HomeLive systemsARIAWorking agentsHow we workOur viewFounderBeyond the build Start a conversation
How we work

Proven in our own business first.

Five strengths, how our agents are organised, the security we run, and a four-step engagement that starts with a conversation.

Our strengths

Access to market

Live systems, live channels and a working sales and operations team reaching the market every day.

Easy to do business with

Clear entry points, many ways to pay, e-signature and fast onboarding.

Support around the clock

AI agents answer 24/7, and people step in during business hours.

Security and data protection first

Encrypted data, verified sign-in, step-up checks on sensitive actions, a sealed outbound gate and a tamper-evident record of every decision.

Built and run by us

Human-led, AI-delivered, and in production rather than on slides.

How we run our operations

Every agent has a job, and people stay in charge.

Each system and working agent below plays a role you would find in any well-run company, from the front desk to the agents working behind the scenes. Select one to see it.

People lead
People decide illustrationPrincipal · Founder and CEOPeople decideSets direction and approves anything that matters.
People step in illustrationSales and operations teamPeople step inTakes over from the agents during business hours.
Medida by Novai Systems demo, sample dataMeasured successMedida by Novai SystemsShows every person their shift, daily goals and call intelligence, measured the same fair way.Open demo ↗
Behind the scenes · runs WorkBench every dayBehind the scenes demo, sample data: email sequences and 12 background agentsAll 13 in one demo ↗
Mail roomEmail sequencesMore than 60 automated emails carry people from sign-up through onboarding, bookings, billing and renewals: a reminder, then a final notice, then a clear next step. Outreach emails go out only when a person says send.Open demo ↗Sales plannerNext-step engineAfter every call, picks the next best step for each lead: call again, switch to email or hand it to someone else. "Do not call" is final.Open demo ↗Call reviewerMorning demand checkReads the previous day's sales calls and flags families who need care, every morning.Open demo ↗Compliance clerkExpiry watchWatches every provider licence, insurance policy and background check, and warns before one lapses.Open demo ↗Contract coordinatorCare contract watchKeeps care contracts moving: flags requests no one has picked up, schedules quality check-ins and escalates open disputes.Open demo ↗Billing clerkBilling and payoutsRuns weekly billing and payouts on the rules people set, keeps held payments current and checks every subscription nightly.Open demo ↗Onboarding coachPartner remindersWalks new partner agencies through sign-up, payout setup and updated terms, and flags anyone stuck on the same step.Open demo ↗Support supervisorInbox watchFlags support, payment and dispute messages left waiting, so a person picks them up.Open demo ↗Message moderatorContent checkFlags requests to take contact or payment off the platform in job messages.Open demo ↗Safety screenStroke and scam checksIn chats and calls, points stroke signs to 911 first and flags scam scripts for a person to check.Open demo ↗Agent examinerNightly agent testsTests every agent against its hard rules each night, and watches fees and terms for drift around the clock.Open demo ↗Alert deskAlert custodianTurns raw alarms into one clear alert for the right person, without repeats.Open demo ↗ArchivistNightly backupsBacks up the whole platform to the cloud every night.Open demo ↗
Callers, visitors and buyers
  • ↑ Up Signals and measures flow up to ARIA and the people.
  • ↓ Down Decisions and approvals flow down to the agents.
  • ■ Stop Risky actions wait at LGD until a person says yes.
The voice desk

Emma answers. Maya and Susan call. People decide.

Three AI voices with one job each, all reporting to the team and to ARIA. Here is how a call runs.

Answers

Emma

Picks up WorkBench calls day and night and answers in plain words. In business hours she passes the caller to a person; if no one is free, she takes a message for a call back.

Call +1 (213) 603-5940

Calls

Maya

Calls customers back and calls new business leads, in English or Spanish, only when a person has approved the call. Anything sensitive waits for a person.

Ask about Maya

Care calls

Susan

Calls licensed home care agencies and care follow-ups in English, Spanish, Tagalog and Cebuano, only when a person has approved the call. If she can't answer something, she brings in our care team.

Ask about Susan

  1. QueueA call is added from a missed call, a website enquiry, a follow-up or a lead list.
  2. A person approvesNothing is dialled until a teammate approves it, one call or a list they chose.
  3. Checks firstDo-not-call, past calls, the calling schedule (8 AM to 5 PM their time, Monday to Saturday) and, for new leads, a business line are checked before every dial.
  4. The callMaya or Susan says who she is, keeps it short and ends the call the moment someone asks.
  5. Back to the teamThe result, a short summary and any warm lead go straight to the people.

