2025-01-18 08:00:00
Last week I was sitting with a friend of mine who is working on an AI co-pilot for X startup.
(I know, I know. A lot of my posts start with “I was sitting with a friend”. This is a gentle way of signalling that I have friends. These are real people, people. And real discussions).
We were discussing agents. I have never let my basic knowledge of things stop me from forming opinions. So why stop now. So this is what I shared with him.
This may not be new to people who are deep into AI and agents. But for an average product person it might be interesting.
There are 2 types of agents. Agent as colleague. Agent as copilot.
Colleague agents are much more expensive. Take Devin for example. He is supposed to be a junior developer who you can also delegate tasks to and he works independently, while Cursor Agent (Composer) and Replit Agent are examples of copilots. They take some tasks off you, but I assume they enhance your ability as a developer rather than working independently.And they are priced accordingly. 500 dollars a month for Devin versus 20 dollars for Composer.
So how do you work with each type of agent? I told my friend that if you think of Devin as a real colleague, an employee in your organisation, then its onboarding should be like a real colleague. Let’s take the example of a new agent. We have decided to buy Jagan, a data analyst agent created by Data Friend Labs. What should the workflow look like?
In an ideal scenario. This may already be happening with some agents today, but let’s say we don’t know of any and we’re using some good old-fashioned first principles thinking. How do you onboard an agent?
A regular employee joins. They have a manager. They have an onboarding buddy. They get an email id. They have a slack. They have an onboarding checklist created by their manager. Now suppose you want this colleague, Jagan, to be as useful as a human employee. How would you do its onboarding?
A normal human would probably do 1-1s with key people. Read a lot of documentation. If they have questions, they could ping their onboarding buddy or manager. Do some onboarding projects. At Facebook, you might even do your first PR on your first day.
No human is going to remember every word of every document they read during their onboarding. They create an onboarding folder. Bookmark all the Google Doc links. Take notes from the 1-1s with key stakeholders. Is the process the same for Jagan? What is Jagan anyway? His brain is powered by some LLM. His memory (local) is the docs he has read. There will be some RAG based system to extract relevant information depending on the task. He will have access to tools. Use of tools. This means his account would be created on Tableau, Metabase, Amplitude, Appsflyer, Google suite, (all relevant tools he needs to do his job). Do we really need stakeholder interviews? Will employees be interested in talking to an AI colleague to provide context? Will we just create the relevant documents to give Jagan context on each stakeholder?
A normal employee would work pretty independently, only going to their manager when they get stuck. I expect Jagan would be stuck a lot. LLMs are good, and with each new generation they get more and more capable. But the context window is still limited. Yes, you can RAG and get relevant information from the docs, but we have seen cursor agents mess up so much. Especially during onboarding, do they keep going back to their manager when they get stuck? Remember that most LLMs today work on pull rather than push. ChatGPT never pings you or sends information itself. You ask ChatGPT. You tag some files and ask the cursor to do something. Even the Agents platform mostly takes a request from a user and completes the task. Do we have a progress md file for Jagan, where he writes after each interaction? People recommend using a progress md for Cursor, where you update the progress, so that when you ask Cursor, it is always up to date on what is happening with the project. Jagan can have many interactions during the day. A human colleague does that. So does it write everything? How does it know what information to write? Find out what is important to discuss on Slack. Does Jagan attend the standups? Does it even have a performance review? Devin was an interesting experiment in how AI colleagues will coexist with human workers. But the reviews aren’t great. And the price means that at least people like me can’t just try it out for fun. Also, Devin has been more like Cursor agent than an actual colleague. People tag Devin to help with PR. And menial tasks as far as I can see on X dot com. You don’t tag your human colleague like that. They work as equals. A manager decides which tasks are assigned to them based on their maturity, skills and experience.
Copilot is also interesting. Copilot is broad. Used by every employee. So shared context. Example: Copilot for a data analyst, let’s call him Data X, needs to understand the schema. Understand the data pipeline. Which data should be retrieved from which table. But setting up the copilot takes time. Even for Cursor, a lot of people just read something on Twitter and figure out how to make Cursor better. Cursorrules, progress md, product description md, asking Cursor to go through the steps before executing, all came from random Twitter users. Democratised learning. But coding is mostly dependent on the language you write code in. So .cursor has rules for different languages. Collecting data and creating artefacts is much more nuanced. For example. GMV may be GTV in some companies. Sales can be calculated differently. I remember reading a report on the state of food delivery from Momentumworks and they have so many disclaimers saying note that Grab’s revenue is calculated this way, Shoppee’s this way and what not. GAAP is only for public companies. And real companies have many more definitions and terminologies. So how do you set up Data X?
