One thing we’ve tried to do in every edition of this newsletter is build AI tools for specific problems. Automate this, extract that, summarise the other thing — basically all the engineering stuff.
We haven’t really gotten into how AI helps with the non-tech side of stuff, specifically growth. That’s finding users, keeping them, and figuring out what to charge.
This edition is all about that. How to apply AI to the growth side of building — and the companies already doing it. Let’s get straight into it.
Make global calls with your AI agent in 10 mins.
Your first international client could be at the other end of an AI call. Stop losing weeks getting the phone line with carrier compliance, global number provisioning, and sub-80ms latency.
Plug into Vapi, Retell, Sarvam or ElevenLabs, get a dedicated number, and make your first call in under 10 minutes in 130+ countries with Vobiz. Razorpay, Shiprocket, and MobiKwik already do. Newsletter readers get ₹2,500 in free credits with code GROWTHXNL-63WEHB.
First, what does growth entail?
Activation, engagement, retention, monetisation, and referrals all fall under growth. That’s the entire process of getting a person to try your product and stick with it. It’s honestly a science.
You’re continuously trying to calibrate your strategy to a few questions across every stage of the product lifecycle.
Who is your user?
What do they actually care about?
Why did they stop using the product?
What would make them pay?
So, where does AI fit into it? Let’s take a look.
a. Acquisition
Think about the last app you signed up for. Something made you stop and act. A line of copy that named your exact problem. An ad that felt oddly specific. Someone spent weeks engineering that moment.
In practice, it comes down to this.
Customer Acquisition Efficiency = (Targeting × Message Effectiveness × Experiment Velocity × Spend Efficiency) / Time
Here, message effectiveness depends on how many people you talk to and how well you recognise and integrate your customer’s voice into your copy.
Now think of how AI can help here for a second.
We know AI can scan the web for information, draw inferences from past experiments, and generate context-aware copy.
Targeting: AI can scan thousands of Reddit threads, app reviews, community posts and search results to understand who has the problem you solve — and how they describe it.
Instead of interviewing 20 people and hoping you’ve heard enough, you can analyse thousands of conversations to spot recurring problems, phrases and buying triggers.
Your differentiator: Telling AI where to look and deciding which patterns actually represent your buyer. AI can surface 50 possible insights. You still need to know which five matter.
Message Effectiveness: Instead of copywriters searching for the right words for ad creatives, they can use AI to generate copy variations with different hooks.
Your differentiator: Set the brief and review outputs with your experience. Catch when AI's highest-performing version wins clicks but overpromises a high-CTR ad that misrepresents the product destroys retention downstream.
Spend Efficiency: Many platforms already use AI to measure, analyse, and forecast your ad spend based on your audience, creative and ad copy. In fact, as of 2026, 62% of all ad spend is driven by Meta’s AI spend-forecasting algorithms.
Your differentiator: Choosing which platforms to be on AI optimises within a platform, not between them. And defining what "conversion" means.
Experiment Velocity: Now you can run acquisition experiments more quickly. You can dynamically change something as small as copy in the actual UI with AI and see what works for your users.
Your differentiator: Setting the strategy. Do you even need ads, or is making the product more comprehensive a better acquisition channel? AI executes within a strategy. Picking the strategy is yours.
b. Activation
Or how you get someone to realise your product is useful and generate that AHA moment.
In practice, the equation looks like this.
Activation Rate = (First Action Relevance × Segmentation Quality × First Output Quality) / Steps to Value
So:
First action relevance is which action you ask the user to take first — and whether it’s the right one.
Segmentation quality is how well you’ve read who this user is and tailored the experience accordingly.
First output quality is how good the product is when the user completes that first action.
Steps to value are every form field, screen, and decision point between signup and that moment.
Now look at where AI is currently.
The biggest shift AI creates in activation is the collapse of the skill gap. Before AI, the quality of a user's first experience depended on their skill level. A non-designer opening Photoshop got a blank canvas and a toolbar they didn't understand. That has changed.
First Output Quality: Take Canva. It asks what you’re making and drops you into a relevant template. But the real shift is the AI prompt path — a user types what they want and receives a finished, publishable design in under a minute.
Your differentiator: Defining what a good first output looks like.
First Action Relevance: Now look at Grammarly. Their aha moment doesn’t happen inside their app. It happens when the browser extension catches a mistake in a real email in the user’s own environment during the first session.
Your differentiator: Choosing where in the user’s existing workflow to insert the first value moment. Grammarly chose “inside the email you were already writing” over “inside our app.”
