About
I am Vishal Kale, an AI product manager based in Abu Dhabi. I build and operate AgentFlow, a platform for creating AI agents that talk to customers by chat and by phone.
Who I am
I have spent twelve years building software and, latterly, owning it. I started as a software engineer on healthcare systems in Pune, moved into leading development, and then into product — first as Chief Product Officer at Axonic Health (CliniVantage), where I built an AI-powered healthcare platform that served more than 1.3 million patients across 300+ hospital deployments.
Today I am Product Manager - AI at Miral Experiences in Abu Dhabi, where I lead enterprise AI and digital transformation across ERP, CRM, conversational AI and intelligent automation. I hold PMP® and Certified Scrum Product Owner (CSPO®) certifications and an MBA in Information Technology.
- Name
- Vishal Kale
- Role
- Product Manager - AI, Miral Experiences
- Location
- Abu Dhabi, United Arab Emirates
- kalevishal96@gmail.com
- linkedin.com/in/vishal-kale1
- GitHub
- github.com/vishal-kale1
AgentFlow
AgentFlow is the application I develop and operate. It is a multi-tenant platform that lets a business build, test and publish AI agents — a chat widget on their website, a voice agent in the browser, or an agent that answers the telephone.
A customer signs up, describes what their agent should do, and gives it three things:
- Knowledge — documents and pages they upload, which the agent searches and cites when it answers, rather than guessing.
- Tools — a strict allow-list of actions the agent may take, such as checking calendar availability, booking an appointment, capturing a lead or handing over to a human.
- A flow — the conversation design, built on a visual canvas, with conditions, error paths and per-step testing.
Once published, the business can watch every conversation in a shared inbox, complete with transcripts, the tools the agent used and the sources it cited.
Why AgentFlow asks for Google access
Appointment booking is one of the most common things a business wants its agent to do, so AgentFlow can connect to a customer's Google account. This is entirely optional: agents work without it, and nothing is read from Google unless the account owner explicitly connects their calendar.
When connected, AgentFlow requests only what booking requires:
| Scope | What it is used for |
|---|---|
openid, email, profile |
Signing in to AgentFlow with Google — name, email address and profile picture. |
calendar, calendar.events |
Reading free/busy times so the agent only offers slots that are genuinely open, and creating, rescheduling or cancelling the booking it agrees with the caller. |
gmail.send |
Sending the booking confirmation to the person who booked, from the business's own address. |
AgentFlow does not read a customer's inbox, does not use Google data for advertising, and does not use it to train any model. The full detail — storage, encryption, retention and how to disconnect — is in the privacy policy.
Contact
Questions about AgentFlow, about anything on this site, or about a privacy request: email kalevishal96@gmail.com. I aim to reply within a few working days.