About Automation Agents

We exist because 72% of AI pilots fail.

Vikesh, Santosh, and Nishan built products at Slack, ZScaler, and actually scaled AI systems. We exist because enterprise AI should never deliver abandoned prototypes and failed pilots. It shouldn't be this broken.

$47M+
Revenue protected for clients
340+
Agents deployed in production
99.7%
Uptime across all deployments
12
Enterprise clients served
Our story

The problem that
got us started.

Why we started

Vikesh was running TexAu when he kept seeing the same failure pattern: enterprises dumping ₹2-5 crore into 6-18 month AI pilots that never shipped. He'd built real products at Slack, Botpress, DoNotPay, Fireflies. He knew what production actually looked like. And he knew the problem wasn't models. It was everything else.

Santosh came from ZScaler and ThriveStack, where he'd built infrastructure for millions of events. That security-first, reliability-obsessed mindset. Nishan was a Principal AI Engineer, not a researcher. An actual builder. He'd taken systems from papers to revenue.

So they founded Automation Agents. Not consultants who bill by the hour. An agency obsessed with shipping production AI in weeks, not years. With evals built in from kickoff.

"Accuracy doesn't matter if you can't deploy it. And we won't deploy anything without evals tied to what actually costs you money."

What we've learned

Built in Mumbai, we realized quickly that India's enterprises had completely different constraints than Bay Area ones:

Multilingual users. 62% of support tickets are Hindi-English code-switching. Bulbul 3.0 exists for this reason.

Data sovereignty. RBI doesn't let you put data in cloud-only solutions. OpenClaw for on-premise wasn't optional.

Cost sensitivity. A rupee per inference compounds across 10M operations daily. Cost-per-token routing became obsession.

Production discipline. Enterprises that survived 2008 understand risk. They want continuous evals, not annual accuracy reports written after deployment.

That's our DNA: India-first engineering, global-grade reliability, eval-obsessed culture.

Our philosophy

Six principles we
don't negotiate on.

Evals before everything

Every system ships with continuous evals tied to your actual business KPIs. Not one-time tests. Living systems that catch problems before they cost you real money.

Business cost over accuracy

We're optimizing for what costs you money, not vanity metrics. A 99% accurate model that's wrong on your biggest transactions is worse than a 90% model that never misses.

Production or nothing

We don't build demos. We build systems that make money. If it's not in production handling real transactions, we're not done. Prototypes are just prototypes.

Model-agnostic, outcome-obsessed

We're not married to any vendor. Bulbul 3.0 for high-volume Indic work. Claude for reasoning-heavy problems. GPT-4o for multi-modal. Whatever solves your problem wins.

India-first, global-grade

Bulbul 3.0. On-premise deployment. Cost-optimized. But we're building to Bay Area reliability standards. Being in India doesn't mean cutting reliability corners.

Own the outcome

It breaks at 2am, we fix it. Not your ops team. We maintain what we ship because our reputation depends on your system staying up.

The team

Built by people who
actually shipped at scale.

Everyone here has failed at production. We've learned from those failures. That's how we know what works.

VT

Vikesh Tiwari

Co-Founder & CEO

Built TexAu. Worked at Slack, Botpress, DoNotPay, Fireflies. 50K+ companies using his growth automation products. Product-market fit intuition. Enterprise sales experience. Obsessed with shipping systems that work.

SY

Santosh Yadav

Co-Founder & CTO

ZScaler, ThriveStack. Security-first infrastructure engineer. Built platforms at ZScaler handling millions of events. Brings the reliability, compliance, and scaling mindset that enterprise AI actually needs.

NJ

Nishan Jain

Co-Founder & Head of AI

Principal AI Engineer. Built RAG pipelines, eval frameworks, multi-agent systems. From papers to revenue. He's the technical backbone of every deployment we ship.

Our methodology

The eval-first
methodology.

We deploy production AI 5-10x faster than big consulting firms. Here's the process.

1

Business Cost Mapping

Before we write any code, we ask: what does a mistake cost? Cost per error. Revenue per automation. KPIs that actually matter to your business.

2

Architecture Design

Model selection. Eval framework blueprint. Guardrails strategy. Designed for your specific constraints: cost, latency, data sovereignty, regulatory complexity.

3

Build & Deploy

RAG pipelines. Multi-agent systems. Guardrails. All production-hardened. We ship evaluated, monitored systems, not prototypes sitting in a demo account.

4

Continuous Eval

Live dashboards tracking performance against your business metrics. Automated alerts when regression happens. Monthly reports tied to your actual P&L.

5

Optimize & Scale

Cost-per-token optimization. Model routing improvements. Guardrails refinement. We maintain and evolve your system as volumes scale up.

Typical timeline: 3-6 weeks from kickoff to production. Weeks, not months. And definitely not years.

Where we operate

India-first, global-grade.

🏙️

Mumbai

Headquarters

Where we started. Our core team is here. This is where the India market focus lives, where we're building Indic language capabilities that actually work.

🚀

Bengaluru

Engineering Hub

Distributed systems and infrastructure team. We're close to the startup ecosystem and the engineering talent pool that powers production AI systems.

🏢

Delhi NCR

Enterprise Sales & Delivery

Enterprise relationships. Compliance meetings. Stakeholder management. Close to India's largest corporations and government institutions.

Let's build AI
that actually works.

Whether you're starting from zero or salvaging a failed pilot, we'll get you to production with evals tied to your business metrics. No excuses.

Schedule a call →

Free 45-minute conversation · No commitment · NDA available