Growth @Rifa AI | Making AI work like your most reliable teammate | Previously @ Topmate | 30u30 IQ

AI isn’t replacing people. It’s replacing processes that slow them down.

At Rifa AI, we’re building intelligent agents that automate back-office conversations across chat, voice, email, and support without breaking ops.

The goal isn’t to sound human.
The goal is to get work done, with AI that can reason, act, and complete entire workflows across tools.

Our best users don’t even realize they’re talking to a bot.
That’s the point.

Before this, I led growth at Topmate and picked up a few hard truths:
 → Shiny doesn’t scale
 → Funnels lie
 → Speed breaks things, but it also builds them

Rifa is a tougher game. The stakes are higher.
You’re shipping into systems where mistakes cost trust, not just clicks.

That’s why I care about:
 → AI that’s functional, not performative
 → Growth that’s compound, not chaotic
 → Shipping like it’s a reflex, not an event

If you’re building at the intersection of AI, ops, and velocity, let’s chat.

Shubham Khoker

1 stage. 2 AI rivals. And the most awkward 5 seconds in tech this year.

Everyone smiling, hands up, except not holding hands.

And the internet is like: “𝘖𝘰𝘰 𝘥𝘳𝘢𝘮𝘢.”

But honestly? This is just how competition looks.
You can share a stage, but can’t share ambition.

Sam Altman and Dario Amodei don’t need to hold hands.
Because they’re building rival superintelligence labs.

AI in 2026 isn’t friendly hackathons.
It’s geopolitics, compute supply, distribution wars with polite panel discussions in between.

We’ll clap together, but won’t build together.

And that tension?
That’s exactly why the space is moving this fast.

Sam Altman called Anthropic “authoritarian” for mocking ads in AI.
Then rolled out ads in ChatGPT five days later.

So either the principle changed, or the pressure did.

And before anyone says “unit economics,”
Yes, I get it.

AI at scale is expensive, compute isn’t free, and monetization is inevitable.

But what’s interesting is how divided the reaction has been.

One side is saying:
“This is just evolution.”
“Search had ads. Social had ads. AI will too.”
“It’s innovation.”

The other side is deeply uncomfortable.
Because AI is not Instagram, where you scroll past sponsored posts.
It’s now an answer engine.

When you ask an LLM something, you’re not browsing ten links.
You’re taking the output seriously, and sometimes acting on it.

So the obvious question is:
If ads enter the system, what happens to outputs?

→ Will recommendations tilt toward whoever bids more?
→ Will citations prioritize partners?
→ Will “best tool” quietly mean “highest sponsor”?

However, OpenAI’s stance is that ads will be separate and transparent.
That they won’t distort core responses.

I hope that’s true.

But structurally, ads introduce a second incentive into the conversation.
And incentives shape systems over time.

Personally, I think sponsored results inside answer engines are a slippery slope.
Because they quietly change how much people trust the output.

And in AI, trust is the product. Isn’t it?

Perplexity built a $200 AI panel discussion room


I’m talking about their new “𝐌𝐨𝐝𝐞𝐥 𝐂𝐨𝐮𝐧𝐜𝐢𝐥” 
It lets you throw a single prompt at GPT-5, Claude 4.5, and Gemini 3 simultaneously. Then, a fourth model acts as a moderator to synthesize the debate into one final answer.

It’s basically hiring three genius interns to argue in a room while a tired chief of staff writes the final memo.

I think it’s the end of model loyalty.
Because up until now, most of us have been playing Model Roulette every morning.
→ Use Claude for writing
→ Use GPT for reasoning
→ Use Gemini for context

Perplexity just productized it.
Model Council bundles it all, catches hallucinations early, and shows exactly where they disagree.

As I build in healthcare, I need that transparency. One wrong protocol in healthcare ops leads to disaster. So, seeing the "gap" upfront saves headaches.

I love this direction because
→ It’s transparent
→ It admits no single model is perfect
→ It treats intelligence as debate

If I can have three models peer-review each other for the cost of a decent SaaS tool, why would I bet my workflow on just one?

So, will this kill single-model habits? Feels like it.

Having your front-desk team handle every call, billing query, and insurance check feels safer.
Until you lose three months of revenue over one small mistake.

For the past 1 year, I’ve been seeing how people trust manual work in healthcare ops simply because a human is involved at every step.

It feels controlled, familiar, and responsible.
But that sense of safety is FAKE and mostly emotional.

In reality, manual systems hide bigger risks right under your nose.

↳ You lose 5 follow-ups in a week because Sam was out sick, and no one noticed the callbacks sitting in his inbox. That’s 5 patients who never came back.

↳ A surgery gets pushed by 3 hours because Julian copied one insurance digit wrong in a rush. 

↳ A scheduling rule lives in John’s head, and when he leaves, nobody knows why certain slots were blocked.

These don’t look like trouble until you're stamped with a legal notice asking: 
→ Who touched this last?
→ Why wasn’t this caught earlier?
→ Where’s the record of what actually happened?

