Jasmine Smart Assistant: local AI that watches, answers, and optimizes your home

Short version: Jasmine Smart Assistant is a physical device that runs a virtual home assistant AI locally in your house. It analyzes your smart-home data inside your home (not in the cloud), answers questions about the home, uses CCTV streams as a local “watchman,” and finds ways to optimize energy, water, and other consumption. Below is a clear, non-technical explanation you can use for customers, family members, or a website.

What it really is 

Jasmine Smart Assistant is a small computer you keep at home. It connects to your Jasmine Smart Switches, sensors, meters, and CCTV cameras. Instead of sending everything to some faraway server, it thinks on the device — so the information and decisions stay inside your house unless you allow otherwise.

It does three main things:

1. Answer your questions about the home, in normal language. (“Which lights are on?”, “When did the fridge run most yesterday?”)

2. Act as a watchman using CCTV streams — detect people, unusual motion, or safety events and alert you.

3. Optimize consumption — learn how you use power, water, and heating/cooling, then change schedules or suggest actions that save money and reduce waste.

Why local processing matters 

From first principles this is how value is created:

• Latency: Local AI reacts instantly. No waiting for a cloud round-trip when you say “turn off the lights.”

• Privacy: Data stays inside your house unless you explicitly share it — fewer opportunities for mass data collection or breaches.

• Reliability: If the internet goes down, core features (security alerts, basic automations, local voice commands) keep working.

• Cost: No per-minute cloud compute billing for everyday decisions — predictable device operation and lower long-term costs.

Those four points are why an on-prem device wins for home control, especially for security and high-value automation.

Everyday benefits

• Ask naturally: “Hey Jasmine, is the stove on?” → immediate answer and optional action.

• Watchman on duty: Jasmine reviews CCTV locally and sends an alert if it sees a person near your back door at 2 a.m. It can also ignore expected motion (pets, curtains) because it learns what’s normal.

• Save money automatically: Jasmine notices HVAC cycles and adjusts a 30-minute pre-cool before you return — cheaper than running full power while you’re gone.

• Accessibility: For someone with limited mobility, voice or simple prompts control lights, fans, and reminders.

• Reduced false alarms: Because it learns your home’s patterns, alerts are more accurate — fewer nuisance calls at 3 a.m.

Concrete, real-world examples 

Morning: “Good morning” runs a routine — bedroom light to 40%, kettle preheat, open curtain motors, and report today’s expected electricity use.
Away / Security: When you leave, Jasmine arms a watchman mode: doors locked, motion model wearing a “notify only if unknown person” rule. If a camera detects an unknown person, you get an alert with a short clip.
Energy saver: It notices the water heater cycles at night when tariffs are low and schedules heavy draws (washing machine) to cheaper hours automatically.
Elder care: If motion sensors detect an unusual inactivity pattern from an elderly person, Jasmine sends an SMS and turns on a bedside lamp to prompt movement.

How it works 

1. Connect: Jasmine plugs into your home network and pairs with your smart switches, sensors, meters, and cameras.

2. Learn: The assistant collects device-level signals (switch on/off, power draw, temperature, motion) and learns normal patterns. This learning happens inside the hardware.

3. Observe (CCTV): Video streams are analyzed using onboard vision models to detect people, faces (if allowed), or unusual actions. The system filters routine events (pets, known family members) so you get meaningful alerts.

4. Decide: When you ask a question or an event happens, the assistant evaluates locally and takes action or sends a notification.

5. Optimize: It runs simple models to reduce consumption — e.g., predictive scheduling, adaptive setpoints, and recommending hardware fixes (bad refrigerator motor, leaky valve) when it detects inefficiencies.

CCTV as the watchman — privacy, accuracy, and practical limits

Using CCTV for security is powerful but sensitive. Here’s how Jasmine approaches it in simple terms:

• Local analysis by default. Video is examined on the device; no raw footage leaves your home unless you choose to upload or back it up.

• Smart filtering. The system is trained to reduce false positives (ignore pets, trees in wind) so you only get important alerts.

• Clip sharing is opt-in. If an alert is sent to you or a trusted contact, the clip shared can be encrypted and controlled by you.

• Retention & review. You control how long clips are kept and who can review them. Old clips can auto-delete after a set time.

• Legal/compliance note. Use of cameras may be subject to local laws (recording in common areas, notices to visitors). Jasmine recommends clear signage and consent where required.

Optimization of home consumption — what it can actually do

Jasmine doesn’t “magically” cut bills — it applies consistent data+control to reduce waste:

• Detect loads: Recognizes high-consumption devices and how often they run.

• Shift timing: Moves non-urgent energy use (washing, EV charging) to lower-tariff hours.

• Predictive control: Anticipates heating/cooling needs based on when people usually return and preconditions only as needed.

• Recommendations: Suggests hardware upgrades (e.g., old AC needs servicing) when inefficiencies are detected.

• Continuous learning: Measures the impact of changes and refines recommendations — a simple feedback loop that compounds savings over months.

Simple setup — what a typical user does

1. Place Jasmine Smart Assistant where it can reach your network and devices (near router or switch).

2. Connect via the Jasmine mobile app. The app will find local devices automatically.

3. Enable “Watchman” for each camera you want monitored (opt in).

4. Let it run for a few days so it can learn normal patterns.

5. Start with a single routine (e.g., “Good night”) and expand.

No coding required. Most users do this within the phone app in 10–20 minutes.

Privacy & security 

• You opt in. Nothing is analyzed in the cloud unless you choose it.

• Control sharing. If you want cloud backup for clips or remote professional monitoring, it’s a separate opt-in service with explicit permissions.

• Encrypt & revoke. Any data that is sent off-site is encrypted. You can revoke cloud access at any time.

• Transparent defaults. The device ships with local-only defaults and only enables anything else after consent.

• Compliance and logs. You can review a simple activity log that shows who accessed what and when.

Simple prompts people can use

Use these as-is when speaking to Jasmine or another integrated assistant:

• “Jasmine, are any doors unlocked?”

• “Set house to energy-save mode.”

• “Show me today’s CCTV highlights.” (returns short clips, if you allowed sharing)

• “Did I leave any lights on downstairs?”

• “Create a ‘Movie Night’ routine: dim living room lights to 30% and turn on air-purifier.”

These prompts can also be typed into ChatGPT/Claude/Grok once you integrate the assistants.

Frequently asked questions 

Q: Will Jasmine record everything on my cameras?
A: No — by default video is analyzed locally and stored only as short clips when an event happens. Full recording is an opt-in feature.

Q: What happens when the internet is down?
A: Local automations, security detection, and voice responses that run on the device continue to work. Cloud-only features pause.

Q: Can I give access to a neighbor or caretaker?
A: Yes — the app supports limited access roles. You decide which devices or alerts they can see.

Q: Is it hard to install?
A: No. Most users set it up in the Jasmine app in under 30 minutes. Installers can do a full setup for multi-unit buildings.

For product & go-to-market teams (short, actionable guidance)

• Core messaging: “Privacy-first home AI that thinks locally.” Lead with privacy + reliability benefits.

• Primary personas: Families with elderly members, security-conscious homeowners, environmentally conscious households, builders of premium apartments.

• Key KPIs to track: false alarm rate, monthly energy savings per household, time to onboard, retention after 90 days.

• Upsell ideas: professional monitoring (opt-in), cloud backup for video clips, energy-audit reports, priority firmware updates.