AI is not the future anymore; it’s already here. At Razorpay, we’ve embraced it wholeheartedly, especially in our Tech Writing team. With a bunch of curious minds and tireless tinkerers, we set out to see what AI could really do for us. The results? A whopping 40% boost in efficiency across our workflows. Curious how we pulled this off? Let’s walk you through our AI journey.
Our Experiments With AI
Listing down the top 5 areas in tech writing where we focused on automating with AI.
Content Creation: From Drafts to Docs
We began with a simple question: Can AI help us write better and faster?
Well, of course, AI can write. But can it write like a Razorpay Tech Writer? With the right training, it can.
We tested popular tools like Claude, ChatGPT, Gemini, Co-pilot, and more. Surprisingly, some were better at writing product documentation, while others excelled at API and integration docs.
We trained the AI models with:
- Our internal style guides
- Real doc samples
- Detailed checklists
- Carefully crafted prompts
With every iteration, our prompts got sharper, and the output got better almost by 3 times.
We even built a shared Prompt Library, so no one starts from scratch. New or old, every writer now has a tested set of prompts to work with.
AI now generates first drafts that are nearly 90% accurate. All it needs is a layer of human review: compliance checks, branding consistency, tone alignment, and factual accuracy. Once that’s done, it’s ready to go live.
Media Creation: Images, Diagrams, Videos—AI’s Got This Too
Our exploration didn’t stop at words. We dove into visuals too, and discovered that AI tools can help us:
- Create diagrams and conceptual images
- Generate text from images or diagrams
- Record screen tutorials
- Produce human-like voice-overs
- Make short concept videos
With these, we’ve slashed media creation time by 50%. Check out the voice-over in this video — it is AI-generated!
Tech + Content: Coding Our Own Solutions
Earlier, making structural changes like updating the docs homepage or menus meant waiting on engineers. Not anymore.
Working with our dev team, we created Cursor rules that let us independently update: Navigation menus, Homepage content, Sidebar links and more. This has made us more agile and self-reliant.
Automating Reports
Tracking traffic and engagement used to take up days. Now, AI does it in seconds.
Our automated system pulls usage reports from Google Analytics, helping tech writers understand their audience better—no more manual effort.
AI Agents for Change Notifications
Imagine a Slack bot that listens to team conversations and nudges writers when a doc update might be needed. Sounds sci-fi? We’re almost there.
We’re building AI agents that monitor internal communications and alert tech writers about possible changes or updates to documents, even before a request is raised. No more last-minute surprises. No more product launches without docs.
These are still being tested, but are showing promise.
The Other Side of AI
Of course, not everything about AI is magical. Here’s what we learned the hard way:
- No single AI model does it all. You’ll need a combination of tools.
- AI struggles with visual content unless explicitly asked to interpret it.
- It can’t read dynamic text or content rendered through code on websites.
- Many tools are powerful but expensive, especially for team-scale usage.
- Free plans are great for experimentation, but lack depth for production use.
- Most tools aren’t compliance-ready, especially in regulated domains like fintech.
- And yes, AI still hallucinates. Always review before publishing.
Final Verdict: So, Is AI Worth It?
Absolutely, if you know what to expect.
AI can supercharge your documentation process. It can handle repetitive tasks, accelerate writing, and even help you become more independent from other teams.
But it’s not a plug-and-play solution. It needs:
- The right setup
- Continuous experimentation
- A strong review layer
- And yes, budget.
At Razorpay, we’re still exploring, refining, and learning—but we’re already seeing the impact.
