← All case studies

EV Energy · Customer Support

Support answers in minutes, not a business day

EV Energy runs customer support across 26 utility programs worldwide. We built an AI agent that answers compatibility tickets in minutes - and customers noticed unprompted. In six weeks, first response went from a business day to 7 minutes, resolution rose 35% to 55%, and CSAT 50% to 70%.

EV Energy logo
7 min
First response (was ~1 business day)
5 hours
Blended category response (was 1–2 days)
35% → 55%
First contact resolution
The client

EV Energy.

EV Energy aggregates EV chargers, batteries, and solar into virtual power plants for utilities worldwide. Its support team handles tickets across 26 utility programs, each with its own compatibility rules - and a poor driver experience means disenrollment.

The challenge

Where it broke.

As EV Energy's program portfolio grew, its support team faced volume spikes, specialist-knowledge requirements, and the manual overhead of researching every ticket. A backlog that grew faster than it could be cleared showed the old approach couldn't scale.

  • Manual lookup per ticketEvery compatibility ticket required individual research — checking program-specific vehicle and charger lists by geography. Each ticket demanded specialist knowledge that wasn't easily shared or scaled.
  • Volume spikes and backlogsStaff leave and ticket volume spikes created backlogs that were hard to recover from. Response times stretched to two business days during peak periods.
  • Specialist knowledge dependencyEach utility program had unique rules not easily transferable across agents. When key team members were unavailable, quality and speed dropped with them.
  • Inconsistent response qualityHuman-drafted responses varied across agents and programs. Tone, accuracy, and structure differed based on who was handling the ticket.
  • Scalability ceilingThe existing model could not scale cost-effectively to support continued growth in utility programs and driver enrollments without proportionally increasing headcount.
The build

What we shipped.

Phased AI automation for customer support — starting with human-in-the-loop validation, then expanding to autonomous responses as quality proves out. Built on EV Energy's own infrastructure, with full data sovereignty and GDPR compliance.

  • Phased human-in-the-loop rolloutBuilt trust incrementally. Helper notes first so the team could see and verify quality before relying on it — then autonomous responses once confidence was established. Higher-risk categories remained in helper-note mode while confidence was built.
  • 26-program compatibility engineThe AI agent manages the full complexity of utility program rules — vehicle compatibility lists, geographic eligibility, business logic variations, and program-specific response requirements — automatically detecting the relevant program from each incoming ticket.
  • Self-hosted on EV Energy's AWSThe entire system runs on EV Energy's own AWS infrastructure using n8n. No third-party data access. Full data sovereignty with the information security standards required to maintain trust with both utilities and consumers. As an EU-headquartered company subject to GDPR, this was non-negotiable.
  • Real-time response under 7 minutesCompatibility tickets are now answered in an average of 7 minutes — including outside business hours. Customers began noticing and volunteering feedback on the speed without being prompted.

How it rolled out

Built in stages.

01

Helper Notes — AI generates a draft response as an internal ticket note. Human agents review and approve before sending. Used to validate quality and refine prompts.

Phase 1
02

Direct Responses — High-confidence, low-risk ticket categories (compatibility and eligibility queries) switch to fully automated direct responses — no human review required.

Phase 2
03

Ongoing Expansion — Higher-risk categories (incentive queries, charging session earnings) remain in helper-note mode while confidence builds. Gradual, controlled expansion across more programs and categories.

Phase 3
04

Infrastructure Hardening — Transition from rapid proof-of-concept to production-grade architecture embedded within EV Energy's engineering infrastructure, enabling their internal team to independently maintain, update, and deploy changes.

Phase 4
We wanted to try and relieve the team quickly, so we had a scrappy approach — which was perfect. And now we're moving across to something that is more secure, instills confidence with our partners, and allows us to maintain and deploy changes effectively. Some pretty compelling metrics. Very pleased with how it's all going, and there's still more headroom to go after.
Sandy Neill
Head of Customer Support · EV Energy

What’s next

The roadmap from here.

01

Expanding AI coverage

Progressively moving more programs and ticket categories from helper-note mode to full direct response.

02

Proactive issue detection

Using AI to diagnose IoT device issues and virtual power plant dispatch problems before drivers need to contact support.

03

Energy optimization with AI agents

Applying AI to optimize the real-time dispatch and control of distributed energy resources across the virtual power plant.

04

AI-powered business case modeling

Enabling utility clients to model the value of energy flexibility in minutes rather than months.

05

Consumer agent integrations

Preparing for a future where personal AI assistants can enroll consumers into smart energy programs on their behalf.

The solution behind it

AI customer support agent for high-volume teams

The same phased AI automation approach deployed for EV Energy. Human-in-the-loop validation, then autonomous responses as quality proves out. Built on your infrastructure.

Start here

Bring one workflow.

A 30-minute working call. We’ll map one operational workflow to an agent stack and tell you honestly whether it’s worth building.