Here you are getting onboarded to Data X (employee onboarding done on a new platform) vs new employee (Jagan) getting onboarded to your company.
Do you start dumping documents to it? Do you need the company (let’s say Data Inc) to come in and create a middle layer to map how you would give tasks to the copilot in natural language vs how it would query? Or can you platformise it somehow? For example, let it ask questions in the first week and get context about your business? You can start throwing documents at it. You need Data Inc to come in and set up the RAG system. You need FDEs, as in the case of Palantir. Palantir is high ticket and the cost structure allows it, but maybe not for a bottom-up Data copilot. How much can you abstract away? Context windows will get bigger. Maybe you will use fine-tuned open source models instead of the frontier models, but every company is unique and you will still need a lot of manual work to set up Data X. I spent so much time on onboarding because onboarding is important. The quicker the onboarding, the smoother the process, the quicker the aha moment, the more likely people are to be wowed by these AI products and actually use them to augment or replace their employees. I think if I were a PM trying to get into AI, I would be obsessed with how to master onboarding. Onboarding PMs, or PMs for activation, was an important role in web2. It’s going to be even more important for people building AI products.
2025-01-17 08:00:00
To get better at product, you don’t need to take more product management courses or read more product management frameworks from Reforge. You need to use more products, think more about various product decisions, and actually ship stuff.
I’ve seen PMs I’ve worked with post on Linkedin about the latest course they took, but they barely cared about their jobs, nor would they spend any time getting better at the craft of building products.
I’ve met founders who bitch about the PMs in their companies, people who’ve never had a new idea in their lives, can’t write a strategy document to save their lives, don’t care about design, don’t care about data, don’t spend time tinkering on side projects. This is why many developers and designers hate PMs.
You don’t need another course. You just need to care more.
2025-01-16 08:00:00
Since I wrote this post on aha moments, I have been thinking about ‘aha’ moments for AI agents.
The best ‘aha’ moment post I have ever read is this one by Fred Wilson on Twilio’s seed pitch.
One principle when it comes to communicating products is ‘show, don’t tell’.
What I see: Every agent website and video shows complicated workflows with flowcharts.
The only good demo/launch video I have seen is Alex Cohen’s Hello Patient.
A good demo of an imaginary product: You are a founder building an insurance voice agent product and you want to sell it to HDFC. Go to their call centre. You probably have a local LLM running on a truffle (simpforsatoshi) like device. You open your briefcase. Take it out. No data leaves. You start making calls. Sell more insurance than the best agent in the call centre. Far higher productivity. Lower costs.
A new form factor is interesting. Of course you can use cloud providers.
I know that is not possible today. You would need to get data from actual prospects before making that call. You would have to follow HDFC’s guidelines (build a RAG solution) and then also fine tune your model to better suit HDFC. And the fine-tuning itself would take time. And they would not share it without signing a contract. So the ‘aha’ moment is not immediate. Nobody’s mind is going to be blown immediately. And that the a problem.
2025-01-11 08:00:00
“O Manas, why did Swiggy launch Snacc? Isn’t it an admission that having multiple brands might be a better plan?”
I have had people DM me this after Swiggy launched their new Snacc app.
Snacc is a separate app because you can’t directly compete with the restaurants you serve (one side of your three-sided marketplace) in the same Swiggy app.
If you’re positioning SNACC as a solution to expensive food delivery, long wait times, and excessive fees, you can’t effectively do this by placing the SNACC icon next to traditional food delivery options on Swiggy’s homepage.
Customers need to think of SNACC as a distinct service for quick, reheated frozen food delivered in 10 minutes. Mixing dark store food with restaurant offerings in the same app could confuse the value proposition.
Consider this: If a customer has a bad experience with a reheated samosa from SNACC, they might question Swiggy’s entire food delivery business. Additionally, restaurants would likely object to having their offerings displayed alongside reheated frozen food, given the perception issues. (Though one could argue that restaurant food is also frequently reheated packaged food.)
The app separation also helps manage customer expectations. You can’t effectively promise both 10-minute and 50-minute deliveries in the same app. The goal is to train users to use SNACC specifically for quick snacks, while using Swiggy for restaurant food – even if it means paying more and waiting longer.
