Skip to content
Vilozi
Talk to us
Talk to us

How we use AI

AI that does the work,
not just the talking.

Vilozi applies AI where it removes real hours from a trades business: answering every call, writing the paperwork, planning the day and checking its own work. Here's how.

  • 24/7

    Every call and message answered

  • 3 + ours

    Frontier AI providers, plus our own tuned models

  • 1,600+

    Automated tests run on every change

  • 100%

    Of money decisions approved by a person

Where AI works in the platform

Six jobs the AI takes on every day, each with a person in charge of the result.

  • Real-time voice

    A voice agent answers calls, understands the job, asks a dispatcher's questions and books the visit.

  • Every message, one inbox

    Texts and website chat are read, sorted and answered, with a person one tap away.

  • Document AI

    Quotes, invoices and insurance claim packages drafted from calls, photos and the work done.

  • Scheduling and dispatch

    Visits planned around crews, travel time and urgency, and re-planned when the day changes.

  • Understanding requests

    Each request is classified by job type, urgency and scope by small models tuned for it.

  • AI-built software

    AI coding agents write much of our own code, checked by engineers and a full test suite.

How a model earns a task

Evaluated first. Trusted second.

  1. Evaluate

    Every candidate model runs against a task-specific test set built from real, permitted job data and checked by people.

  2. Canary

    The winner handles a small share of live work while its answers are compared with the model it would replace.

  3. Monitor

    Once live, answers keep being checked against fresh cases, and the task moves back if quality drifts.

Read the Lab

Questions

About our AI

Do you build your own AI models?
Yes, for narrow tasks where a small model tuned on industry work beats a general one. For open-ended work we use frontier models from Anthropic, OpenAI and Google. Our routing layer picks the best one for each task.
Which AI providers do you work with?
Anthropic (Claude), OpenAI (GPT) and Google (Gemini), plus open models we tune ourselves. We add or swap models when our evaluations show a better one for a task.
Does the AI talk to customers?
Yes, on calls, texts and website chat. It always says it's an AI assistant, and a caller can ask for a person at any time.
What data does the AI see?
Only what a task needs: for example the call, the job details and the business's own rates for a quote. Outside providers never train on customer data. Our [Security](/security) page lists every company we use.
Can AI make mistakes?
Yes, which is why anything that costs money waits for a person, every draft shows its sources, and every task is measured. See our [Responsible AI](/responsible-ai) principles.

See where AI fits in your business.

Tell us how the work runs today, and we'll show you what the AI can take on.

Talk to us