AgentAnywhere Swaraj

Photo

INDUSTRIES · BPO AND IT SERVICES

Delivery at scale, inside each client's boundary.

Client contracts decide where data may live. Run AI coding, testing and service operations in your cloud or your client's, with the work checked and every step recorded.

IN SHORT

For BPO and IT services firms, AgentAnywhere runs inside each client's boundary. Seva reads a service conversation and returns a summary, a quality scorecard, compliance flags and a suggested reply for a person to use, AgentAnywhere Vaani is designed to handle voice notes in Indian languages, Taksha gives delivery teams a sovereign AI engineer with specialists for legacy modernisation and testing and an independent referee on its work, and Aavaran data masking, the Agent Gateway and a recorded trail keep each client apart. Each deployment sits in your cloud or your client's, region-pinned or designed to deploy air-gapped.

01 · Contracts bind the data

A client's code and tickets cannot be sent to a tool the contract does not allow.

02 · Every account is a boundary

What is allowed for one client is forbidden for the next. The controls have to be per client, not per company.

03 · The client will audit

You need to show what the AI did on their work, in a form they can check.

WHY IT FITS

Sovereign, our models, and the Trust Platform.

Three things, in the terms of BPO and IT services: where it runs and who controls it, which models do the work, and the controls and the record around every call.

01 · Sovereignty

Each client's data stays inside the boundary that client's contract draws.

  • Each deployment sits in your cloud or your client's, in an Indian region or on hardware you name. Region-pinned, and designed to deploy air-gapped.
  • Each client is its own workspace, with its own agents, chosen model, stored keys, runs and audit events. What one client permits does not carry into another's account.
  • Swaraj, the sovereign deployment, is designed to pin a deployment to the region the contract names.

02 · Our models

The line we select, measure and serve, for the service floor and the delivery team.

  • Seva reads a service conversation and returns a summary, a quality scorecard, compliance flags and a suggested reply, in Indian languages and English mixed the way customers speak. It is shown on synthetic conversations.
  • AgentAnywhere Vaani is designed to understand, mask, classify and route voice notes in Indian languages. Its masking, routing and tally are ours; speech recognition and synthesis are open-licensed external components used for now.
  • Taksha gives delivery teams a sovereign AI engineer with specialists for legacy modernisation and testing. Seva and Taksha are built on open foundation models, measured and served by us. You can bring your own open model.

03 · The Trust Platform

The controls the client asks to see, and the record that shows them.

  • Aavaran, our data-masking layer, masks card and ID numbers in a transcript before the model reads it, with a signed Trust Receipt for each masking call.
  • Kavach is designed to screen each request for prompt injection and unsafe output. The Agent Gateway applies the client's policy at the point of the call.
  • Every step is recorded. A workspace's audit events can be filtered and downloaded as CSV, so the client can inspect its own record.

INSIDE YOUR PERIMETER

What happens to one client request.

Follow one request from the moment it arrives to the receipt an auditor can check. Raw values are designed not to cross the boundary.

One request

Illustration

Your perimeter

What the client sent“I am Priya Sharma. My card 4111 1111 1111 1111 was charged twice.”

Step 1 of 5

A request arrives

A question comes in. It carries a name, an account, a card number: the things that must not travel.

THE RULES

The rules that bind you.

What each instrument requires in plain words, where it stands today, and what it means for AI in BPO and IT services. Dates that have not arrived are written as dates.

  1. DPDP Act 2023, section 17(1)(d)

    From May 2027

    Read the Act

    What it requires

    For personal data of people outside India, processed in India under a contract with a person outside India, most of Chapter II and Chapter III and section 16 are disapplied. Sections 8(1) and 8(5), the security safeguards, still apply.

    What it means for AI

    The carve-out is narrow. It does not cover Indian customers' data, so work for Indian banks, telcos and insurers carries the full duties. Per-client controls matter on both sides of the line.

  2. DPDP Act section 8(5) and DPDP Rules 2025 (G.S.R. 846(E))

    From May 2027

    Read the Rules

    What it requires

    Commenced by G.S.R. 843(E) of 13 November 2025. Section 8(5): protect personal data, including data with a processor, by reasonable security safeguards. Rule 7(2)(b): a fuller breach report within 72 hours. Rule 6(1)(e): logs and personal data kept one year.

    What it means for AI

    A BPO is usually the processor, and a client can pass these duties down by contract. Masking before the model reads and a record per client are how you show the safeguards and rebuild an incident.

