Sparsh Gupta

Founder @ | Helping B2B companies unlock growth through AI agents & Automations

[Sparsh Gupta @sparshai]

@attio's founders had a CRM with paying customers and steady revenue.

They shut it down anyway.

Not because it failed. Because it worked in a market that could never get big enough.

Here is the origin story, because it explains a lot about the product.

@nicolasosharp ran business development at Passion Capital. His job was relationships, and his tools were painful. Ancient enterprise systems built for big sales teams, or lightweight tools that could not handle real deal flow. So he started building his own. VC by day, developer by night.

In 2017 he teamed up with @byteofbits and built Fundstack, a CRM made for investors. It worked. Real customers, real revenue, plans to expand.

Two years in, they hit the ceiling. The problems they were solving were not VC problems. They were CRM problems. But the VC market was too small for the product they knew should exist.

The product worked. The market was a cage.

November 2019. They scrapped it and went after Salesforce and HubSpot instead.

Then came the three year silence. From 2019 to 2022, Attio built in the dark. No press. No launches. No marketing. In Sharp's words, they were not in stealth. They just were not doing marketing.

Three years went into one thing: the data model. Most CRMs assume your business works like their template. Attio built a system where any object, any relationship, any workflow could be modelled. Flexible like Notion, powerful like Salesforce.

It nearly broke them. They interviewed 400 engineers and could not convince anyone to join. Christie got a job offer from Amazon and turned it down at the last minute.

The question that kept them building: if you were starting again today, would you still start this company?

The answer kept coming back yes.

December 2022, still in beta, they crossed $1M ARR on word of mouth alone. March 2023, Attio launched publicly with a $23.5M Series A led by Redpoint.

Here is why this matters if you are choosing a CRM in 2026.

Legacy CRMs are bolting AI onto data models designed 20 years ago. Attio spent three years building the foundation before the AI wave arrived. That is why AI inside Attio feels native instead of taped on.

The patience was the product.

At seed, this is the infrastructure you will scale on. At Series A, this is the migration you will not have to do later.

At Automation Jinn, we build AI native GTM systems on Attio as an Official Attio Expert Partner. We did not choose it for the story. We chose it for the foundation the story explains.

DM me if you are evaluating Attio and want to walk through what an AI native GTM stack looks like around it.

[Sparsh Gupta @sparshai]

@attio's founders had a CRM with paying customers and steady revenue.

They shut it down anyway.

Not because it failed. Because it worked in a market that could never get big enough.

Here is the origin story, because it explains a lot about the product.

@nicolasosharp ran business development at Passion Capital. His job was relationships, and his tools were painful. Ancient enterprise systems built for big sales teams, or lightweight tools that could not handle real deal flow. So he started building his own. VC by day, developer by night.

In 2017 he teamed up with @byteofbits and built Fundstack, a CRM made for investors. It worked. Real customers, real revenue, plans to expand.

Two years in, they hit the ceiling. The problems they were solving were not VC problems. They were CRM problems. But the VC market was too small for the product they knew should exist.

The product worked. The market was a cage.

November 2019. They scrapped it and went after Salesforce and HubSpot instead.

Then came the three year silence. From 2019 to 2022, Attio built in the dark. No press. No launches. No marketing. In Sharp's words, they were not in stealth. They just were not doing marketing.

Three years went into one thing: the data model. Most CRMs assume your business works like their template. Attio built a system where any object, any relationship, any workflow could be modelled. Flexible like Notion, powerful like Salesforce.

It nearly broke them. They interviewed 400 engineers and could not convince anyone to join. Christie got a job offer from Amazon and turned it down at the last minute.

The question that kept them building: if you were starting again today, would you still start this company?

The answer kept coming back yes.

December 2022, still in beta, they crossed $1M ARR on word of mouth alone. March 2023, Attio launched publicly with a $23.5M Series A led by Redpoint.

Here is why this matters if you are choosing a CRM in 2026.

Legacy CRMs are bolting AI onto data models designed 20 years ago. Attio spent three years building the foundation before the AI wave arrived. That is why AI inside Attio feels native instead of taped on.

The patience was the product.

At seed, this is the infrastructure you will scale on. At Series A, this is the migration you will not have to do later.

At Automation Jinn, we build AI native GTM systems on Attio as an Official Attio Expert Partner. We did not choose it for the story. We chose it for the foundation the story explains.

