A promise to send an invoice. A partnership discussion buried in a group. A store manager’s update that needs to reach a Google Sheet. Individually, these look like small tasks. Together, they become a significant part of running a business.
On Dilse Omni Talks, I explored this challenge with Swapnika Nag, Founder and CEO of Periskope and PerisClaw. An IIT alumna and former McKinsey consultant, Swapnika builds both companies with her husband as co-founder. Periskope automates WhatsApp operations for more than 5,000 businesses across 100 countries. PerisClaw brings an AI executive assistant to your own WhatsApp number.
I began with a simple question: what problem does this actually solve? Her answer took us from personal productivity to relationship management, privacy, and a compelling example of automation across 57 retail stores.
My biggest learning: the opportunity is not simply to answer messages faster. It is to turn conversations into completed work.
Table of Contents
1. Start Where the Work Already Happens
2. What Does an AI Executive Assistant Actually Do?
3. The Real Value Is in Important Work That Gets Forgotten
4. Why Voice Notes Are More Than a Convenience
5. Privacy Comes Before Delegation
6. The Ideal User May Not Be Tech-Savvy
7. A Retail Masterclass: From 57 WhatsApp Groups to Business Insights
8. Dashboards Should Begin With a Business Question
1. Start Where the Work Already Happens
We often imagine workplace productivity through email, calendars, and collaboration platforms. But for many founders, operators, and traditional businesses, a substantial part of work happens on WhatsApp.
That is where commitments are made, documents are exchanged, and relationships develop. Unfortunately, it is also where context gets scattered.
Swapnika described Perisclaw as an AI executive assistant that understands the context of your chats and works with tools such as email, Google Drive, Google Sheets, Notion, and Calendar.
The distinction matters. A reminder tool depends on you remembering to create the reminder. A context-aware assistant can help identify a commitment within the conversation itself.
If you promised an invoice by Friday, that promise should not depend entirely on your memory. If someone owes you a note, the assistant can bring the pending follow-up back to your attention.
For me, this connects with the broader discussion around conversational AI and customer engagement: the conversation becomes more useful when it connects to action.
2. What Does an AI Executive Assistant Actually Do?
I asked Swapnika to unpack the phrase “executive assistant.” Many people do not have one, so the title alone does not explain the value.
Her explanation was practical: remove manual, repetitive administrative work.
She illustrated this through Arjun, a power user whose work spans investing, entrepreneurship, and content creation. His use falls into three clear areas:
Meeting coordination: Handling the back-and-forth involved in arranging conversations.
Daily briefings: Pulling together scattered information so he enters meetings prepared.
Tasks and follow-ups: Bringing pending commitments back into focus.
The briefing use case stood out to me. Preparing for a meeting is rarely just about opening the calendar invite. It means remembering the previous discussion, finding the relevant document, and understanding what remains unresolved.
Swapnika compared that preparation to part of what a chief of staff does: ensuring you enter the room with context.
A Simple Framework: Coordinate, Prepare, Close
My way of summarizing these examples is a three-part framework. This is a synthesis of our discussion, not a named PerisClaw methodology:
COORDINATE: Get the meeting or next step organized.
PREPARE: Assemble the context needed to make the interaction useful.
CLOSE: Track what was promised and follow through.
Coordination → Context → Follow-through
This is a useful way to evaluate an AI assistant. Does it merely produce a response, or does it help move work through these stages?
3. The Real Value Is in Important Work That Gets Forgotten
I shared an observation from building this podcast. My guest outreach conversion rate is around 5%. Busy people may intend to respond, but the message disappears beneath everything else.
Arjun was unusually responsive, often getting back to me in the morning. After I started using PerisClaw, I noticed that morning reminders and task summaries could help explain that pattern. I could not establish that as the reason, but the connection was interesting.
It brought us to a belief I hold strongly:
“Important but not urgent is where the real opportunity lies.”
Saurabh Agrawal
Urgent work usually gets attention. The important follow-up without an immediate deadline is easier to postpone.
That is where an assistant can become valuable. It reduces the mental burden of carrying dozens of open commitments and makes them visible again.
The benefit is not simply a cleaner task list. It is greater reliability in the relationships and responsibilities that matter.
4. Why Voice Notes Are More Than a Convenience
Swapnika shared an interesting adoption story. An early investor user invoked Claw while arranging a meeting with Noon CEO Faraz Khalid. Faraz became interested because he was looking for something similar for Emaar founder Mohamed Alabbar.