  3. DPDP Act section 16(1) and Rule 13(4)

    From May 2027

    Read the source

    What it requires

    Government may restrict transfers of personal data to notified countries, and may require Significant Data Fiduciaries to keep specified data in India.

    What it means for AI

    As far as we read, no list or categories have been notified. A deployment can be pinned to a region if a client asks for India-only processing.

  4. CERT-In Directions No. 20(3)/2022-CERT-In, 28 April 2022

    In force

    Read the source

    What it requires

    Report listed incidents within 6 hours of noticing. Keep ICT logs for a rolling 180 days within Indian jurisdiction.

    What it means for AI

    The AI tools on a delivery floor produce ICT logs. A recorded trail that stays in the deployment, per client, is the material for the log duty. Reporting stays with you.

  5. TRAI Telecom Commercial Communications Customer Preference Regulations, Third Amendment 2026 (05 of 2026), 18 September 2026

    Issued in September 2026

    Read the source

    What it requires

    Defines an A2P call as a voice call placed by an application or automated platform without direct human dialling, including robo-calls and artificial-voice technologies. Senders declare A2P use in advance.

    What it means for AI

    Our contact-centre work today assists a person on the call. An automated voice agent placing calls is what this definition describes, so read it before planning one.

This is what the rules say, not legal advice. Your compliance function owns the reading and the filing.

WORK SCENES

What it looks like on a working day.

The problem, what runs, and what the person at the desk gets. Each scene is an illustration of how the work is designed to go, not a customer account.

Scene 01

A card number said aloud on a call

The problem

A customer reads a card number to an agent while an assist tool listens, and the number must not reach a model.

What runs

Aavaran masks card and ID numbers in the transcript before the model reads it, and Seva suggests the next step.

What they get

The agent sees a suggested reply and compliance flags, and the client's record shows a signed Trust Receipt for each masking call.
Masking in the contact centre

Scene 02

Quality on every contact, not a sample

The problem

A quality team can review only a few calls, so most go unseen.

What runs

Seva returns a summary and a quality scorecard with compliance flags for each interaction.

What they get

The team leader opens the flagged contacts first and reads the scorecard.
Seva

Scene 03

The same workflow for two clients

The problem

What one client allows, the next one forbids.

What runs

Each client is its own workspace with its own agents, chosen model, stored keys and audit events.

What they get

Each client downloads its own record as CSV, filtered by person, action and date.
One workspace per client

Scene 04

A helpline that takes voice notes

The problem

A client's helpline receives voice notes in Hindi, Tamil, Telugu and Marathi, and the recordings are sensitive.

What runs

AgentAnywhere Vaani is designed to understand, mask, classify and route each note on Indian infrastructure, keeping only an anonymised tally.

What they get

The right queue receives a masked, categorised note, and urgent or uncertain ones go to a person.
AgentAnywhere Vaani

Scene 05

Client code that cannot leave

The problem

A client's contract forbids sending its code to an outside tool.

What runs

Taksha runs on your hardware or in the client's environment, and an independent referee checks each claimed success.

What they get

Your reviewer receives each change with the referee's verdict, and your release process decides what ships.
Sovereign coding

THE SOVEREIGNTY TEST

The sovereignty test, for BPO and IT services.

If an AI system cannot answer these from its own records, you are renting it. Beside each question: what a rented AI service usually answers, and how we answer. Read the test.

  1. 01Where does the data run?

    What a rented AI service usually answers

    “In a region near you, per the contract.” The vendor decides when the model changes.

    How we answer

    In your cloud or your client's, in an Indian region or on hardware the contract names, region-pinned and designed to deploy air-gapped. One workspace per client keeps their data and rules apart.

  2. 02Which model made that decision?

    What a rented AI service usually answers

    “The AI.” The version is in the vendor's release notes, not in the client's file.

    How we answer

    Model Hub is designed to register each model with an owner and lineage, and each client workspace has its chosen model. Seva and Taksha are built on open foundation models, measured and served by us.

  3. 03What did the model actually see?

    What a rented AI service usually answers

    The full transcript, with the card number in it.

    How we answer

    Aavaran masks card and ID numbers in the transcript before the model reads it, with a signed Trust Receipt for each masking call.

  4. 04What stops the agent doing harm?

    What a rented AI service usually answers

    A prompt that asks the model to follow the client's rules.

    How we answer

    The client's policy is applied outside the model, at the Agent Gateway, and Kavach is designed to screen each request for prompt injection and unsafe output.

  5. 05Who approved the consequential step?

    What a rented AI service usually answers

    “The system.”