Segmentation Quality: Notion uses signup answers to pre-populate a workspace with relevant templates based on behavioural segmentation. They do this by clustering users based on what they actually do in the first session, rather than what they select in a dropdown, to route them to the path most likely to activate them.
Your differentiator: Writing intake signals that give AI accurate information to work with. If the signup question is vague, every downstream branch is built on a wrong assumption.
c. Engagement
Think about why you go back to an app you love. Say Instagram.
Your content’s fresh, your feed is personalised, and you’ve built a habit like watching reels in the bathroom around it. And it’s intuitive, meaning it takes up 0 cognitive load.
So, if engagement had to have an equation. It’d look like this.
Engagement Rate = (Content Freshness × Content Relevance) / Cognitive Load
So, what can AI do here?
Be the personalisation algorithm? Recognise engagement patterns for specific users and predict when a similar user’s engagement will drop? Maybe personalise the UI trigger for the person as well. AI can do all of this today.
Content Relevance: AI can continuously adjust what a user sees based on their interests.
Take Spotify. Two people can open the exact same app and see completely different recommendations based on what they listen to, skip and return to. Or take Meesho. Spend time browsing a certain kind of product, and the homepage can start surfacing more of what you’ve shown interest in.
Your differentiator: Deciding what should actually become more personalised. If you’re Zomato, is relevance showing a restaurant similar to your favourites — or deliberately showing you something new?
Content Freshness: AI can also keep finding new reasons for someone to return.
For Spotify, that could mean new music. For Zomato, restaurants you haven’t tried. For Groww, a new insight about what happened to your portfolio this month.
The point isn’t just to show users more. It’s to make the product feel useful again every time they return.
Your differentiator: Knowing what “fresh” means for your product. More recommendations aren’t automatically more value.
Cognitive Load: AI can analyse where users struggle, abandon a flow or spend unusually long completing a task and help you identify where the product is making them work too hard.
If users consistently spend five minutes comparing nearly identical restaurant options before checking out, that’s a signal. If they keep bouncing between three screens to complete one task, that’s another.
Your differentiator: Knowing which effort is friction and which effort is useful. If you’re Swiggy, should AI narrow 300 restaurants to 20? Five? Or would that remove too much choice?
d. Retention
Think about the last subscription you cancelled. Here’s how that might have gone down.
You opened the app less often, skipped a few sessions, and stopped caring about the new features. By the time you hit cancel, the decision was already made.
So, the equation to nail retention probably looks like this.
Retention Rate = (Value Consistency × Early Intervention Speed × Re-engagement Relevance) / Churn Signal Lag
Now think about what AI specifically does here.
It detects behavioural shifts. It predicts which users are about to churn. It personalises the intervention to the specific signal that triggered the risk. And it automates the entire workflow, including detection, intervention, and escalation, without a calendar reminder directed to Customer Success.
Churn Detection: AI can spot when the way someone uses your product starts changing.
Maybe they used to open the app five times a week, and now they do it once. Maybe they still log in, but have stopped completing the action that used to give them value.
AI can compare these behavioural shifts against patterns from other users and flag someone before they actually disappear.
Your differentiator: Defining what a churn signal means in your product. You need to know which behavioural changes actually indicate that value is disappearing.
Intervention Relevance: Once AI spots that change, it can personalise what happens next.
Say someone stops opening Netflix after finishing a show they used to watch every night. Showing them another similar series makes more sense than sending a generic “We miss you” email.
The same logic could apply to fintech, SaaS or commerce: identify why someone is drifting away, then match the intervention to that signal.
Your differentiator: Deciding which signal deserves which intervention. AI can personalise the response while you still decide what response makes sense.
Intervention Speed: AI can also act much earlier. Traditionally, you might wait until someone hasn’t logged in for 30 days or until they hit “cancel.” But by then, the problem may have started weeks ago.
If AI detects the behavioural shift earlier, you get a chance to intervene while the user still cares.
Your differentiator: Deciding when to act. Too late and the user is gone. Too early and you’re interrupting someone who wasn’t actually at risk.
To conclude
The pattern across all four stages is pretty consistent.
AI can find the signal, generate the output, personalise the experience, and automate the intervention. But someone still has to decide what signal matters, what “good” looks like, and what the system should do next.
BTW, you can get your hands dirty doing exactly this at our next Build Sprint.
Over two weeks, you’ll build an AI product from scratch, put it in front of real customers, figure out what they actually want, and try selling it.
Basically, everything we just talked about, but with your own product.
Cohort starts October 2.