According to Swapnika, three days later, Alabbar met the team in Delhi, where they helped set it up. He subsequently became a power user, using the assistant to manage relationships and understand who needed a response from him.
His preferred interface was particularly revealing: long voice notes.
“He doesn't type to Claw. He just leaves long voice notes.”
Swapnika Nag
I had also noticed that a dictated voice note could become a task. You can express what is on your mind without first organizing it into a carefully typed instruction.
That lowers an important barrier. The user does not need to become comfortable with a new dashboard before getting value.
The lesson I took away is straightforward: effective AI should accommodate how people already communicate. Voice is not an extra feature when it is someone’s natural way of working.
5. Privacy Comes Before Delegation
An assistant becomes useful when it understands context. But that immediately raises a difficult question: who can access that context?
I asked Swapnika how someone managing sensitive business relationships thinks about privacy. She said it is usually the first question people ask. Faraz had raised it before recommending the product to his boss.
She described two central design choices:
Your own WhatsApp number: PerisClaw operates on the user’s number rather than requiring conversations with a separate assistant number.
Isolated access: Each number, she explained, is associated with an isolated, encrypted container intended to restrict data access to the user and their agent.
Swapnika also said this architecture limits the team’s ability to inspect issues directly, so the agent has tools to help diagnose problems. She cited ISO certification and GDPR compliance as part of their security approach.
These are the company’s stated safeguards. My takeaway is not that operating on your own number automatically resolves privacy. It is that data access, isolation, and support processes belong in the evaluation from the beginning, alongside productivity features.
6. The Ideal User May Not Be Tech-Savvy
I initially asked about founders and investors. Swapnika broadened the picture to anyone whose work depends heavily on WhatsApp.
She mentioned an operator working on partnerships at JioHotstar, where staying on top of conversations and responding on time was a substantial responsibility.
Then came the more surprising insight: she said their biggest segment was people who were not especially tech-forward.
PerisClaw had spread through word of mouth in Kolkata’s business community, reaching traders, retail business owners, and people in conventional industries such as steel. Swapnika said roughly 70% of users discovered the product through word of mouth.
These businesses may not organize themselves around the latest software, but they depend heavily on WhatsApp, Sheets, and Drive. The fit comes from solving a familiar operational problem, not from enthusiasm for AI.
7. A Retail Masterclass: From 57 WhatsApp Groups to Business Insights
The example that most strongly connected with my omnichannel work involved an operator running 57 retail stores across Northeast India.
Every morning, stores were expected to report attendance, stock, revenue, and specified compliance updates. Previously, the owner manually checked groups, chased missing submissions, and copied information into a Google Sheet.
Swapnika described a workflow he could delegate through a voice instruction: check the updates at 9 a.m., follow up with managers who have not responded, request missing values, update the sheet, and generate insights.
The Conversation-to-Insight Framework
I would break this workflow into five stages:
COLLECT: Find the daily updates across store groups.
CHASE: Follow up where submissions are missing.
CHECK: Identify incomplete information and request the missing values.
CONSOLIDATE: Add the completed information to Google Sheets.
INTERPRET: Surface stock movements, unexpected revenue differences, and relevant trends.
WhatsApp updates → Complete data → Google Sheets → Insights → Dashboards
The important detail is the middle of that sequence. Collecting messages alone does not complete the workflow. Missing updates and incomplete fields still need attention.
This is closely related to the operational questions in our discussion on retail store technology and AI strategy: how do we make everyday store information useful for decisions?
8. Dashboards Should Begin With a Business Question
Swapnika extended the retail example to custom dashboards. Rather than treating every reporting request as a separate technical project, an operator could ask for three metrics and how they are trending over a chosen period.
The agent, she explained, can create a dashboard around that request.
What stayed with me was the complete journey: a voice instruction initiates a process that gathers information, follows up, structures data, and presents it in a useful form.
I work with omnichannel brands, and store coordination can be a nightmare. There are regional groups, city groups, morning checklists, afternoon checklists, and constant updates.
The opportunity is to reduce the manual effort between a conversation and a decision. That is a far more meaningful promise than simply adding another chatbot.
I am Saurabh Agrawal and we come with a new episode on Dilse omni talks every fortnight and cover different aspect of omnichannel with amazing speakers.