    How we answer

    A person approves what goes out to a customer, and the approval is recorded in that client's workspace.

  6. 06What will you show the regulator?

    What a rented AI service usually answers

    A slide in the quarterly review.

    How we answer

    A workspace's audit events, downloaded as CSV, with a signed Trust Receipt for each masking call.

STATED PLAINLY

What this is not.

  • Seva is shown on synthetic conversations. Automated resolution of routine contacts is work we build with you, with a person approving what goes out.
  • We describe Seva and AgentAnywhere Vaani without accuracy figures. AgentAnywhere Vaani's first pilot deployment is being built, and its speech recognition and synthesis are open-licensed external components used for now.
  • The DPDP carve-out in section 17(1)(d) is narrow. It does not cover Indian customers' data, and whether any of your work falls inside it is for your counsel.
  • The referee checks Taksha's work. Your own review and release process still decides what ships.
  • Masking reduces what a model reads. It does not take data outside any law.

THE OFFER · THE SOVEREIGN PILOT

Bring your hardest workload. We come to your data.

Pick the one workload your risk committee keeps deferring because the data cannot go anywhere. We run the pilot where the data already sits, with our architects beside your team, and you decide at the end whether it goes live.

For BPO and IT services firms: bring the client account with the hardest contract. We set up inside that client's boundary, and leave a record the client can inspect.

What the pilot is

  • One workload, one team, six to ten weeks.
  • A sovereign environment by default.
  • Solutions Architecture paired with your team for the duration.
  • It ends with one agent in production under your governance, or a documented decision not to proceed.

What we add

  • The Trust Platform is on from the first call: data masking, policy screening, the Agent Gateway and a recorded trail.
  • The record stays with you and is exportable as CSV.
  • You can bring your own open model.
  • A reply within one business day.

A risk committee has no reason to refuse, because the data does not have to go anywhere to find out. The pilot ends one of two ways: one agent in production under your governance, or a documented decision not to proceed, with the record behind it.

FAQ

Frequently asked questions.

Can a BPO use AI on client data without breaching the contract?
Only where the contract lets the data go. AgentAnywhere is designed to run in your cloud or your client's, region-pinned, with Aavaran masking sensitive fields before a model reads them, so you can show a client what the model saw.
Does the DPDP Act exempt BPOs?
Only in a narrow case, and only from May 2027. Section 17(1)(d) disapplies most of Chapter II and Chapter III and section 16 for personal data of people outside India, processed in India under a contract with a person outside India. The security safeguards in sections 8(1) and 8(5) still apply, and Indian customers' data is not covered. Your counsel decides how it applies to you.
How do we keep one client's data and rules apart from another's?
Each client is its own workspace, with its own agents, chosen model, stored keys, runs and audit events. What one client permits does not carry into another client's account.
How does a BPO show a client what the AI did on their work?
Every step is recorded, each masking call carries a signed Trust Receipt, and a workspace's audit events can be filtered and downloaded as CSV. The client can inspect its own record.
What can Seva do on a contact-centre floor today?
Seva reads a service conversation and returns a summary, a quality scorecard, compliance flags and a suggested reply for a person to use, in Indian languages and English mixed the way customers speak. It is shown on synthetic conversations. Automated resolution of routine contacts is work we build with you. Seva is built on open foundation models, measured and served by us.
Do the new TRAI rules on automated calls apply to AI voice agents?
TRAI issued amendment regulations in September 2026 that define an A2P call as a voice call placed by an application or automated platform without direct human dialling, including robo-calls and artificial-voice technologies, and senders declare A2P use in advance. An automated voice agent placing outbound calls is what that describes. Assist and review tools that sit with a person on the call are a different case. Your compliance function should read the regulation before planning automated outbound voice.
Can a delivery team use AI coding tools on client code?
Only where the client's contract allows the data to go. Taksha runs on your own hardware or in your client's environment, region-pinned or designed to deploy air-gapped, so the code is designed to stay inside the boundary the contract draws. An independent referee checks its work, and your own review decides what ships.
What does the Sovereign Pilot involve for a BPO?
One workload, one team, six to ten weeks, in a sovereign environment, with Solutions Architecture paired with your team. It can run inside one client's boundary. The Trust Platform is on from the first call, the record stays with you and is exportable as CSV, and it ends with one agent in production under your governance or a documented decision not to proceed.

Bring your hardest workload.

We come to your data. One workload, one team, six to ten weeks, in a sovereign environment. Tell us the workload and a solutions architect will reach out within one business day.