Case Studies

Real bots. Real people.
Real work getting done.

Each bot below is live, used by real users, and would be missed if it stopped working. They’re shaped completely differently — and together they show what’s possible when an agent is built to ship, not to demo.

Case 01 — Multi-tenant SaaS

Barty

The WhatsApp bot for residential building committees / HOAs — by Menti Technologies.

The problem

Every building has the same chaos. A WhatsApp group with 40 neighbors. The committee chair fielding the same questions for the hundredth time — when does the cleaner come, who do we call about the boiler, how much do I owe this month. Tasks that fall through the cracks. Payment reminders nobody wants to send manually.

What Barty does

Lives inside WhatsApp — where the building already talks. Answers questions about the building 24/7. Tracks open issue tickets. Sends building-dues reminders automatically. Each building has its own private context — its own rules, contacts, and history.

Under the hood

Multi-tenant architecture

One platform, many buildings, full isolation between them.

WhatsApp Business API

Production messaging infrastructure, not a chat widget.

Native-language NLU

Understands how real people actually write WhatsApp messages, typos and all.

Reactive + proactive

Answers when asked, reaches out when something needs attention.

Where this pattern fits

Anywhere a small group of people repeats the same questions, chases the same tasks, and already lives inside a messaging app:

  • HOAs and property management — same shape, different building
  • School parent groups and community managers — recurring questions, payment cycles, light coordination
  • Franchises and multi-location SMBs — one bot per location, central platform underneath

If your users already live in a group chat, that group chat can become a product.

Case 02 — Personal AI

Mazker

Mazker (Hebrew for “reminder”) — your personal assistant. By Menti Technologies.

The problem

Reminder apps are dumb storage. You pick the date, you pick the time, you type the text — and they parrot it back. That’s not an assistant; it’s a sticky note with a clock. Real life is messier: “remind me to call my dad the day before his flight,” or “every workday at 9, but not when I’m in a meeting.”

What it does

You write to it the way you’d ask a person:

“Remind me to take my vitamins every morning.”
“Day before my flight to Berlin, remind me to check in.”
“Every first of the month, remind me to send the invoice.”

It figures out the schedule, picks sensible times (not 2am, not while you’re in a meeting), and lets you reply “done,” “snooze 1hr,” or “reschedule to tomorrow” in plain language. No forms. No date pickers. No app to install — it lives where you already type.

Who uses it

Right now, me and the people closest to me — family and friends. It’s been quietly running in real lives every day, which is the strictest QA there is. The bugs you find when your mother uses your bot are the bugs that matter.

Under the hood

Natural language understanding

For fuzzy time expressions like “day before” or “every workday.”

Context-aware delivery

Knows not to interrupt at the wrong moment.

Two-way conversation

Replies are commands, not just acknowledgments.

Persistent memory

Recurring patterns survive between messages.

Where this pattern fits

Any product that today asks users to fill in a form when it could ask them a question:

  • Habit, wellness, and medication adherence — proactive nudges beat passive logs
  • Coaching, accountability, and onboarding sequences — conversation, not checklists
  • Internal ops — compliance deadlines, renewals, sales follow-ups, reviews

From “the user does the work” to “the bot does the work, and only asks when it needs to.”

What both bots have in common

Two completely different bots.
Same underlying discipline.

Both ship to the channel where users already are. Both treat memory as a first-class feature, not a bonus. Both do work, not just talk. Both have been running long enough that breaking them is no longer a hypothetical.

If your product has forms, dashboards, or repetitive support questions, there’s a bot pattern hiding inside it. The interesting question isn’t can we build it? — it’s what should the bot do, and what should still stay a button?

Get In Touch

Start a
Conversation

Roy personally reads every message. Tell him about your challenge and he'll suggest the best path forward — no commitment required.

Response time
Within 24 hours

Your message goes directly to Roy. Expect a personal reply within 24 hours.