DM me if you are evaluating Attio and want to walk through what an AI native GTM stack looks like around it.

[Sparsh Gupta @sparshai]

Most teams don't have a CRM problem.

They have a single source of truth problem.

Revenue data spread across 8 tools. Product usage here. Billing there. Calls somewhere else.

No one sees the full account in one place.

@attio fixes this by becoming the system of record everything else feeds into.

The stack:
Snitcher→ anonymous site traffic becomes named accounts.

@ClayRunHQ → every record auto-enriched. Industry, headcount, contacts, filled in.

Segment (@twilio) → live product usage on the account.

@stripe → subscriptions and invoices on the same record as the relationship.

@meetgranola → every meeting note lands on the right account.

@intercom's Fin → works inbound 24/7 with live Attio access, hands over qualified leads.

@Mixmax → fires the right sequence the moment a record changes.

@SlackHQ → ask Attio anything, get pinged when something moves.

Now every account is one record.

Who they are. What they do. What they pay. What was said. What's next.

All in one place.

At seed: one founder seeing the whole customer without 8 tabs.

At Series A: a team working off the same truth, not arguing about whose dashboard is right.

A CRM you log into vs a CRM that runs your revenue motion.

Attio isn't where you store deals. It's where your revenue stack comes together.

DM me to design your Attio as the single source of truth.

[Sparsh Gupta @sparshai]

Over 3 million developers use @Railway.

For years, no CRM could model how a company like that makes money.

They sell on metered billing. Usage based, not seat based.

@salesforce. @HubSpot. Every legacy CRM pushed the same rigid deal and contact model.

None could represent it.

Their Head of Product Marketing called it "friction, friction, friction."

Outdated data structures. A full-time admin just to run it.

So their engineers coded their own adoption scores and piped alerts into chat.

Then they moved to @attio.

The Workspaces object mapped to their customer teams out of the box.

No bending their model to fit the tool.

An enrichment bot they wrote themselves.

New workspace signs up. A script checks the domain against their ICP database. Fills in industry, seat count, ARR, lifetime usage.

A home-grown adoption score.

Blended from Segment (@twilio), BigQuery, and manual inputs.

Piped into Attio as a live attribute.

Then the workflows.

Spend crosses a threshold. An alert streams to GTM chat.

Someone clicks "talk to sales." A deal record is created automatically and picked up in triage.

Product usage in. Pipeline out. One system.

The result: 100% GTM adoption. Zero admin overhead. Revenue, usage, and adoption at a glance.

This is the move legacy CRM can't make.

Salesforce and HubSpot were built around deals and contacts.

Never built to hold live product usage per account.

Attio's Workspaces object was.

At seed: a founder runs sales assisted on top of self serve, no RevOps hire.

At Series A: a real expansion motion, run where your product data already lives.

DM me to build a Railway style usage to pipeline setup on your Attio.

[Sparsh Gupta @sparshai]

The renewal you lost last quarter wasn't lost at renewal.

It was lost 40 days earlier, when the account stopped logging in.

You couldn't see it. Usage data lives in your product DB. Accounts live in your CRM. The two never speak.

Here's the workflow to make @attio see churn coming.

Start with the data model.

Attio ships two objects most teams ignore: Workspaces and Users.

Built for software companies tracking product accounts.
Pipe signup + usage into them.

Segment by @twilio, ETL like @HightouchAI or @polytomic, or a direct API push from your backend.

Now every account carries real product activity. Not just deal history.

One trigger.

Weekly active users on a Workspace drops past a line you set.

Say 30% in 14 days.

One motion.

A workflow watches that attribute. When it crosses the line, Attio:
flips the account to At risk
assigns the owner
creates a "usage dropping, reach out" task
posts the drop to @SlackHQ

The rep experience:

40 days before renewal, your CSM sees "usage down 38%, renewal in 6 weeks."

Still time to act.

Not a post-mortem. A heads-up.

This is why teams leave legacy CRM.

@HubSpot and @salesforce were built around deals and contacts.

Never natively built to hold live product usage per account.

Attio's Workspaces and Users objects were.

At seed: a founder spots a quiet account before it cancels.

At Series A: CS runs on leading indicators, not churn post-mortems.

Your product already knows which customers are drifting.

The job is getting that signal in front of a human early enough to matter.

DM me to turn Attio into your churn radar.

[Sparsh Gupta @sparshai]

@nicolasosharp @attio Pretty interesting. Love how you shipped a lot of new integrations in addition to the massive New Workflow update.

