68%
Queries resolved without a person
→ All case studies
Energy · NetherlandsAgentic AI · SAP hypercare
ERP Hypercare for an Energy Company, Answered by an Agent
A European integrated energy company came out of a major ERP go-live with a hypercare problem: finance teams, field engineers, customer ops and renewable-asset managers all asking very different questions of one overloaded support team. We gave the questions somewhere better to go.
80%
Fewer misrouted escalations
8 → 4
Hypercare FTE, via attrition
7 weeks
Build, 2024
How the system runs
The inputFour user groups
→
Finance, field engineers, customer ops and renewables, one overloaded support team.
We builtRAG agent
→
Trained on the full ERP documentation set, in the portal and Teams.
We builtPersona routing
→
Answers adapt to the asker’s role, from finance to field.
We builtGap analysis
→
Questions with no good answer flag missing documentation, 45 found.
The outcome68% auto-resolved
80% fewer misrouted escalations; team halved by attrition.
The situation
One Generalist Support Team, Four Very Different User Groups
The documentation existed, but not in the flow of work, and the fast-growing renewables business was creating new asset types and trading processes faster than documentation could keep up. A hypercare team of eight was under sustained query pressure, with no AI layer to triage or contextualise before escalation.
What we didOne Agent, Four Personas, Live in Seven Weeks
A lead consultant and one AI developer shipped the agent from knowledge base to phased go-live in seven weeks:
- Trained a RAG agent on the full ERP documentation set, including the renewables-specific process guides.
- Built multi-persona routing: the agent adapts its answers to the user’s role, from finance to field engineering.
- Integrated it into the company portal and Microsoft Teams, where the questions actually arrive.
- Set up dynamic knowledge updates, so new documentation is auto-indexed, and gap analysis that flagged 45 missing process documents.
The design principle
The result
"The agent does not just answer. It notices which questions have no good answer yet, and says so."
A Permanent Operational Tool, Not a Hypercare Patch
- 68% of hypercare queries resolved without human intervention.
- Misrouted escalations down 80% thanks to persona routing.
- Renewables asset-management queries, previously unserved, handled at 85%+ accuracy.
- The hypercare team went from 8 to 4 through natural attrition, saving $350,000–550,000 a year.
The agent outlived hypercare and is now cited internally as a flagship of the company’s digital transformation programme.
Our fee
$45,000 – $70,000
Duration
7 weeks + ongoing
Team
Lead + 1 AI developer
Client anonymised under NDA. Figures are engagement-reported ranges, measured against the baseline set at the start.
Your number next
Want a Savings Estimate for Your Workflow?
A baseline findings report puts a real figure on the table, before you commit to anything bigger.