Situational Awareness

## Everyone agrees sustainability matters. So why does almost nobody change their daily habits?

Because this is not just a data problem—it is an **engagement problem**.

Buildings already have utility bills, dashboards, statistics, and reminders. Yet most of these tools explain what happened after the fact. They rarely give ordinary people a compelling reason to act differently in the moment.

That insight inspired **Situational Awareness**: an AI-powered sustainability game and retail digital twin that makes resource-conscious behaviour safe, social, measurable, and fun.

## What it does

Situational Awareness turns everyday sustainability actions into live challenges across:

- Energy
- Water
- Waste
- Food
- Transport
- Buying habits

A manager launches a game day and shares a QR code with staff. Players join from their phones, snatch available challenges, complete tasks, earn points, and compete on a live leaderboard.

An **AI Game Master** recommends relevant challenges based on previous participation and delivers the right prompt at the right moment. The system records claims, completions, timing, points, and available evidence so managers can understand which challenges actually engage people.

Our first scenario, **Green Close**, focuses on reducing avoidable resource use during a store’s closing shift. Staff can complete safe actions such as switching off non-critical lighting or separating reusable packaging, while refrigeration, safety systems, customer service, and other protected operations remain outside their control.

## More than another leaderboard

Gamification can create short-term excitement, but novelty alone does not create lasting behaviour change. We designed Situational Awareness around a repeatable learning loop:

1. **Challenge:** Give each player a clear, achievable action.
2. **Nudge:** Recommend it at an appropriate moment.
3. **Compete:** Use points, individual progress, and a leaderboard.
4. **Measure:** Record what was attempted and what evidence exists.
5. **Learn:** Analyse friction and participation after each game day.
6. **Improve:** Adapt future recommendations within strict limits.

The AI can improve task recommendations and make bounded adjustments to point incentives, but it cannot weaken safety rules, authorize protected equipment, fabricate measurements, or make employment decisions.

## Test before you pilot

Situational Awareness also includes a retail digital twin that allows operators to compare their current workflow with a proposed intervention before trying it in a live store.

Managers can enter utility information and operational assumptions, then run matched baseline and intervention simulations. The system estimates ranges for energy, cost, emissions, staff time, task completion, customer-service incidents, and operating impact.

A 3D replay makes the result understandable rather than leaving it buried in a spreadsheet. Every event is reconstructed from an append-only ledger, allowing operators to see what changed, when it changed, and why the Game Master accepted or rejected an action.

This connects the two sides of the problem:

- The **digital twin** helps determine whether an intervention is safe and worthwhile.
- The **game** helps motivate people to carry it out consistently.

## How we built it

We built the backend with **Python, FastAPI, Pydantic, and SQLite**. It contains the authoritative simulation engine, project and utility-data workflow, live game system, impact analysis, and append-only event ledgers.

The frontend uses **React, TypeScript, Vite, React Three Fiber, and Three.js**. It provides the manager workspace, mobile staff experience, QR-code handoff, live leaderboard, impact results, and interactive 3D replay.

AI providers—including OpenAI and local Ollama-compatible models—can propose structured actions, explanations, and post-game analysis. However, the language model never directly changes the simulated store. Every proposal passes through deterministic permissions and safety rules enforced by the Game Master. If an AI provider is unavailable or returns invalid output, the experience continues using a deterministic fallback.

We also separate evidence into four clear categories:

- **Measured:** Confirmed real-world inputs
- **Derived:** Reproducible calculations
- **Assumed:** Editable modelling inputs
- **Simulated:** Behavioural hypotheses

This prevents an engaging AI experience from turning uncertain estimates into misleading sustainability claims.

## Challenges we faced

The hardest challenge was balancing **fun, intelligence, and trust**.

A sustainability game needs excitement, but it must not encourage unsafe behaviour or reward people for interfering with critical building systems. We solved this by separating AI recommendations from authoritative control: the AI may suggest, but deterministic rules decide.

We also had to keep the simulation, mobile game, leaderboard, and 3D replay consistent. We used sequence-numbered event ledgers so every interface reads from the same history instead of inventing its own version of events.

Another challenge was measuring impact honestly. Completing a challenge is evidence of engagement, but it is not automatically proof of energy or carbon savings. The platform therefore preserves measurement gaps and labels estimated or unverified outcomes clearly.

## What we learned

We learned that a credible sustainability product must answer three questions:

1. **Is the action safe?**
2. **Is it worth doing?**
3. **Will people actually want to do it?**

Traditional dashboards mostly address the second question. Situational Awareness combines operational guardrails, impact modelling, timely nudges, and game mechanics to address all three.

Most importantly, we learned that AI is most valuable when it improves relevance and responsiveness—not when it is given unchecked control. Our Game Master architecture uses AI for personalization and explanation while keeping safety, permissions, scoring, and impact calculations transparent and testable.

## What’s next

Next, we want to connect the platform to real building sensors and equipment data, introduce team streaks and social challenges, and validate behaviour change through multi-day pilots.

We also plan to expand beyond Green Close with reusable challenge packs for offices, schools, campuses, and other building types.

Our goal is simple: **stop giving people another sustainability dashboard to ignore, and start making the greener choice the action they want to take.**