Sumit Saxena | AI-Led Customer Research Platform

Founder at & Inxise Datalabs | Building AI-Powered Market Research Tool | Customer Research | Ex AC Nielsen, Saxsons Group | Download FREE Guide👇🏻

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

How does Maya AI work?

Using Maya, you don’t need any research or technical skills. Maya is meant to augment your research process.

✅Create an instant discussion guide
✅Deploy instant smart, conversational interviews
✅Access reports without long wait times

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

@sidin They should have just sung, "hurry home, Hari". Would have been a great origin story for the QCOM category

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

India’s Deeptech ecosystem meets GCC opportunities:

Being a part of the Innovator Hub Showcase as one of the 40 start-ups was a privilege.

It has been a delight to be a part of the NASSCOM GCC Summit 2026.

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

Learn more about Maya AI here:
https://t.co/ZyPiwgdT5c

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

A pilot session costs you one interview slot.

Skipping the pilot costs you the integrity of your entire fieldwork.

You will always find something that needs adjustment: better pilot than sorry.

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

The best moments come after the first answer.

Build in space for follow-up.

"Tell me more about that."

"What do you mean when you say..."

"Walk me through what happened."

These are the questions that produce insight.

Don't be so focused on advancing through your discussion-guide that you miss them.

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

I understand the excitement of a novice researcher, but don’t try to cover everything with a discussion guide.

They come into a 60-minute IDI (in-depth interview) with 45 questions.
They're going to cover all the topics.
They're going to be thorough.
They're going to give the client their money's worth.

What they actually produce is a surface-level skim across 45 topics.

A discussion guide is a roadmap, not a questionnaire. Its job is to ensure coverage of the most important areas while leaving room for the conversation to go where it needs to go.
Those two things are in tension, and managing that tension is the craft.

Here is the TLDR version of the framework I understood over 2 decades:

🔹Sections, not questions.
Organise your guide around topic areas, not individual questions. Under each section, have two or three anchor questions and a set of probes.
You'll never get through all the probes. That's fine.

The probes are there so you have options, not obligations.

Test every question for embedded assumptions.
❌"Did you find the onboarding confusing?" presumes confusion.

✅"How would you describe the onboarding experience?" is better.

Avoid leading questions. Review every single question in your guide for any assumption in the phrasing. Then rewrite.

Pilot it.
A pilot session costs you one interview slot. Skipping the pilot costs you the integrity of your entire fieldwork. You will always find something that needs adjustment: better pilot than sorry.

Start with the business question, not the research brief.
Before you write a single question, write one sentence: "This research needs to be answered..."
If you can't complete that sentence cleanly, your guide will be unfocused irrespective of the individual questions are.

The best moments come after the first answer.
Build in space for follow-up. "Tell me more about that." "What do you mean when you say..."

"Walk me through what happened."

These are the questions that produce insight. Don't be so focused on advancing through your guide that you miss them.

A precise, well-crafted guide of 8–10 key areas will produce richer data than a comprehensive guide of 25 questions.

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

You may ask, who created Merren's Maya AI?

Maya AI is an AI-powered qualitative interviewer created by Monalisa Saxena and Sumit Saxena.

As a co-founder, we curated a research tool that thinks like great researchers and moves at the speed of business with technology.

Businesses are making decisions in days BUT traditional research takes weeks or months.

To close the gap between getting research results and helping businesses take quicker actions, Maya was built.

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

In an AI Moderator, there are three components that work together in real time:

1. A large (smart) language model trained on qualitative methodology that generates follow-up questions which sound natural, not scripted

2. A structured discussion guide that keeps the research on track and ensures comparability across respondents

3. An analysis layer that produces thematic summaries automatically once fieldwork closes

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

What exactly is an AI moderator? (and what it definitely is not)

The term "AI moderator" can be thrown around and it can create more confusion than clarity. Let me be precise.

When we were considering AI integration in our qualitative research methodology, our first instinct was scepticism.
Fast forward, here's what I can tell you.

What an AI moderator is not:
>It is not a chatbot serving a screener questionnaire.

>It is not a survey with a chat interface bolted on.

>It is not a tool that follows a rigid script and records responses.

What a genuine AI moderator actually is:
A software system that conducts qualitative research interviews autonomously.
It follows a structured discussion guide, asks follow-up questions based on what the respondent actually says, maintains conversational context across multiple turns.

It concludes the interview when coverage is complete.

▶️The key distinction from a simple survey tool is ADAPTABILITY.

▶️The key distinction from a support chatbot is RESEARCH INTENT.

A support bot routes issues. An AI research moderator is designed to elicit rich, honest qualitative data using techniques drawn from qualitative methodology.

Where it genuinely outperforms human moderation:
Consistency. A human moderator has good days and bad days, fatigue, personal biases.