A fixed calling schedule. Maya and Susan can only place calls from 8 AM to 5 PM, Monday to Saturday, in both our time zone and the time zone of the person being called, because calling-hour rules exist to protect people from calls at the wrong time and we hold ourselves to a tighter window than those rules require.* LGD checks every call against that schedule as it starts, so a call that drifts outside it is recorded and the team is emailed at once. An admin can also switch calling off until the next scheduled start, and the dialer turns itself back on at 8 AM, so a pause never depends on someone remembering to restart it.

* The calling window is a setting in the build, so it can be set to the rule a call falls under, for example the FDCPA's limits for a collections call or the TCPA's for telemarketing. Today every call uses the tighter 8 AM to 5 PM window.

Our own listening check sits between the phone transcription and the AI on Susan's calls: it checks spelled emails and read-backs before she answers, and gives the team an English line for every Spanish, Tagalog or Cebuano turn.

Try Emma, Maya and Susan in the voice desk demo, or see the voice desk in the ARIA demo (sample data). Scripts, call data and how they are built stay private.

Operating suite, in daily use

The same systems we would build for you run our business every day.

The demos use sample data. Live admin screens, real data and how they are built stay private.

ARIA demo with sample business-health widgets: revenue pulse, cash runway, pipeline and ops hygiene
The cockpit

ARIA

CEO and founder command center: business indicators, performance health and hygiene, every line and agent in one view.

Open the ARIA demo
Intelligence Hub demo showing live visitors, sessions and the day at a glance with sample data
The signals

Intelligence Hub

What is happening right now: visitors, calls, chats, leads, reps, agent health and email delivery, with a daily digest.

See the Hub demo
NSL Navigator demo: cash on hand, burn, runway and a revenue and cost forecast chart with sample data
The plan and the money

NSL Navigator

Execution and governance for every initiative, with costs, burn, runway, revenue forecasts and ROI.

Try the Navigator demo
Illustration with sample data: AJ Analytics Console with KPI tiles for SLA, CSAT, first-contact resolution, site traffic, response time and process sigma, a control chart flagging one point, and the morning brief
The quality champion

AJ Analytics Console

A Master Black Belt agent: tracks KPIs, SLAs, customer satisfaction and traffic, flags what drifts out of control, answers in plain language and sends a brief every morning. With Medida by Novai Systems, measured success for your people: shifts against work, daily goals and call intelligence, measured the same fair way for everyone.

Try the AJ demoSee Medida by Novai Systems
LGD site: Governance-as-a-Service that holds AI agent actions until people approve them
Governance-as-a-Service

LGD

Risky AI actions wait for a quorum of named people to approve, and every decision is kept on a tamper-evident record.

See LGD
Illustration with sample data: Comms layer timeline of delivered emails, a transcribed call, a confirmed text, a verified sign-in and a tamper-evident audit trail
On the record

Comms layer

Follow-up emails and notifications go out like clockwork, and every one is tracked: email delivery, text messages, voice calls with transcripts, two-factor sign-in and an audit trail.

Try the Comms demo

What you measure, accurately and effectively, improves.

The principle behind every system we run
Tried and tested

Workflows we already run.

Outcomes we deliver today, in our own operations.

  • A caller asks for someone by name and reaches them in business hours; after hours, an AI agent answers.
  • A buyer who asks for help on a sales call is flagged to the team the next morning.
  • Every outgoing email is tracked through to delivery.
  • A client's payment is held until they approve the work.
  • A care agency signs its agreement and uploads documents online, start to finish.
  • Providers are identity- and license-checked before they take work.
  • Every website enquiry is saved and the team is alerted within seconds.
  • Every chat and phone reply is checked; when a caller asks for a person or is unhappy, the team is told the same day.
  • Every call our phone agents take is reviewed daily for mistakes, and anything off goes to a person.
  • A caller who describes stroke signs is told to call 911 at once; scam attempts are stopped and reported.
  • No AI phone call goes out until a named person approves it, and every outreach batch is signed by a person before it sends.
  • Anything inside our AI agents or accounts that changes without a person, or drifts from its approved setup, is flagged within minutes. Anything serious is emailed to a person at once, and it stays open until a person marks it checked.
  • Every call recording is kept, with personal details masked, and each play is logged.
  • A lesson an AI agent learns is used only after a person approves it. Until then it is held, and anything that breaks a rule is raised for review.
Security, as it is

We guard our own systems the way we would guard yours.

No promises we cannot show. These are the protections running across Novai Systems today.

Locked admin screens

Every admin and finance screen sits behind a sign-in that needs a one-time email code and a second step such as Touch ID. Only named people get in.

Sealed outbound gate

Every scheduled job on Novai Systems’s own machine, and every one of our cloud services that sends data out, our AI voice agents and our platform among them, reaches the internet only through LGD’s outbound gate. Each new site needs the founder’s passkey, and secrets are refused every time, encrypted traffic included.

Keys that can’t pass alone

Our own systems sign each call to LGD, so a copied key used on its own is refused. Sentinel also watches our email and payment accounts for use we didn’t make, and raises an alarm if a key may be in someone else’s hands.

People approve risky actions

LGD, our Governance-as-a-Service, holds consequential AI actions until a quorum of named people approve, and chains every decision into a tamper-evident record.

Step-up checks

Sensitive actions such as payouts ask for a second check by phone code. Providers are identity- and license-checked before they take work.

Protected data

Data is encrypted, and public demos run on sample data only. Website forms check for bots, and every email, sign-in and AI call is kept on the record.

LGD Super on watch

A watcher over our live AI agents that can only look, never change. Any change inside an agent that a person didn’t make, or that has no record, goes to a person, and only a person marks it checked.

Sentinel on watch

Every attempt against a Novai Systems site is recorded, and an address that attacks twice is stopped on every Novai Systems site until a person lifts it.

Scored on ourselves

We answer the same ten-question AI agent safety assessment we offer others, from our own records.

  • * Nov 12, 2026: 23%
  • * Nov 17, 2026: 67% (before LGD Super)
  • * Nov 18, 2026: 93% (LGD Super live)

We measure it again after 30 days.

Take the safety assessment

Live

Sentinel: the watch on every Novai Systems site

Sentinel is part of LGD, our Governance-as-a-Service. It watches every Novai Systems site and records each attempt made against them. Sentinel stops an address on its second attempt: it is challenged on every Novai Systems site until a person at Novai Systems lifts it, and blocked outright if it gets through and attacks again. Real people can pass the challenge, and only a person can lift a stop. In our own security operations, AI recommends and a person decides: every decision carries a human passkey.

What we capture, we can prove.

Sentinel records every attempt against our sites the moment it happens, and every call our AI agents make is recorded and kept on our side. Sentinel's decisions sit in a ledger where each entry is sealed to the one before it, so nothing can be changed or removed without it showing. That matters more as voices and messages become easy to fake: if a call or message claims to come from Novai Systems, our records show whether it did. When it matters, the facts are on the record, and a fact on the record is how a wrong gets put right.

Hunts for hidden filesAttempts to find secrets, settings and credential files.
AI crawlersBots collecting pages for AI models, and whether they follow our rules.
Page copyingScripts that copy our pages wholesale.
Relay attemptsTries to use our servers to reach other services.
Blocked and refusedWhat the firewall stopped and what the LGD gate refused.
Repeat attackersAn address that attacks a second time is stopped across every Novai Systems site.
Scams and fake buyersScam scripts in our chats, and fake buyers on our forms, flagged for a person to check before anyone replies.
Our own repliesEvery chat and call reply our agents give is checked. Code shown to a customer, or a complaint answered with a link instead of a fix, is flagged for a person to fix.

Sentinel went live in October 2026. We will publish figures once there is enough history to show a real trend. Addresses and details stay private.

How an engagement runs

Four steps, at your pace.

A conversation

About your goals, constraints and what success looks like.

A readiness assessment

Where AI helps first, and what has to be true for it to work.

We build it with you

In production, with your people, not as a slide.

We run it, or hand it over

We operate it for you, or transfer it to your team with training.

Coming in 2027

Specialised services, coming in 2027.

2027

Outsourced operations

We run defined operations for you, measured on outcomes.

Register interest

2027

Sales and GTM build-operate-transfer

We build your sales engine or agency, run it until it performs, then hand it to your team.

Register interest

Start with a conversation.

We meet you wherever you are on the journey, from first idea to enterprise scale.

New · Medida by Novai Systems

Measured success for your people.

Medida by Novai Systems shows every person their shift against the work, their daily goals and what their calls are telling them, measured the same fair way for everyone. It helps people win, and a person decides what to do next. We run it on our own team every day.

Medida by Novai Systems demo with sample data: team today (142 calls, 33% connected, 19 outcomes) and three people with their shift status, daily goals and call intelligence