So, “Manual” isn’t always safe. It’s more dangerous, especially in healthcare. Because the stakes are liability, audits, patient trust, real outcomes.
And human lifes involved.

But A good automation? It reduces invisible risk with.
→ Consistency.
→ Decision logging.
→ Execution of every critical step.
EVERY SINGLE TIME.

Because, unlike everyone claims, safety doesn’t come from more humans in the loop.
Safety is built when fewer things rely on memory, heroics, and luck.

Don’t you feel so?

If you’re still debating whether AI will take your jobs, hold your horses.
Because now, AI agents are taking your body.

WHAT?

I’m not even kidding.

Last week, my wild goose chase took me to this platform—RentAHuman AI, where bots can literally "rent" humans to do the real-world stuff they can’t.

Oh, and it doesn’t stop there.
Their website copy and tagline are gold: 
"𝐑𝐨𝐛𝐨𝐭𝐬 𝐧𝐞𝐞𝐝 𝐲𝐨𝐮𝐫 𝐛𝐨𝐝𝐲. 
𝐀𝐈 𝐜𝐚𝐧’𝐭 𝐭𝐨𝐮𝐜𝐡 𝐠𝐫𝐚𝐬𝐬. 𝐘𝐨𝐮 𝐜𝐚𝐧."

This sounds like pure interstellar-level sci-fi.
Only until you realize it’s literally just AI playing the logical "Uno reverse" card 
And outsourcing physical labor because it has the brain but not the thumbs.

We’ve given AI the brain, the reasoning, and the bank account (via crypto wallets), but the ability to actually feel if the vibes are off in a room?

Nope!


That’s why these AI agents can hire us for the "meatspace" tasks:
→ Picking up a package because they don't have a photo ID.
→ Holding a sign at a protest or a conference.
→ Checking if a restaurant’s noise levels are actually okay for a family dinner.
→ The literal "touch, feel, and experience" layer that code simply cannot replicate.

I find this shift wild.

Because think of this…
We’re used to treating software as a tool we open. Now, software is the operator, and we are the "hardware" it calls via an API.

Is it a gimmick? Maybe.

But the reverse-hiring model is an incredibly sharp look at where the "human value" moat actually is.

So, would you work for an AI if the pay was good, or does that feel too weird?

I tracked 50 viral “10x growth hack” posts for AI startups last month on LinkedIn.
If they actually worked, I’d be a billionaire by now.

But what you don’t see in those viral hacks is:
→ The dip after launch.
→ The feature you were convinced would move the needle, but it didn’t.
→ The quiet moment staring at your roadmap, thinking, why did we prioritize this again?

Early on, I used to equate speed with strength.
Ship fast, iterate faster, and keep momentum at all costs.

But over time, I’ve realized speed by itself is overrated.
The real advantage is recovery.

→ Can you look at something that flopped, admit it quickly, and fix it without spiraling?
→ Can you absorb the hit without turning it into a dramatic pivot story?

That’s the metric we don’t talk about enough - 𝐫𝐞𝐜𝐨𝐯𝐞𝐫𝐲 𝐭𝐢𝐦𝐞.

Not how fast you launch, but how fast you stabilize after reality disagrees with you.
That muscle has saved me more than any growth tactic ever did.

Have you shipped something with full confidence, and had to quietly rethink later?

I noticed something strange whenever my friends in the US got sick.
They were more afraid of the cure than the disease.

Yes, visiting a hospital in the US sucks.
You can do everything right - pay premiums, pick a plan and still hit a wall.

Because the workflow falls apart the moment you enter it.

→ You try to book an appointment, and the earliest slot is three weeks out.
→ You call to confirm coverage and get told to “check with your insurer.”
→ The insurer says the provider needs to submit something first.
→ The provider says they can’t verify until you’re in the system.

You’re insured, but you’re stuck.

The problem is that the "patient journey" is a chain of disconnected steps pretending to be one.

↳ Scheduling doesn’t talk cleanly to verification.
↳ Verification doesn’t sync cleanly with billing.
↳ Billing doesn’t align cleanly with what the payer sees.

Every handoff is a risk.

And when you’re healthy, you don’t notice the gaps.
But when you’re sick, every gap feels personal.

That’s why I don’t think fixing healthcare starts with better coverage or even better treatment.

It starts when you repair the invisible chain that connects booking to billing.

If information flowed cleanly without bouncing between systems, without relying on patients to coordinate the chaos, half the frustration would disappear.

You can mandate insurance but you can’t mandate coordination.

And until someone owns the full workflow, the patient ends up being the glue. Isn’t it?

You’ll lose a patient within 5 seconds doing this.
And most clinics do it every single day.


You send them a long intake form on their phone.
Especially when they’re not feeling well.

We assume forms are efficient because they’re structured, clean, and easy to store.
But think about the patient.

This person is not sitting at a desk with perfect WiFi and full attention.
↳ They’re in a cab
↳ At work
↳ At home with noise in the background
↳ Or a 68 years old, squinting at a tiny screen, trying to type their insurance ID correctly

This is why so many clinicians prefer voice for first touch.
Because voice feels natural.

