/Paul Cooper
05Sourceful Energy / Case Study

Sourceful Energy Agent

A conversational control layer for home energy. Homeowners, prosumers, and SME property owners can talk to their solar, battery, meter, and EV system in plain language, then turn intent into safe automation rules optimized for cost, self-consumption, and savings.

Client
Sourceful Energy
Role
Product design + AI orchestration + iOS build
Year
2026
Status
In progress
Natural-language analytics over home energy history. Ask a practical question, get an actionable answer.

Why conversation beats dashboards

Home energy systems now produce more data than most people can parse in real time: inverter telemetry, battery state, spot prices, EV load, household demand. Most apps answer with charts. Most users need decisions.

Sourceful Energy Agent reframes the interface around outcomes. Homeowners, prosumers, and SME property owners can ask what is happening now, what changes next, and what action to take, then move directly from understanding to safe automation.

No one wants to stare at charts. They want clarity, confidence, and the next best action.

Product direction, 2026

Agentic workflows with guardrails

The core interaction model is not "chat as search." It is chat as control with a safety envelope. Users can create automated rules to optimize for cost efficiency, self-consumption, and energy savings, while system rails protect fuse limits and operational safety.

A demo mode simulates a fully integrated setup (EV, solar PV, battery, inverter, home meter) so users can learn the shape of the product before touching live infrastructure.

  • Home chat interface in Sourceful Energy Agent

    01 / Home / chat

  • Command examples in the chat assistant

    02 / Guided commands

  • Support panel and help navigation in Sourceful Energy Agent

    03 / Support

Mobile-first interaction model. Prompt guidance and support are designed as part of the flow, not afterthoughts.

2

Modes: live + full demo simulation

5

Personalities users can choose from

1

Conversational surface for action + context

Rails

Fuse and system-safety constraints

Cost-aware intelligence as product design

Model economics were a product problem, not just an infrastructure problem. Simple requests route to lower-cost models for speed and efficiency. Complex, multi-step reasoning escalates to stronger models. Tool-result summarization shifts back to a cheaper tier where possible.

That routing strategy keeps the experience responsive without letting inference costs balloon as usage grows. The AI stack effectively optimizes itself for both user experience and operating cost.

Comparative analysis flow. The agent explains trend deltas and suggests what to change next.

Personality, trust, and adoption

Home energy is a daily habit product. Tone affects trust. Users can choose from Professional, Friendly, Nordic, Pirate, and Robot personas to match how they want to interact, without changing system constraints or safeguards.

The project is currently in closed beta with controlled real-world testing by trusted community users to validate reliability and safety before broader rollout.

  • Personality selection settings

    01 / Personality settings

  • Bookmarked conversation threads

    02 / Bookmarked threads

Tone can be personalized while safety and control boundaries stay fixed.
AI-generated interface exploration
AI-assisted UI exploration used to accelerate pattern iteration.
AI-generated optimisation strategy output
Optimization outputs emphasize trade-offs and concrete actions.

What I led

I led product design, AI orchestration, and iOS implementation, with design craft as the center of gravity. The work spanned interaction model, conversation design, orchestration behavior, and app delivery in Xcode, with Figma and Claude Code used throughout the cycle.

  • iOS
  • SwiftUI
  • OpenRouter
  • Claude
  • Agentic tools
  • Demo/Live modes
  • Figma
  • Claude Code

Outcome (current)

In progress

Closed beta, controlled conditions

Focus

Accuracy + cost as core product metrics

Design-led

End-to-end ownership across stack

AI-native

Conversational + agentic workflow

The result is a conversational interface that turns dense telemetry into clear decisions and safe automation. Instead of navigating charts, users can ask, understand, and act.

Related