Razorpay's Slash AI: Inside the Fintech Giant's Agentic GTM Playbook

GTM Deep Dive

How Razorpay built an internal AI coworker that processes 14,800+ tasks a week — and what it means for the future of GTM teams.

14,800+ Weekly Tasks

2,150+ PRs/Week

₹3,783 Cr FY25 Revenue

45% No-Rework PRs

The Story

Razorpay cofounder Shashank Kumar recently revealed how the fintech giant internally built an AI assistant called Slash, now deeply integrated across the company’s workflows. Accessible directly through Slack, Slash can read Razorpay’s codebase, debug production incidents, write code, analyse logs, and raise PRs autonomously.

In its first week, Slash handled 122 tasks. Six weeks later: 14,800+ tasks per week. That’s a 121x growth in adoption in under two months.

The Old Way

Fintech companies traditionally operated with rigid engineering hierarchies. A support ticket → triage → senior engineer → fix → QA → deploy. Each handoff introduced delays. GTM teams waited days for product fixes, feature requests got buried in backlogs, and the cost of internal tooling ate into margins.

Razorpay, like most fintechs, had separate teams for infrastructure, backend, frontend, data, and security. Cross-team coordination was slow. GTM experiments that required engineering support had 2-4 week lead times.

The New Way

Slash changed the equation entirely. Now, anyone at Razorpay can ask Slash to investigate a production issue, optimize a query, or ship a feature — directly from Slack. The 14,800 weekly tasks include:

• Incident resolution — Slash reads logs, identifies root cause, applies fix
• Kubernetes optimization — Auto-scaling, resource allocation, cost reduction
• Testing & security — Automated test generation, vulnerability scanning
• SQL pipelines — Product and analytics teams self-serve data requests
• Research & automation — GTM teams use Slash for competitive research, CRM automation

GTM Implications

For GTM teams, the implications are profound. When engineering bottlenecks disappear, GTM velocity increases:

• Sales enablement: Custom demos, POCs, and integration support in hours instead of weeks
• Marketing automation: SEO analysis, content personalization, A/B testing at AI-speed
• Customer success: Real-time issue resolution without engineer handoffs
• Revenue operations: Automated pipeline analysis, forecast modeling, CRM enrichment

“The question was never how do we do the same with less. It’s what we can build that was previously impossible.”

— Beerud Sheth, CEO, Gupshup (on AI-driven productivity)

The Bigger Picture: AI as GTM Leverage

Razorpay’s Slash is part of a larger pattern. Across the ecosystem:

• Freshworks cut 500 roles while citing AI-led productivity — but also redirected spend toward higher-priority GTM verticals
• Coinbase reduced workforce by 14% while positioning as an “AI-native” company
• Oracle reallocated capital from headcount to AI infrastructure and cloud expansion

Andrew Ng, AI researcher and Coursera cofounder, argues that the “AI jobpocalypse” narrative is disconnected from reality. Companies have a strong incentive to frame layoffs as AI-driven — it makes them look forward-thinking. The reality is more nuanced: pandemic-era overhiring, rising capital costs, and genuine AI productivity gains are all happening simultaneously.

Key Takeaways for GTM Leaders

• Ship GTM experiments 10x faster — When AI handles the engineering overhead, your iteration cycle shrinks from weeks to hours
• Self-serve data & research — GTM teams can run their own queries, research, and analysis without waiting for data engineering
• AI is a force multiplier, not a replacement — The companies winning are the ones using AI to expand, not just cut costs
• Build the bridge early — Razorpay’s Slash took 6 weeks to go from 122 to 14,800 tasks. The adoption curve is steep when you remove friction
• Measure what matters — 45% of Slash’s PRs ship without rework. That’s not just productivity — it’s trust. Build systems that earn it.

“For companies with vision, AI-driven superproductivity unlocks entirely new markets and product lines. AI is fundamentally a growth engine, not a cost lever.”

— Beerud Sheth, CEO, Gupshup

GTM Framework: The AI-Native GTM Flywheel

Step 1: Remove engineering friction — AI coworker handles infra, data, and code tasks (like Slash)

Step 2: Accelerate GTM experiments — Faster iteration on messaging, channels, and segments

Step 3: Compound learnings — AI captures and synthesizes every experiment result

Step 4: Expand surface area — New channels, new segments, new products become viable

Step 5: Repeat — Each cycle compounds speed and intelligence

Sources

• Shashank Kumar (Razorpay Cofounder) — X post on Slash’s adoption metrics
• Inc42 — “AI Jobpocalypse Or Market Correction?” — May 20, 2026
• Andrew Ng — LinkedIn post on AI and layoff narratives
• Freshworks, Coinbase, Oracle — public statements on AI-led restructuring
• Beerud Sheth (Gupshup CEO) — Interview with Inc42