How an EU energy & utility giant built repeatable automation it can control

A complex corporate structure, operations in 30+ countries, and serving 50+ million customers

Our client is one of Europe’s largest energy and utility companies operating across the EU. Their business is genuinely complex: they sell energy, produce energy, sell goods, and manage internal transactions, all mixed together. This complexity shows up in their data structures, which vary significantly by company code and business function. No two regions execute processes the same way, and there is no simple answer to which way is "right."

Fragmentation and lack of visibility turned automation into guesswork

The overall nature of the business makes straight-through processing hard to achieve and maintain due to complex energy-utility invoices, high service contract volumes, multi jurisdictional VAT requirements, and a 50/50 PO/non-PO document split.

With 500+ company codes to manage, the same workflows are executed completely differently across regions. There was no documented visibility into where processes diverged or why.

Teams were stuck in manual work: extracting supplier names from documents, searching SAP for matching records, manually checking payment status all per document. The capabilities of older technology did not allow for context understanding in real time, so applying RPA-era rules and practices to a complex, non-standardized business structure, capped automation success.

Our client defined five specific performance KPIs upfront, with targetautomation rates-- including auto rejection--between75% and 95% depending on task complexity. Any solution would be held to these from day one.

Governance first, then iteration: scaling from one problem to many solutions

Instead of forcingour client's data into standard patterns, we co-designed a data governance structure based on several custom data structures specific tothem, respecting the incredible corporate complexity they operateunder. We worked with them to discover the underlying blockers that can now be solved with this new data and process design.

We deployed AI Agents for end-to-end invoice processing across our client's SAP environment, handling the complexity of mixed invoice types with integrated VAT tax coding and compliance validation from the start.

Automation rates tracked post go-live revealed a critical insight: inaccurate and duplicate supplier master data in SAP was the single biggest factor capping performance. When the system couldn't uniquely match a supplier, automation defaulted to human intervention every time. E.g.: Out of 1,400 supplier records reviewed in a single company code, 572 required remediation.

As a result, webuilt the Supplier Master Data Validation agentenriching SAP supplier records from public sources through a custom UI.

From that foundation, the solution footprint expanded:
‍Dunning Letter Automation was implemented to eliminate a fully manual invoice-adjacent process and cut handling time by over two minutes per letter. Contract Accounting planned to follow, with AI agents reviewing SaaS contracts to advise on software cost capitalization.

Towards repeatable touchless workflows

From manual chaos to methodical agent deployment - the process redesign enables sustainable change from the ground up. Client team scan now standardize how work gets done while respecting their in-house complexity. The results speak for themselves: