On-premise deployment

AI voice agents inside your own data centre

AI Rudder runs single tenant in your data centre or your own cloud account. Customer data stays in your environment, and releases follow your change windows.

Two ways to run AI Rudder

In our cloud, or in your data centre

The stages are the same. Where the agents run, who runs the infrastructure and how releases reach you are different.

Cloud

AI Rudder cloud

Where it runs
AI Rudder cloud on AWS or Alibaba Cloud, in the region you select
Infrastructure
Run and monitored by AI Rudder
Releases
Scheduled releases, rolled out region by region
Capacity
Agreed with you as your call volume grows
Suits
Most lenders, fintechs, BPOs and service teams
How cloud deployment works
On-premise

Your data centre

Where it runs
Your data centre or your own cloud account, for you alone
Infrastructure
Your servers and GPUs, sized with AI Rudder; set up and kept current by AI Rudder’s engineers with your IT team
Releases
In your change windows, through your approval process
Capacity
Set by the hardware you run
Suits
Banks and other regulated companies whose data must stay in their own environment
What you get

Built for your security review

Single tenant

Separate SIT, UAT and production environments, for you alone

Your data stays with you

Calls, recordings, transcripts and customer data stay in your environment

Encryption

Encryption of personal data

Your release process

Maker-checker approval, in your change windows

A full audit trail

Audit logs and monitoring you can download and keep

High availability, if you need it

Redundant production services, set up when your requirements call for them

On-premise deployment

From proof of concept to go-live

Seven stages. Your technology and security teams sign off each one.

  1. 01Proof of conceptYour calls, your languages, at your scale
  2. 02Technical designArchitecture, security and integrations, signed off
  3. 03EnvironmentsSIT, UAT and production, each for you alone
  4. 04IntegrationsLines, systems, sign-in and encryption
  5. 05Security reviewYour questionnaire, our reports and documents
  6. 06TestingSIT, then UAT, with fixes inside the UAT window
  7. 07Go liveIn your change window, then close support

Then releases in your change windows, and back to design for the next set of calls

What sets the timeline

Each on-premise deployment is built for one customer, so the plan is agreed with you, project by project, and both sides track it. How long it takes depends on:

Your requirements

Security, data protection and availability standards

How much is customised

Integrations and features built for your environment

Your infrastructure

When your servers, GPUs, network and access are ready

Your approvals

Design sign-off, security review and UAT on your side

Stage by stage

What happens, and who does what

Stage 01Proof of conceptEnds with results and hardware sizing

What happens

  • The agents run on your real call types, in your languages
  • Scale tests in the markets you will call
  • The hardware needed is sized from the results

Your team

  • The calls, scripts and success measures for the test

AI Rudder

  • A test set-up and the results, by market
Stage 02Technical designEnds with your sign-off

What happens

  • A technical design document: architecture, environments, network, accounts, integrations and security controls
  • The design is reviewed with your technology and security teams
  • Your teams sign it off before anything is built

Your team

  • Your architecture standards, network rules and security requirements
  • Sign-off on the design

AI Rudder

  • The design document
  • Server and GPU sizing for your call volume, languages and voices
Stage 03EnvironmentsPaced by your infrastructure

What happens

  • Separate environments for SIT, UAT and production, each single tenant
  • Built in your data centre, or in a cloud account you own
  • AI Rudder’s infrastructure team builds each one, then our engineers configure it

Your team

  • Servers, GPUs and memory, or the cloud account
  • Network access for our engineers (VPN or on site)

AI Rudder

  • The build of each environment
  • Monitoring, logs and alerts in your environment
Stage 04IntegrationsWith the environments

What happens

  • Telephony: your SIP trunks and session border controllers, with encrypted signalling and media where you need them
  • Your systems: CRM, core banking, message queues, email, file transfer, single sign-on and logging standards
  • Messaging through the official WhatsApp Business API

Your team

  • Access to each system, and the formats it uses
  • Your carrier lines and load balancers

AI Rudder

  • The integrations, adapted to your systems and standards
  • Two-step sign-in, encryption of personal data and audit logs
Stage 05Security reviewAlongside testing

What happens

  • Your security questionnaire answered
  • Third-party penetration test and vulnerability scan reports, with findings fixed
  • Design, network, code-review and test-case documents
  • Maker-checker approval on every release

Your team

  • The questionnaire and the evidence you need

AI Rudder

  • The reports, documents and fixes
Stage 06TestingEnds with UAT sign-off

What happens

  • SIT on your test data, then UAT with your team
  • Engineers on standby; fixes turned around inside the UAT window
  • The approved UAT version is then moved to production

Your team

  • Test cases, test data and UAT sign-off

AI Rudder

  • Test runs, fixes and the move to production
Stage 07Go live and afterIn your change window

What happens

  • Go-live in your change window, through your release approval
  • Daily checks of every call run and line in the first weeks
  • Later releases follow your release process and windows

Your team

  • Release approvals and change windows
  • Changes to scripts and rules in BotLab, without us

AI Rudder

  • Product updates merged into your version on a fixed baseline
  • Engineers on site or remote, as you prefer
Who provides what

Your infrastructure, our engineering

Your team provides
  • Servers, GPUs and memory, or a cloud account you own
  • Network access for our engineers: VPN or on site
  • Your carrier lines, SIP load balancers and session border controllers
  • Access to your CRM, core banking and other systems
  • Your security questionnaire, release process and change windows
  • Test data, and sign-off at each stage
AI Rudder provides
  • Sizing of servers and GPUs for your call volume, languages and voices
  • The build of every environment: SIT, UAT and production
  • Integrations adapted to your systems and standards
  • Encryption of personal data, two-step sign-in, audit logs, monitoring and alerts
  • Design, network, code-review and test documents; penetration test and vulnerability scan reports
  • Engineers on site or on standby, and product updates merged into your version
Sizing

Hardware sized before the build

The servers and GPUs you need depend on how many calls run at once, the languages and the voices. AI Rudder sizes them with you from the proof of concept, and they are fixed in the technical design.

There is no one standard specification: requirements differ from customer to customer, the AI hardware most of all.

After go-live

Updates on your schedule

Your version stays on a fixed baseline. AI Rudder merges product updates into it and releases them through your process.

Your release process

Every release goes through your review and approval, in your change windows

Monitoring and alerts

Call runs, lines and business data watched, with alerts to both teams

Support

Engineers remote or on site. Response times are agreed in your contract

What you control

What your team changes without us

Your team sets the rules, the scripts and the limits, and changes them in BotLab.

Scripts and call flows

In BotLab

Rules and limits

Per deployment

Consent and disclosure

Per market

Handoff to people

When, and to whom

Knowledge sources

Knowledge Studio

Responsible AI

Responsible AI on every call

How Responsible AI works
Your rulesScripts, limits and contact times set by your team
Disclosure and consentPer market, on every call
Handoff to peopleWith the context, when a person is needed
A full recordRecording, transcript and QA check for each call

How data is protected and where it is hosted

Data protection and residency

Tell us which market, and which calls.

Our in-market team will plan the first deployment with you, in the language your customers speak.