Deployed AI is the practice of building AI agents inside a company's real workflow and running them there, rather than delivering a tool and leaving. Vikrama Studio audits the workflow, builds one agent at a time on the company's data, and stays to run it.

How is this different from an AI automation agency?

Automation agencies wire tools together with triggers. We build agents that read, decide and draft, with an approval step and an eval, and we price the outcome, not the hours.

What do you need from us to start?

Read access to the systems in scope, one owner on your side, and two weeks for the audit.

Which models and tools do you use?

Frontier models through their APIs with zero-retention settings where offered, your existing ERP, accounting and CRM systems, and our own orchestration. We do not sell a platform.

Do you work with companies outside India?

Yes. Most Deployed AI work is for US companies and accounting firms. Calls run in ET, GST and AEST.

What happens after the pilot?

Either your team runs it with our documentation, or we stay on a monthly run retainer and queue the next workflow.

Deployed AI

Agents that do real work inside your business. Built on your data, approved by your people, run by us until your team owns them.

[ 00 ] In short

Deployed AI is Vikrama Studio's applied AI practice. We build and run AI agents for finance and operations teams: month-end close and reconciliation, invoice and document extraction, quoting from RFQs, lead qualification, reporting. Every system ships with an approval step, an audit log and evals. A pilot takes six to eight weeks and starts at $20,000.

Sandeep Kumar, Co-founder and CTO · Updated 2026-09-12

Key facts

Pilot length: 6 to 8 weeks, one workflow, software running on your data.SRC · Vikrama Studio method · 2026-09
Approval design: Your controller or ops lead approves before anything posts.SRC · Vikrama Studio method · 2026-09
Pricing: Audit from $5,000. Pilot from $20,000. Run from $5,000 a month.SRC · Vikrama Studio price list · 2026-09
Dizios beta: 10 studios on real data.SRC · Dizios beta · August 2026
Lead engineer on every build: Sandeep Kumar, Co-founder and CTO.SRC · Vikrama Studio · 2026-09
Where we work: Gurugram, with call hours in ET, GST and AEST.SRC · Vikrama Studio · 2026-09

The engineer on every build

Sandeep Kumar
Co-founder and CTO
Forward-deployed AI
LinkedIn ↗

Three rules every build follows.

One. Nothing posts without a human approval step, and the step is designed before the agent is.
Two. Every agent has an eval. If we cannot measure it, we do not ship it.
Three. Your data stays yours. No training on client data, ever. Deletion on exit, in writing.

Hours · Tuesday to Thursday, 8:00 to 12:00 ET, arranged on WhatsApp

WA ·+91 63955 52618

What we build

Six offers. Each starts with a two-week audit and a fixed price.

Audit, Build, Stay

One practice, three stages. We audit before we build, and we stay after we ship.

[ 01 ]

Audit

Two weeks inside your workflow. We leave with the three tasks agents should take first, an approval design, a plan and a fixed price.

[ 02 ]

Build

One workflow, one pilot, six to eight weeks. Software running on your data, with evals on a held-out period before any demo.

[ 03 ]

Stay

We run it. Monitoring, evals every month, exceptions handled, the next workflow queued. Until your team can, or for as long as you want us to.

How we build →

What an agent looks like when we build it

The Matcher from the finance system: a spec sheet, an approval step and an eval before it runs on your books.

Agent / spec

Matcher

Purpose: Matches lines to invoices, proposes splits for partial payments, flags exceptions with a reason.
Inputs: Bank feed or statements · AR and AP ledgers · Remittance emails
Output: A matched set and an exception queue.
Approval: Controller clears exceptions.
Eval: Auto-match rate, false-match rate on a held-out month.

The system it belongs to

The orchestrator runs three agents. Their work stops at the approval step before anything posts to the ERP.

Finance operations system, architecture

Use cases

Eight workflows agents take first.

Industries

Accounting, bookkeeping and fractional-CFO firmsClient close and reconciliation, Invoice extraction, Monthly pack assembly
D2C and ecommerce brandsAgentic storefront readiness, Order and returns extraction to the ledger, Reporting
Real estate teams, brokerages and boutique staysSpeed-to-lead, Booking and enquiry qualification, Monthly reporting
Manufacturers and industrial suppliersRFQ-to-quote, PO and invoice extraction to ERP, Inbound enquiry qualification
Gyms, yoga, dance and wellness studiosMember enquiry agent, Retention insights, Billing reconciliation
Banks, NBFCs and fintechsReconciliation at volume, Document extraction for onboarding, Risk and credit model operations

Security

Read-only access first, scoped credentials, no training on your data, an audit log on every agent action and deletion on exit, in writing.

How we handle your data →

Questions

What is Deployed AI?: Deployed AI is the practice of building AI agents inside a company's real workflow and running them there, rather than delivering a tool and leaving. Vikrama Studio audits the workflow, builds one agent at a time on the company's data, and stays to run it.
How is this different from an AI automation agency?: Automation agencies wire tools together with triggers. We build agents that read, decide and draft, with an approval step and an eval, and we price the outcome, not the hours.
What do you need from us to start?: Read access to the systems in scope, one owner on your side, and two weeks for the audit.
Which models and tools do you use?: Frontier models through their APIs with zero-retention settings where offered, your existing ERP, accounting and CRM systems, and our own orchestration. We do not sell a platform.
Do you work with companies outside India?: Yes. Most Deployed AI work is for US companies and accounting firms. Calls run in ET, GST and AEST.
What happens after the pilot?: Either your team runs it with our documentation, or we stay on a monthly run retainer and queue the next workflow.

Talk

Talk to Sandeep, or message us.

WA ·+91 63955 52618 Free written teardown →

Hours: IST 10:00 to 19:00, Monday to Saturday. Calls in ET, GST and AEST, arranged on WhatsApp.
Where we work: Gurugram, with call hours in ET, GST and AEST.
Free written teardown: No call. Delivered in writing in 48 hours.