That Read Context
Campari deployed specialized AI agents that read context the way humans do — across email, ERP, CRM, and document attachments — and then act.
Agent 1: Exception Detection and Context Reading
The first agent monitors incoming transactions and flags exceptions with full context attached. When a key customer's payment was late, the agent surfaced an email from the customer's CFO announcing a merger, recognized it as a temporary liquidity event rather than a credit deterioration signal, and routed the case with a recommended temporary term extension — instead of a collections escalation.
Agent 2: Payment Block Resolution
The second agent investigates and resolves payment blocks autonomously. It reviews payment history, current exposure, and open contract negotiations, then proposes a resolution to the credit manager — who can approve it in minutes rather than reviewing from scratch. Payment block resolution dropped from days to hours.
Agent 3: Invoice Matching and Reconciliation
The third agent reads remittance advices (including PDFs and free-text emails), matches payments to invoices, and posts the cash. Invoice matching cycle time compressed dramatically — from a multi-day process to same-day.
The Results
After deploying AI agents on cash application and exception resolution, Campari saw a significant increase in first-pass match rate, dramatically faster invoice matching, a substantially reduced dispute backlog, and meaningful improvement in Days Sales Outstanding — all of which translated into real working capital freed without changing payment terms or customer behavior.
The finance team's time on mechanical exception work fell significantly, freeing capacity for the strategic finance work — credit policy refinement, customer profitability analysis, business partnering — that actually requires their expertise.
"The Collaboration Agent has significantly enhanced our operational efficiency and improved transparency across our order-to-cash process. What used to take days now happens in hours." — Laura Buseghin, Process Optimization & Automation Director, Campari Group
How This Translates to Your Business
The Campari pattern is repeatable. If a meaningful share of your finance team's time is on exception work and AI agents can significantly reduce that, you redeploy substantial capacity toward analysis, vendor negotiation, and strategic finance — work that pays for itself many times over.
Typical timeline:
- Weeks 1–4: Connect data sources, shadow-mode pilot on one workflow
- Weeks 5–8: Go live on remittance matching and exception triage
- Weeks 9–12: Expand to payment blocks and collections; measure DSO impact
The Blueprint
- Start small. One workflow, one geography, one team — prove the model before expanding.
- Measure everything. Baseline DSO, exception rates, and resolution times before deployment, not after.
- Focus on context. The win is in the email-plus-ERP-plus-CRM read, not in any single system.
- Redeploy, don't reduce. Move people to higher-judgment work — that's how you sustain organizational support.
- Scale after proof. Use the first-quarter numbers to fund the next four workflows.
Frequently Asked Questions
Q: Does this approach require replacing Campari's existing ERP or CRM systems? No. The AI agents operate on top of existing systems — reading from and writing to SAP and their CRM without replacing them. The orchestration layer connects to systems of record via governed APIs, so there's no rip-and-replace, no data migration, and no disruption to existing workflows during deployment.
Q: How did Campari handle the change management of introducing AI agents to a finance team? The deployment started in shadow mode: agents proposed actions, finance team members reviewed and confirmed them. This built trust in the agent's decisions before autonomous execution began, and gave the team visibility into exactly what the agent was doing and why. The result was that the team saw the AI as a tool that handled mechanical work — freeing them for the cases that genuinely required their expertise.
Q: What was the biggest obstacle in the deployment, and how was it resolved? The core challenge was that context needed to resolve exceptions was scattered across multiple systems — email, SAP notes, CRM records, shipping confirmations — and invisible to any single automation tool. The solution was connecting the orchestration layer to all of those sources simultaneously, giving the agent the same visibility a senior analyst would have when reviewing a case. Once the context problem was solved, resolution rates followed.
Q: How long did it take to see results after deployment? The initial pilot focused on a single workflow — cash application and exception triage — and ran in shadow mode for several weeks before switching to autonomous execution. Measurable improvements in resolution cycle time appeared in the first full month of autonomous operation. The expanded scope (payment blocks, dispute resolution) followed in subsequent quarters using the same deployment playbook.
See how AI agents transform Order-to-Cash → or read the full end-to-end O2C breakdown →.
