Skip to content
MAISONLABS
PROJECT_KRYPTON

Krypton Onboarding Core

Qualifying a home for installation meant video calls and scattered emails to gather technical details customers didn't know how to give. We replaced that with AI assistants that collect it all over text and WhatsApp — filtering out the homes that won't work before anyone on the team gets involved.

Before a crew could be sent out, someone had to confirm the home could take the hardware — electrical supply, cable distances, physical clearances. Gathering those details meant chasing customers across calls and emails for scraps of information.

We built AI assistants that gather those details straight from the customer, over the apps they already use. The system sorts the technical facts on its own, drops the homes that don't qualify, and passes only the serious, workable ones to the engineering team.

SECTOR / Renewable Energy & Hardware InstallationDISCIPLINE / Serverless & Multi-Agent OrchestrationBUILT WITH / AWS Lambda · DynamoDB · API Gateway · WhatsApp / SMS Gateways · Agentic Orchestration
Chat frames feeding a branching teal constellation of agent nodes to one lit qualified node and a human-handover node on obsidian
Property data captured in full
100%
Qualified leads converting
+42%
Unqualified enquiries reaching staff
-87%
01

Qualifying homes without the video calls

Checking whether a home can take an installation is a technical job — supply amperage, clearances, layout limits, cable runs. It used to mean booked video surveys and long email threads. Now a team of AI assistants collects it all in conversation: every property comes through with complete data, qualified enquiries convert 42% more often, and engineers no longer lose hours to homes that were never going to work.

02

Why the details never arrived clean

The details needed to quote a job were real, technical, and scattered everywhere.

Data customers can't easily give

Knowing if a home could handle the hardware meant technical details most homeowners can't produce without being walked through them.

Scattered across channels

Staff ran phone calls, live video reviews, and email threads just to get a photo of a fuse board or a cable measurement — every job spread across a different channel.

The human cost

The team burned hours chasing people who were never going to buy, and mistakes crept into the CRM along the way.

03

How the assistants work

One conversation thread, several focused assistants.

In their own apps / SMS + WhatsApp

The conversation happens in the text apps people already use every day. Customers reply in their own time, with nothing to download.

A team of focused AI assistants

Not one catch-all chatbot, but several small assistants, each responsible for getting one part of the job right.

Straight into the CRM / Lambda + DynamoDB

Every answer is checked and written straight into the CRM as it's captured — so nothing is mis-typed or missed, and the system handles as many conversations at once as you throw at it.

A clear hand-off point

When a home clearly meets every requirement, it's approved automatically. Anything unusual is packaged into a tidy summary and handed to a person — with all the detail already gathered.

04

What changes when the team only sees real jobs

Where the hours go once the AI does the qualifying.

Near-zero capture cost

With AI doing the first round of data gathering, the cost of qualifying a property fell to almost nothing.

The team only sees real jobs

With 87% of unqualified enquiries filtered out first, staff spend their time only on homes that are genuinely worth a visit.

No wasted call-outs

Because every booked job comes with complete, consistent property data, crews stop arriving to installations that can't actually go ahead.

Next PieceNexus Logistics Core
The Audit

This is what we'd map for your firm.

Tell us where the time goes. We'll tell you plainly whether we can flatten it, what it would involve, and what it would return.