A CA we work with — Vikram, Bengaluru, sole practitioner, ~120 clients — used to keep his phone face-up on the dinner table. Not because he wanted to. Because if he didn't reply within 20 minutes to a client's "ek baat puchni thi" message, the next message would be the same client calling at 11pm with a question that could absolutely have waited until 9am the next morning.
What he set up after we built the desktop app was unremarkable. An after-hours auto-reply. Greeting between 7pm and 9am. One paragraph. The phone could stay in another room.
What was remarkable was the second-order effect.
The thing that actually changed
Within two weeks Vikram noticed three things, in this order:
Clients stopped chasing. When the first message they send at 10pm gets an instant reply — even an automated "thanks, we're closed, will respond at 9am" — they don't call. They don't send three more messages. They put the phone down. The unread anxiety on both sides goes away.
Late-night urgency dropped. Most "urgent" client messages aren't actually urgent. They feel urgent because the client doesn't know if anyone's listening. Once the auto-reply tells them they were heard, the perceived urgency drops. Real emergencies — GST deadline tomorrow, notice received yesterday — stayed urgent. The rest stopped pretending to be.
Vikram's mornings got better. He used to wake up to 12-15 unread messages and a queasy "what did I miss" feeling. After the auto-reply, the messages were still there, but the implied accusation of unresponsiveness wasn't. Same volume of work, lower cortisol.
This is not a feature pitch. It is, to my knowledge, the single highest-leverage thing a small CA firm can set up on WhatsApp, and most haven't.
What "good auto-reply" actually looks like
The bad version: "Hi, I'm not available right now, please leave a message." Universally hated, signals nothing, creates no value.
The version we run on the AI4CA desktop app, slightly tuned for each firm:
- After-hours (7pm–9am, Sun all day): Acknowledge receipt, set the response window, give an escape hatch for true emergencies. "Thanks for the message. Office is closed until 9am tomorrow. We respond first thing. If it's deadline-tomorrow urgent, reply 'urgent' and we'll get a partner on it."
- First message from a new number: Treat them as a prospect, not a known client. Ask the three questions you'd ask anyway in the first call: "Welcome — to help fast: (1) your name and firm, (2) what you need help with (GST / ITR / advisory / audit), (3) your city. We'll get back within business hours."
- Keyword triggers for repeat questions: "fees" → standard fee bands, "documents" → checklist, "office address" → address + Maps link. Saves your assistant 40 minutes a day.
The keyword triggers are where most of the time-saving lives. About 60% of incoming questions to a typical CA firm are the same five questions, asked by different people. Reply to them once, properly, and let the system answer them after that.
What we built into the desktop app
The auto-reply engine in the AI4CA desktop app does three things:
- Time-windowed rules. Different replies during business hours, after hours, and weekends. The engine reads your machine's timezone, so if you're in Pune the rules trigger on IST regardless of where the WhatsApp servers think you are.
- Keyword matching with priority ordering. "GST notice" matches before "GST" so the more specific reply fires. Multiple match modes: exact, contains, regex (for the technically inclined).
- Per-contact suppression. If you've replied manually to someone in the last 60 seconds, the auto-reply won't fire — it would feel weird to the client to get your real reply followed by a robot. Cooldown timer per contact prevents auto-reply spam to people who message you several times in a row.
These three together are what make it feel less like a chatbot and more like a quiet, well-trained junior. The client doesn't notice it's automated. They notice they got a fast, useful reply.
What this is not
It's not a chatbot. It's not AI-driven. It's not going to file a GSTR-3B for someone. The auto-reply is a thin, deterministic layer that handles the 60% of incoming traffic that's repetitive, so you can give the other 40% your full attention.
We have customers who tried bigger, fancier "AI WhatsApp assistants" before us. The pattern was the same: enthusiasm in week one, frustration in week three when the AI made a wrong call about a tax matter, removal in week four. Smaller, deterministic, predictable beats bigger, AI-flavored, occasionally wrong — every time, in regulated work.
The setup
If you're going to do this, the realistic path:
- Install the desktop app, scan QR with your existing WhatsApp number (no separate business number needed).
- Spend 20 minutes writing your after-hours rule and three keyword rules. Use your real voice, not "we appreciate your business."
- Turn it on. Watch what happens for a week.
- After a week, check the logged messages. Tweak the keyword rules based on the questions you actually got. Add 2-3 more keywords.
- After two weeks, you'll have a setup that handles most of the repeat traffic and you can put the phone in another room at 8pm.
That's the whole thing. No agency, no implementation engagement, no "transform your practice with AI" — just a system that catches messages while you're at dinner.
The auto-reply engine is part of the AI4CA WhatsApp desktop app. Free tier supports unlimited rules and basic keyword matching. ₹499/month adds bulk send, full CRM logging, and the time-window scheduler. Connects to your existing WhatsApp number — no separate account, no per-message fees.