AI Oversight WhatsApp SaaS

Whassist

A 15-person sales team was running on WhatsApp — and losing 20-30% of revenue to chats nobody could read. We built the AI agent that sees what the CEO can't.

Client UK Sales Agency · Whassist SaaS  ·  Live at wh-assist.com →
Case Study Video
Video Breakdown · Coming Soon
30s
Onboarding (QR Scan)
70k+
Messages Per Org
$2k/mo
Leak Caught On Day One

You Can't Manage What You Can't Read.

Lead lost

Our client runs a sales agency in the UK. Setters, closers, supervisors, managers. Everyone communicates through WhatsApp group chats and DMs. Dozens of active conversations at any given time.

If you're running a team like that, you have a visibility problem. You're paying fifteen, twenty people. They're having hundreds of conversations a day. And you have no idea what's actually going on in those messages. Are your setters following scripts? Are supervisors training new hires? Are leads being followed up with?

His estimate: 20 to 30 percent of revenue was falling through the cracks. Leads going cold. Bottlenecks invisible. By the time he found out, the money was gone.

The only way to know is to read everything — and that's physically impossible. Most CEOs running teams on WhatsApp just don't. They trust and hope.

"What you can't see on WhatsApp is costing you money."

He'd been pasting his WhatsApp exports into ChatGPT and asking questions. He wanted an AI that did this automatically. Reads every message. Understands context. Tells him what matters. Summaries, search, performance tracking. And not just for himself — he planned to launch it as a SaaS product.

Every Agency Pitched The Same Thing. Every One Of Them Was Wrong.

Meta Forms
QR · 30s

Before us, he'd spoken to other agencies. Every single one pitched building it on the official WhatsApp Business API. On paper that makes sense — it's Meta's official tool. In practice, it kills the product.

Why The Official API Killed The Product
  • Meta business verification for every user
  • Company registration + docs + phone verification
  • Rate limits + template-only messaging
  • No access to full chat history at scale
  • Built for support bots, not org-wide oversight
What We Did Instead
  • Open-source bridge — same path as WhatsApp Desktop
  • User scans a QR code, done in 30 seconds
  • No verification, no rate limits, no template walls
  • Full access to every message, group, contact
  • The discovery that made the entire product possible
For a SaaS where someone signs up and connects their WhatsApp in five minutes? Filling out Meta compliance forms is a dealbreaker. Nobody does it just to check what their team said today.

Every agency trying to mix WhatsApp with AI hit this same wall. They either didn't know another path existed, or didn't know how to get around the restrictions. We went deeper — and that's the only reason this product exists.

Access Was Step One. Accuracy Was The Real Work.

whassist · query
> what did mark do this week? [search] 12 conversations · 3 groups [filter] Mark + last 7d · 24 messages [verify] 1 person matched · Mark followed up on 4 of 7 hot leads.

You've got access to 70,000 messages. Simple now, right? Feed it into ChatGPT and ask questions. That doesn't work — and this is where experienced developers get it wrong.

PROBLEM 01
Data Compliance
Private conversations. Employee messages. Business secrets. You can't just ship it all to an external model. Architecture has to keep data controlled before any agent logic exists.
PROBLEM 02
Scale Beyond Context
70k messages don't fit in any model's memory. Dump them in and the AI silently ignores most of them and gives a confident wrong answer.
PROBLEM 03
Accuracy Or Nothing
A CEO making real decisions can't accept hallucinations, blended conversations, or confused names. Verification layers, not vibes.

We built a retrieval system that finds the right messages for each question. Ask "what needs my attention today?" and it pulls only what matters. Ask "is this employee doing their job?" and it surfaces relevant exchanges across every group chat they're in.

If the AI isn't sure, it asks a clarifying question instead of guessing. Time-filtered, conversation-isolated, role-aware. CEOs, managers, and team members each see only what they're allowed to.

Onboarded Two Managers. Caught An Insider On Day One.

Is this supervisor working with someone outside my team?
Cross-referenced 6 chats over 3 weeks.
Pattern: side operation for a competing agency.
Fire him today.

Nobody expected the AI to pay for itself this fast.

Our client had onboarded two managers and a supervisor. He was testing the system, trying to break it. He asked the AI to summarize what one of his supervisors had been doing.

Something came up unusual. The supervisor was talking about someone who didn't appear to be part of the agency. He pushed further: "Is this person working with someone outside of my team?"

The AI pieced it together across multiple conversations. The supervisor was running a side operation for a competing agency. Full-time employee. Full access to every client, every lead, every internal process.
"It's like being the recipe guy at McDonald's and going to work at Burger King."

Confronted. Admitted. Fired the same day.

$2k/mo
Revenue Leak Stopped
1
Data Breach Prevented
Day1
ROI Achieved

No human finds that by reading group chats. The signals were scattered across weeks of messages, buried in normal conversation. The AI connected it because it could search meaning across the full history — something physically impossible to do manually.

The Product Is Live.

Live
+$2k
Active
70k+
Today
12

Users sign up, scan a QR code to connect their WhatsApp in thirty seconds, and get AI-powered oversight of their entire team immediately. No Meta compliance forms. No phone verification. No template walls.

What The Product Does
  • Reads every message, every group, every DM
  • Summarizes what needs the CEO's attention today
  • Tracks performance and follow-up cadence per rep
  • Answers natural-language questions across history
  • Role-aware: CEO, manager, and team see different views
What Made It Shippable
  • Open-source WhatsApp bridge (not Business API)
  • Compliance-aware data architecture
  • Retrieval system tuned for 70k+ message orgs
  • Verification layers — clarifies instead of guessing
  • 30-second onboarding from scan to insight
We build AI agents that handle real data at real scale — with the engineering to make them accurate, compliant, and useful.

Have A Problem
That Looks Impossible?

We've shipped agentic AI products since early 2025. If the standard tools don't solve it, we go deeper.