The AI Consultant

Working on LLMs since 2018. 50+ satisfied clients. Talk to me if you’re looking to adopt AI for your business.

[The AI Consultant @AIWhizz]

@abhi_agarwal4 @openroundai Pretty cool.

[The AI Consultant @AIWhizz]

@abhi_agarwal4 One on the left is AI. Lower quality audio lol

[The AI Consultant @AIWhizz]

@abhi_agarwal4 Never thought there’s an RoI on business class tickets 😳

[Abhishek Agarwal @abhi_agarwal4]

I’ve been consuming YC content for 6 years now.

And I’ve launched like 10 products in the last 3 years. Most of them didn’t earn a dollar in revenue.

Why?

We were stuck in a loop of thinking of a cool idea, grinding for 8 weeks, launching a crappy MVP and then moving on when no one paid us. We thought the customer doesn’t want our product and we picked the wrong problem to solve.

While building Fabric I’m realising, it’s so important to get one customer to pay you. Just don’t stop till you get one customer.

It’s going to be a hard grind selling a crappy product with zero brand recognition. But it does get easier.

Most people give up too soon!

[The AI Consultant @AIWhizz]

Would you use the new Siri knowing all of that data is going to OpenAI?

[The AI Consultant @AIWhizz]

guys no i don’t want to send cold dms. but why are all of them using the exact same script?

[Abhishek Agarwal @abhi_agarwal4]

Here's how I used Twitter to validate our startup idea and get early customers!

When I first posted this 2 months ago, we were just testing the idea for Fabric.

We knew D2C founders had trouble with data and reporting but that was it. We had zero lines of code written.

This tweet got some attention and attracted a few design partners.

We worked closely with these design partners, shipped our MVP, and converted them into paying customers!

Working with these design partners, we learnt a lot about how broken e-commerce analytics for D2C brands really is.

To summarise, it's Excel hell.

Manual data aggregation. Reports full of errors. And no real data driven decisions!

We're building Fabric to solve analytics with an automation first and AI powered approach.

Super grateful that Twitter exists lol.

P.S. If you're a D2C founder struggling with reporting and analytics, my DMs are open!

[The AI Consultant @AIWhizz]

@EricWollberg You are your mom’s wrapper.

[Abhishek Vijayvergiya @abvijayvergiya]

I analyzed 402 pitches of Shark Tank to find patterns behind what makes a pitch successful. I found -

- Which sharks are most and least likely to invest in a pitch? (my guess was wrong for both)
- Several other interesting insights.

A thread. 🧵

[Abhishek Agarwal @abhi_agarwal4]

I see a lot of VCs virtue signalling on Twitter about how India needs to up its tech innovation game. And yet they are the first ones to back out of risky bets.

The simple reason why the Indian startup ecosystem lacks innovation is because no VC is ready to take a bet on ambitious founders.

Instead of first principles thinking, we think of copying or Indianizing what works in the west. We want to follow the trend, co-invest, and second guess our entrepreneurs instead of boldly backing competent founders.

No you can’t build a better LLM with $250k in seed capital.

It’s not that Indian founders don’t know how to innovate. You need capital to build groundbreaking tech. And there’s very few risk taker tech VCs in India.

Hence, Indian founders resort to business application oriented tech startups rather than foundational tech companies.

The smart ones who are serious about something meaningful don’t rely on the Indian investor ecosystem and move to the Bay Area instead.

It’s a wake up call for Indian VCs more than anything to fund ambitious tech companies.

Who will lead the way?

[Abhishek Agarwal @abhi_agarwal4]

Think about how we consume news.

Imagine if newspapers were your only source of news and you lost access to Google.

If you want to learn more about something, you have a reporter's number to call and ask questions to.

Wouldn't it be frustrating AF?

Yet most businesses seem to operate this way when it comes to Business Intelligence and Analytics.

Dashboards are digital newspapers - great for skimming through the top headlines but unable to give you a deep dive on interesting topics.

Data analysts are like reporters, meticulously curating stories for you to consume from data. You wouldn't go to a reporter every time you want a deep dive on your news, will you?

Just like the solution to deep diving on news is not to read more newspapers, the solution to deeper BI is not to build more dashboards.

Because dashboards only give you a 1000ft view of your business and fail to answer the "why" question, you have an adhoc factory of analysts to cater to adhoc requests.

Need a quick data pull? Email an analyst.

