How we build and run agents
An agent is only as good as its procedures and the people behind it. These are the standards every ViaAI agent is built to.
Written procedures
Every agent runs on written operating procedures you approve: what it may answer, what it may do, and when it must hand off.
Real actions, least access
Agents work inside your tools, such as your calendar, inbox or CRM, with only the permissions their procedures need.
A person for judgment calls
Anything outside its procedures goes to a named person on your team, with the conversation attached.
Every action on the record
Each conversation and action is logged, so you can see what the agent did and why.
Tested before customers see it
Your team tries the agent on a private link with real questions before it goes live.
Measured by what it resolves
You get a regular report: what the agent resolved, what it handed off, and what customers asked most.
Changed only with your approval
On a care plan we review conversations and propose improvements. Nothing changes until you approve it.
Always says it's an AI
Your agent introduces itself as an AI assistant and offers a person whenever someone asks.
From first conversation to a running agent
Scope
We map the handoff with you and send a written scope: the procedures, the tools it connects to, a delivery date and a fixed fee. Scoping is complimentary.
Build
We write the procedures, connect your tools and set up the channels: phone, chat or email.
Test
Your team works the agent hard on a private link. We tighten the procedures until it handles your real cases.
Launch and run
We launch on accounts in your business's name and hand over the logins and procedures. An optional care plan keeps it reviewed and improving.
Questions
What does it cost?
Each build is a fixed fee agreed in writing before work starts. Running costs, such as AI usage and phone minutes, are listed separately in your scope. An optional monthly care plan covers reviews, reporting and changes. Scoping is complimentary.
Which AI models do you use?
We build mainly on Claude by Anthropic and choose the model for each task. Voice agents use a dedicated voice platform for natural speech.
Which tools can an agent connect to?
Most tools with an integration, an API or a Zapier connection: calendars, email, CRMs, spreadsheets, booking systems and store platforms. We confirm yours during scoping.
Will customers know they're talking to an AI?
Yes. Your agent introduces itself as an AI assistant and offers a person whenever someone asks for one.
Who owns the agent?
You own your procedures, your data and your accounts. Agents run on accounts in your business's name wherever possible, and you get the logins and documentation at handover.
Can an agent handle sensitive information?
We don't build agents that handle patient health records, card numbers or Social Security numbers. Payments go through your payment provider, never through the agent.
How long does a build take?
Your scope includes a delivery date. An agent on one channel is quickest. Phone agents and agents that act in several tools take longer, because they are tested on more cases.
What won't you build?
Outbound robocalls or texts without consent, spam, agents that pretend to be a person, and agents that give medical, legal, tax or financial advice.