An AI moderator probes identically for respondent number 1 and respondent number 200. For multi-market studies or large-scale consumer insight work, this consistency is an enormous asset.

Speed. Studies that take three weeks with traditional fieldwork can go from brief to analysis in three days.

Scale. You cannot run 100 simultaneous IDIs with human moderators. You can with AI.

Where human moderation still wins:
Complex, politically sensitive or emotionally nuanced research.
Situations where a respondent goes somewhere unexpected and you need human judgment to decide to follow.
Executive interviews, where rapport and improvisation matter acutely.

The most sophisticated research designs I'm seeing now combine both: AI for the breadth phase, human moderators for targeted follow-up with the most interesting or extreme respondents. This is a better design than either method alone.

The industry is changing. The question isn't whether to engage with AI moderation. It's how to integrate it intelligently.

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

@rameshsrivats Because IPs don't survive in India, despite the IP protection laws

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

When Maya is interviewing in multiple languages, it is native.

It is trained to understand cultural nuance, context and local expression which is critical in qual-at-scale deployments.

This is done to prevent a robotic voice from asking questions in qualitative research.

In an interesting turn of events, a respondent began candidly interacting (and diverting) with Maya despite being told that she is an AI agent.

Maya was using the local nomenclature during the interaction.

Due to the conversational nature of Maya AI, the respondent got carried away in the interaction, a testament to how natural these AI interviews can feel.

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

In the AI-moderated research era, “informed consent” can’t be a checkbox buried in T&Cs. It has to be clear, upfront and real.

Participants should know:
– they’re speaking to AI, not a human
– what data is being collected (transcripts, voice, etc.)
– how it’s stored, who can access it and when it’s deleted
– that they can opt out anytime, no consequences

The consent process should be simple, written in plain language, and presented before the interview begins.

Burying disclosure in a terms and conditions page that 95% of respondents will not read is not informed consent in any meaningful sense.

Merren's standard consent flow presents a plain-language disclosure message at the start of every interview.

It requires an affirmative response before proceeding and gives respondents a clear way to end the session at any point.

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

Can users customize Maya AI's behavior for different research contexts?

yes.

You can define the tone, depth and probing style. Maya can behave like a typical researcher, a friendly peer or a product tester.

It can be changed with regards to language, tonality and other factors that people usually decide when onboarding a human moderator for a particular target group.

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

If you can articulate a business question, you can use Maya. You do not need to be a market researcher.

Non-researchers can get clarity without any sort of methodological training. It is designed to be very simple so that professionals access industries can use it sans training.

https://t.co/JgvhNrei7u

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

What kind of data analysis does Maya AI provide?

The process is very simple.

You enter your research objective, you customize and upload your discussion guide.

Maya will conduct the interview and you can watch the insights build in real time.

Stakeholders can download the structured report which will have the basic themes, verbatims and the recommendations. It is very in-sync with how a typical qualitative research happens.

The user journey is made in a way that is accessible to anyone irrespective of technical know-how.

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

How does Maya’s conversational AI differ from traditional survey or chatbot tools?

Surveys ask but Maya listens and probes, beyond asking. That is the key difference.

Chat-bots follow scripts whereas Maya follows the meaning.

She will adapt her next question based on what respondents say just like a skilled human moderator. These are the vivid differences between Maya’s operation vis-a-vis a chatbot.

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

Maya is very efficient as an AI agent.

Maya can conduct thousands of simultaneous interviews across markets, languages, formats (voice or text) which is the very definition of qualitative research at scale.

There is a consistent quality, zero scheduling friction, there is no fatigue and bias.
It is very efficient as an AI agent and much more than what human-researchers can do.

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

There are qualitative studies that take 4 weeks to complete being done in under 4 days.

In one of the early cases, the interviews revealed a hidden product barrier sooner.

It allowed the client to pivot the messaging before the launch itself that saved them media spend.

It helped with the brand equity overall.

Speed is not just about efficiency but also about risk mitigation that was proven in this case.

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

Who can use Maya AI ?

It is designed for anyone who needs fast and reliable customer insights.

Research teams who are looking to scale, marketing teams that need very fast validation,

Product teams that are testing concepts,

CX leaders who are tracking sentiments in real time.

Maya’s AI-driven qualitative research capabilities are for people who cannot afford to delay business decisions who delay insights.

[Sumit Saxena | AI-Led Customer Research Platform @sum_saxena]

Researchers can also have big misconceptions around the use of AI. We had the same prior to Maya.

The initial fear was that it would make the process robotic.

Qualitative research is about the element of human insights in research that captures the soul and essence of people’s experience and perspective.

However, Monalisa noticed that when you design well, AI removes mechanical work and enhances depth.

AI does not replace researchers but makes them more productive.