Need to see the trend of a metric before your meeting in an hour? Call an analyst and pull a personal favor.

Need to understand why churn is high in your loyal customer segment this week? Tell an analyst the CEO is asking for it and it's urgent.

In the meantime, the analytics team fails to work on interesting, needle-moving projects merely becoming a service arm within your org.

One fine day, the CFO questions what's the RoI on the analytics team. And when things go south, half the analytics team is let go.

This is how analytics becomes a cost center instead of a business growth vehicle.

We're working to change that with Fabric. By using LLMs to enable self-service analytics for business users. By bringing a Google like interface to BI.

If you resonate with this sorry state of affairs and consider yourself an innovator, we're onboarding 5 design partners. Do feel free to reach out via DM. 🤟🏽

[The AI Consultant @AIWhizz]

@abhi_agarwal4 @BowTiedSalesGuy You should check this out @BowTiedSalesGuy

[Abhishek Agarwal @abhi_agarwal4]

4 years ago, I had never sold anything for a penny in my life. Then I learnt sales and have closed $2M+ worth of deals since then for my own business.

I attribute a lot of my sales success to following @BowTiedSalesGuy's methods and his content has shaped my views on sales a lot. He's definitely unconventional and his advice is hard to follow but it worked wonders for me.

As a token of appreciation to BTSG and spreading his reach, I built an AI Chad Sales Trainer GPT on Fabric.

This bot has access to most of BTSG's advice that he has published on X and Substack.

Use the AI chad sales trainer to:

✅ Be more effective at sales
✅ Answer real, situational questions
✅ Educate your sales team

I've been using the bot and can't wait to share with y'all. It's smart, witty, and has effective (but unconventional) sales advice.

I'm not publishing it online yet but will do so once this gets retweeted by @BowTiedSalesGuy himself AND reaches 500 likes.

Happy to open it up for everyone to use after that.

[The AI Consultant @AIWhizz]

@abhi_agarwal4 @Ashneer_Grover 🤣

[Abhishek Agarwal @abhi_agarwal4]

Wish he had listened to his own advice. 😔

[Abhishek Agarwal @abhi_agarwal4]

Folks looking to integrate AI into your product - I have news for you. 📰

RAG (Retrieval Augmented Generation) is not the silver bullet you think it is. Most people think that building a smart AI assistant is just about integrating an LLM with a Vector DB.

Sure, if you only have a couple of PDFs - this approach might suffice. But if your data is more complex, an LLM + VectorDB won't be adequate to answer questions users will ask.❌

Last week, I chatted with 5 business owners who had the same problem - RAG was proving to be inadequate to answer simple questions on their custom data.

How so?

Let me demonstrate this with an example.

Let's say that I want to create an AI agent to analyze the Generative AI startup landscape. I'll give it access to a dataset on ~400 GenAI startups which has names, funding amounts, description, websites etc.

Now, if I ask it a question like "which GenAI startup has received the highest amount of funding?" - RAG will fail.

Try asking this to any of your favorite "Chat with Your Data" app and tell me how it goes. 😉

Reason is that vector DBs aren't made for answering such questions. This seems like a natural question to ask though, doesn't it?

This is why you need an AI agent that goes beyond RAG to answer questions on your data.

I actually built this in 10 mins on Fabric. As you can see in the video, not only does the Fabric agent correctly answer the question on the startup with the highest funding but also answers usual RAG type questions, e.g. name some startups in the customer service segment. ✅

Create your AI agent today by signing up on Fabric. DMs open for inquiries. 🚀🚀

[Abhishek Agarwal @abhi_agarwal4]

Nobody changed the world working 40 hours per week.

[Abhishek Agarwal @abhi_agarwal4]

@vkhosla Okay this blew up.

Customary shameless plug:

We're raising for @fabrichq_ai - a platform to build AI agents that automate complex business workflows.

Feel free to DM if you want to invest in our seed round or become a customer.

More details here - https://t.co/DHlycv1Vv6

[Abhishek Agarwal @abhi_agarwal4]

Such a shame. They raised ~18M in capital only to desert their customers.

See the desperate comments from customers on a random 2 yr old LinkedIn post from the CEO.

Yet the CEO’s entitlement knows no bounds.

They could’ve handled this so much better.

[The AI Consultant @AIWhizz]

All that is cool but why is she practising a handjob?

[KRK @kamaalrkhan]

Full story of Israel and Palestine in few seconds. Brilliant.