The separation enables different demand and supply strategies. On the demand side, having both quick, cheap SNACC items and restaurant food in one app creates competing user flows. For example, a customer looking for a samosa on SNACC might compare it with a nearby restaurant’s samosa on Swiggy, which could be cheaper despite the 25-minute delivery time. Separating the apps eliminates this direct comparison. Supply-side operations also differ. Dark store delivery workers handling 10-minute deliveries may require different incentive structures, necessitating a separate supply spending strategy.
Finally, organizational dynamics play a role. A separate app gives the SNACC team autonomy. They don’t need to compete for visibility on Swiggy’s home screen or negotiate with the design team about animations and features. They can move quickly, ship features rapidly, and maintain independent control over branding, communications, and overall product development. This illustrates a key advantage of multiple apps. However, it comes with challenges: building a user base from scratch, training customers to use multiple apps, and managing cross-platform features. For instance, I still can’t use my Zomato wallet money from an Infinia gift card on Blinkit, nor can I earn and use points across their product portfolio.
Both approaches have their merits and drawbacks.
Having said all this, no one really knows the exact rationale behind this decision. Based on my decade-plus experience in tech, most decisions ultimately stem from CEO preferences and influential stakeholder input, shaped by various biases – recency bias, confirmation bias, and many others – plus competitive pressures. The true decision-making process often remains opaque.
2025-01-10 08:00:00
It was ~15 years ago. I was in Kota, staying in a paying guest accommodation, cramming organic chemistry before a placement test at Bansal’s that would determine my batch assignment for the next few months.
I still remember the day clearly. I was hanging out with a couple of guys from the PG accommodation. We were talking about movies. I had never been a big movie enthusiast - not the kind of person who would go for first-day-first-show screenings. But like all teenagers, I had opinions. I remember one of them bringing up an Akshay Kumar movie. I mentioned casually that I thought the movie was bad and made a comment about Akshay Kumar.
Immediately, I saw the mood shift. One of them got upset, and things escalated from there. Coming from the east, where people don’t typically engage in hero worship, it was my first time seeing someone get so upset over something so external. I couldn’t understand why he got so agitated over an offhand remark about Akshay Kumar. He was apparently a big fan, and it wasn’t acceptable to say anything negative about his idol. He was defending his hero’s honor. Until the end, it seemed like satire. Surely, I thought, he must be joking. Why would someone get so upset? Our relationship never recovered. Over time, this incident, along with other issues, led to us avoiding each other. There was always tension between us.
I’ve been supporting Liverpool since 2002 and was a big Sachin Tendulkar fan. But I would never take offense if you said anything about Sachin or Mohamed Salah. After all, I don’t personally know Salah. He doesn’t even know I exist. This is why I find the culture of hero worship very strange, whether it’s actors, politicians, or political parties. I’ve never understood it.
At least Akshay Kumar doesn’t get paid from taxpayer money - you vote for him on Friday when his movie releases, with your wallet.
2025-01-01 08:00:00
It’s that time of the year again. End of year reflections.
I was thinking about this podcast I heard a few years back. This valley founder who had it all - the headlines, the success, the network. Then his company started going downhill. And suddenly, all those people who used to hang around? Gone. He talked about how he had never felt more alone. Eventually packed up and left the SF startup scene altogether. Spent years trying to find himself while rebuilding his company from scratch.
Makes me wonder sometimes. What if I moved from Bengaluru to Hubli tomorrow? Left this whole startup bubble behind. No more fancy meetups, no more “what are you working on” conversations, no more startup ecosystem drama. If work wasn’t my identity anymore, how many people from my current circle would actually stay in touch?
Not asking for LinkedIn engagement or Twitter DMs. Just genuine “hey, how are you doing” calls. People who’d want to talk because they enjoy our conversations, not because there’s something to gain or lose.
In Bengaluru your worth is tied to what you’re building, what company you work at, what your designation is. Take all that away and who really cares about grabbing that coffee just to catch up?
Maybe I’m being cynical. Maybe I’m overthinking. But these thoughts hit different during December when everyone is posting their year-end achievements and goals for next year.
Not that I’m planning to move to Hubli tomorrow. But it’s a good thought experiment. Makes you think about what really matters, who really matters.
Anyway, I’ll also be back to posting about product and growth next week. Let me have this moment of existential crisis.