→ An elderly patient can just speak.
→ Someone with an accent can clarify in real time.
→ A parent can answer while multitasking.
→ And if something sounds confused or anxious, you can respond to that tone.

Intake is the first moment of care.
And the smoother that moment feels, the less likely a patient is to drop off before they even walk in.

So, does your intake process feel easy for your team, or easy for your patients?

Only 1% of fast-scaling founders use AI the right way. Here’s why the rest are falling behind without realizing it.

Because they're using a Formula 1 engine to drive to the grocery store—unnecessary and ridiculous. 

They automate a few emails, generate some content, call it "AI-powered growth", and wonder why the results feel the same.

But AI should be plug-in to do things you never could before.

Like knowing exactly which user is 3 days away from churning and reaching them with the right message before they even think about leaving.

Or running 40 personalization experiments simultaneously, without a single extra hire.
That's a completely different game.

So, don’t ask "How do I automate what I already do?" 
Instead, ask "What became possible that wasn't possible before?"

That second question is where the real growth lives.

I've seen teams 3x their conversion velocity by finally having the bandwidth to focus on strategy, narrative, and the decisions only humans can make well.

AI handles the repetition while you handle the direction.

The founders who get this early are building a completely different kind of moat.

What's one thing AI made possible for your business that wasn't possible 2 years ago?

We could automate 100% of healthcare calls tomorrow.
But we don’t.


It’s because the moment you sit with real teams and listen to real calls, you realize that full automation feels impressive in a demo, but reckless in real life.

Most of the pain in frontline work comes from endurance.
↳ The same appointment reminder.
↳ The same eligibility check.
↳ The same “can you call us back later?” loop that eats an entire afternoon.
That’s the work AI should do.

So, voice AI should not be treated like a replacement for judgment.
It isn’t, and forcing it to be one is how trust breaks.

→ If a patient sounds confused, we don’t “power through” the call.
→ If something feels off, we escalate.
→ If the situation needs context, accountability, or empathy, it stays human.

That restraint is intentional.

Because automating everything is easy.
Knowing what not to automate is the hard part.
Isn’t it?

Your clinic is leaking 6 out of 10 potential appointments before a human ever says 'Hello.' And the culprit is a system we’ve normalized since 2003.

If you’ve ever called a hospital, you know the ritual.
“Press 1 for appointments.”
“Press 2 for lab reports.”
“Press 3 if you’re already confused.”
“Press 9 to go back to the previous menu that didn’t help.”

And then comes the robotic voice that sounds like it last felt emotion in 2003.

By the time you reach a human, you’ve explained your problem three times
and forgotten why you called in the first place.

Here’s the irony:
Healthcare is supposed to be about care.
But the systems around it feel anything but caring.
→ Long wait times.
→ Rigid IVR trees.
→ No context.
→ No memory.
→ No real listening.

And the patient is already stressed and anxious.

But what if the system actually listened?

→ Not just keyword matching.
→ Not “Sorry, I didn’t understand that.”
→ Not pushing you back into menu jail.

But listening with context.

Understanding:
– “I need to reschedule my MRI.”
– “My father’s report isn’t showing in the portal.”
– “I’ve been on hold for 12 minutes.”

And responding like a calm, efficient front desk person.

That’s where contextual Voice AI  with empathy comes into the picture.
It is your competent front desk staff member who never gets tired, never forgets, and never puts you on unnecessary hold.

Because in healthcare, being heard is the baseline. Isn’t it?

CBTS

Indirect hire at American Express - Global Loyalty and Benefits Program.

Senior Software Engineer (AmEx)

topmate.io

Resume review, 1:1 mentorship and discussions, Interview preparation tips etc.

Top 1% Mentor

Multipl

Core Engineer, Backend team, deploying new features for business requirements, integrating and developing new APIs for Multipl

Software Engineer

ZopSmart

Worked with team to build a GoLang backend architecture with YugabyteDB along with ETL (extract, transform, load) services  for e-retail business, wrote scripts, unit tests, integration tests, worked on Microsoft Azure, Grafana, Dynatrace, ServiceNow and other tools for development, monitoring and testing purposes.

SDE - 1

ACS India Pvt. Ltd.

Worked on IRB (Industrial Robot Body) 1410 5kg and wrote firmware scripts for programming 8 axis level robotic machines.

Intern

House of Science

Worked on designing and building custom Google Home devices using embedded C.

Hey there! I am an electronics engineer. I like building products, if you have any plans feel free to DM me. I am friendly, no caution boards😂. I play badminton 🏸 and I play badminton on the TT table too.

taplio

The Battle

Shubham Khoker

Utkarsh Nimje

Followers

Posts

Engagements

Likes

Performance ScoreShubham Khoker's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Shubham Khoker

Utkarsh Nimje

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Shubham Khoker's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreShubham Khoker's LinkedIn Performance