Built by operators
We’re an AI studio focused on Autonomous Operations.
Enterprise operations are made of hundreds of small decisions every day — which vendor to use, when to run a discount, which route to dispatch, which account to chase first, which shift to fill, which price to set. These decisions are rarely made well. Most teams default to gut feel or to a process written eighteen months ago, and the cost compounds quietly into the P&L.
AI agents are good at decisions like these — often better than the humans who make them today. But the hard part is not building the agent. It is deciding what to delegate to it. We’re operators; we’ve run these processes. We start with the operation — the workflow, the people, the data, the constraints — and calibrate the agent to the right level of autonomy for every decision it touches.
- Factory floor
- Supply chain
- Fleet
- Customer support
- Back-office
01 — Calibration
Every operational decision belongs in one of three lanes.
Getting the lanes right is the whole game. Some decisions should be fully delegated to an agent. Some should be prepared by the agent and approved by a person. Some should never leave human hands. Knowing which is which is the work — and the part most AI vendors skip.
The agent decides; your team audits.
High-frequency decisions with narrow stakes, clear guardrails, and a full audit trail.
The agent prepares; a person approves.
Structured decisions that cross systems and end in a judgment call.
Stays human.
Stakeholder-facing, irreversible, or ambiguous. The agent assists at most.
Vendors who put everything in the first lane build systems you can’t trust. Vendors who put everything in the third build expensive dashboards. The value is in drawing the lines correctly — and only someone who has run the operation can draw them credibly.
02 — Work
Real operations. Real production systems.
We keep client detail off the public site — most of this work runs inside other companies’ operations, and it is theirs to talk about. What we can show you: who we work with, which operation, and where the decisions sit. Judge us on deployments, not decks.
Supply-chain operations with a Swiggy company.
Weekly account-health reporting for India’s largest transport-management platform — the agent drafts, a person signs off.
Supply chain and demand chain optimization.
AI infrastructure for property transaction data.
Collections operations for a debt-resolution platform.
Fleet and dealership operations.
Outbound voice operations at scale, in seven languages. Read more →
Support rebuilt end to end after two vendors failed. The full story is public. Read more →
Support triage where the customers already are — WhatsApp.
Visual quality inspection on the line. Try the same models yourself. Read more →
03 — How we work
Most vendors start with the agent. We start with the operation.
Process, team, system — then AI. In that order, on purpose. If the process is broken or the team is wrong, AI accelerates the wrong thing. We map how the work actually happens before anything gets built, and let AI earn its place decision by decision.
- 01
Diagnostic
Short and paid. We sit inside one operation — the workflow, the people, the data, the constraints — and classify its decisions into lanes. You get a written point of view, not a sales document. If AI is not the right answer, we say so and stop there.
- 02
First loop in production
One decision loop, built and shipped to a real production surface. Instrumented from day one, guardrails and audit trail included. The first loop does most of the heavy lifting for everything after it.
- 03
Scale
Adjacent decisions join the system once the first loop pays for itself. The lanes get recalibrated as trust builds — some decisions graduate from surface to delegate. Some never do, on purpose.
- 04
Your team owns it
Documentation, training, transfer. The engagement ends with your team running the system and us gone. An agent your team audits — not a black box, and not a dependency on us.
04 — Our own operation
We run our own company on the AI we build.
We run Applied AI Studio on the same systems we sell, calibrated with the same lanes. This is the one operation we can show you from the inside. If it didn’t work for us, we wouldn’t sell it to you.
Plans our content. Every Sunday it reads the analytics, audits everything already published, and writes the week’s calendar. A person reviews the plan; that is the entire human involvement.
SurfaceWrites and publishes the content Mitra plans — research, drafts, visuals, audio, publishing. Every piece on our blog shipped through it.
DelegateRuns our books. Invoices arrive in Slack, get extracted, checked for duplicates, and routed. It never approves a payment — that stays human, on purpose.
SurfaceThe governed layer the agents run on — who can read what, which tools they can touch, what gets logged. The same architecture we build for clients.
HoldBuilt and maintained the same way — sections, blog, decks. What you are reading went through the system it describes.
Delegate05 — Who is behind this
A team of operators, backed by a community of 3,000.
Applied AI Studio is built by people who have run the operations we now automate — supply chains, support floors, back offices. The studio was founded by Bala Panneerselvam: seventeen years in product and operations, co-founder of ZORP — a workflow platform Zomato, Zepto, and Porter used to run field operations — and product leadership at Bounce at 100,000 rides a day and at PaySense through its exit to PayU. ZORP went too broad, too early; that lesson is why our engagements start with one operation, not a mandate.
We give back to the ecosystem we work in. Applied AI Club — the community we run — is over 3,000 operators learning to put AI to work: talks, teardowns, working sessions. It keeps us honest. The people we teach are the people we build for.
We don’t do strategy decks. We build systems, calibrate them, and leave when your team owns them.
Institutions we work with
Contact
Bring us one operation.
Tell us which decisions eat your team’s week. We’ll tell you, in writing, where AI earns its place there — and where it doesn’t. You’ll hear back within a day.
Prefer to talk? Book 30 minutes directly.