[Sparsh Gupta @sparshai]

@attio just shipped the new Workflows.

This is a genuine step change, not an incremental update.

Most CRM workflow features are glorified if-this-then-that.

What Attio just released is a different category.

Four things stand out.

Custom Agents. You set a goal and the agent runs the work. Not a fixed sequence you map out in advance. You describe the outcome and it figures out how to get there using the context already in your CRM. This is the shift from automation to delegation.

Ask Attio builds the workflow for you. You describe what you want in plain language and Attio constructs it. The gap between having an idea and shipping it just dropped.

MCP tools. Connect agents to every tool in your stack. Your Attio agent can reach into the rest of your tooling instead of living on an island. Most teams will underestimate this. It unlocks the most.

Multi-triggers, retries and more on a new engine. The unglamorous infrastructure that makes everything above reliable in production. This is what tells me they rebuilt the foundation, not just the surface.

Here is why this matters for revenue teams.

The bottleneck in GTM automation was never ideas. It was the gap between imagining a system and shipping it. Attio just narrowed that gap dramatically.

But lowering the barrier raises the ceiling.

Describing one workflow is now easy. Architecting a system where custom agents, MCP connections and multi-trigger automations run your entire revenue motion is a different skill entirely.

The teams that win will not be the ones who build one agent. They will be the ones who design the whole system.

Kudos to @byteofbits and the Attio team.

Building an AI-native workflow engine on a flexible data model is genuinely hard. This is the most ambitious release I have seen from them.

[Sparsh Gupta @sparshai]

Hiring a head of sales before your data model is clean is like hiring a chef before you build a kitchen.

The chef is talented.
But they will spend three months finding ingredients, working around broken equipment and compensating for a space not designed for how they cook.

Most Series A companies hire a head of sales into this exact situation.

CRM is a mess.
Pipeline stages mean different things to different reps.
No qualification framework anyone follows.
Forecast built on gut feel dressed up as data.

The head of sales arrives.
They are good.
They start fixing the data model while building the team, running the process and closing deals.

Three months in they are exhausted.
The board is asking questions.

The problem was never the head of sales.

Here is what most founders miss.

The data model is not just a reporting problem.
It is the onboarding manual for every rep who joins after you.

When your head of sales arrives, they learn how you sell by reading your CRM.
Which questions matter.
Which deals you won and why.
What your ICP actually looks like.

If the data model is incomplete, that knowledge never transfers.
Every new rep reinvents the motion.

Fix the data model first.

One pipeline. Clear stage criteria. A qualification framework that enforces itself.
Fields that populate automatically.
A CRM that reflects what is actually happening, not what someone remembered to type.

Then hire the head of sales.

They will move faster, build better and stay longer.
Because the infrastructure underneath them actually works.

[Sparsh Gupta @sparshai]

Most VCs are not seeing 80% of their inbound deal flow.

They are seeing what an associate had the bandwidth to review.

This is not a people problem.
It is a system problem.

First pass triage at most funds still works the same way it did ten years ago.
Associate reads decks.
Makes a call on thesis fit.
Decides what reaches the partners.

20 to 30 minutes per company.
Multiplied by every inbound that week.

The best deal in your next fund will not announce itself. It will arrive on a Wednesday afternoon, get a quick skim, and sit in "review later" until the round closes without you.

Here is what changes when the system is built properly.

Every inbound creates a record in @attio automatically.

An AI step assesses each one on arrival against three things.

Thesis fit. Stage, sector, geography. Not a keyword match. A judgment call.

Signal quality. Warm intro? Trusted co-investor reference? Prior founder pattern?

Timing relevance. Actively raising? Right stage for your deployment cycle?

The associate opens a ranked list. Not 50 unread pitches.

The ones worth a first call are flagged.
The ones outside thesis are filed.
No deal buried because it arrived on a busy week.

And the slow leak most funds do not talk about gets fixed too.

A company you passed on quietly grows into exactly the kind of company you would back.
Nobody noticed because it was sitting in a parked pipeline.

A workflow watches those records.
New funding announced.
Key hire made.
Traction signal fired.

The deal comes back to the top at exactly the right moment.

This is not a replacement for judgment.
It is infrastructure that makes judgment faster and more consistent.

[Sparsh Gupta @sparshai]

If you run a PLG motion on @HubSpot, your CRM and your product are telling two different stories.

HubSpot can handle PLG.
It was not built for it.

Designed around contacts and deals.
Product data lives in custom objects you configure and maintain yourself.
Every usage signal is a layer of middleware to manage.

The specific problem.

In PLG the unit that matters is not a contact.
It is the account and the people using the product inside it.

HubSpot has no native Workspace or User object.
You build it yourself.

@attio ships both as standard objects.

Users are people in your product.
People are your contacts.

Keep them separate and a champion keeps their full history even when they change jobs.
Product behaviour and sales context in one place.

Connect through Segment by @twilio or your warehouse via @HightouchAI or @polytomic.

Workspace crosses 80% of its seat limit.
Expansion play fires before a rep would have looked.

Free user invites four teammates.
Flagged as a PQL while intent is warm.

Usage drops for two weeks.
Churn risk task opens while there is time to act.

Not reports you check.
Signals that find you.

HubSpot can be configured to run a PLG motion.
Attio was designed to run on one.

[Sparsh Gupta @sparshai]

@meetgranola. @modal. @replicate .

Some of the fastest moving AI native companies right now.

All of them chose @attio.

Not because of brand recognition. Because of how the product is built.

@theshre, Head of Business at Granola, put it plainly.
"Connecting our whole stack was a real lightbulb moment. Suddenly, we could see the right people, the right context, at the right time. That's when Attio felt magical."

That is not a review of a feature. That is someone who launched the Granola Enterprise motion.

Here is what AI native companies ask when picking a CRM.

Not: how good is the UI? But: can an agent read and write to this cleanly?

Not: does it have good reporting? But: can an agent make decisions using this data?

Not: what does onboarding look like? But: will this still fit at 10x scale?

Legacy CRMs were designed for humans entering data manually. The schema is fixed. AI got bolted on top of architecture never built to support it.

Attio is different in three ways.

The data model is a graph. Every record relates to every other in any direction. Nothing is siloed.

The objects are flexible. You define the schema, not the vendor. Modal used this to sync product data for outreach. Granola cut lead triage time by 83%.

The API was built for agents. @clay_hq enriches records the moment they arrive. @claudeai reads transcripts and writes back to deals. No human in the loop.

AI native companies are not picking Attio because it is the newest option.

They are picking it because it is the only CRM where the infrastructure matches the way they actually build.

[Sparsh Gupta @sparshai]

Your @affinity__crm workarounds are not a you problem.
They are an Affinity problem.

Notes drafted elsewhere and pasted in after.
A Google Sheet for the second fund.
Meeting context that never quite made it into the CRM.

That is not your team being undisciplined.
That is a product hitting the edge of what it was designed to do.

Affinity was built to answer one question. Who do you know?

Relationship scoring. Network pathing. Auto data capture. For that job, genuinely good.

But the question most funds are asking now is different.

What can your stack actually do with that data?

You want @clay_hq enriching deal flow automatically. You want @claudeai scoring companies against your thesis.
You want meeting notes landing in the deal record without copy paste.

Affinity's AI reads your notes and answers questions.
It does not score, write to fields, or trigger actions.
It surfaces information. It does not act on it.

So the workarounds stack up.
The parallel systems multiply.
The CRM that was supposed to be the source of truth becomes the thing the team works around.

@attio was built for the world those tools exist in.

Flexible objects that model how your fund works.
An API that Clay and Claude navigate without friction.
Relationship scoring on your terms, not a black box algorithm.

The migration is simpler than most expect.
Contacts, companies and notes export cleanly.
Most funds are live in two to three weeks.

If your team works around Affinity more than in it, that is the signal.

[Sparsh Gupta @sparshai]

This is from @kylecnorton's chapter in @attio's GTM Atlas. Free resource. Operators from @Lovable, @vercel, @framer sharing how they actually run GTM today. One of the best GTM resource in the market right now.

[Sparsh Gupta @sparshai]

Most Series A teams are doing AI wrong.

Not distributing it wrong.
Distributing it at all.

Giving every rep @claudeai and a few prompts is not an AI strategy.
It is chaos with a chatbot.

@kylecnorton, CRO at @ownercom, took their GTM team from $2.5M to nearly $100M ARR in under four years. He wrote about this in @attio's GTM Atlas.

His take: somebody has to own the centralisation.

Take in all the good ideas.
Build them to production quality.
Deploy one version to every rep.

The output difference between a specialist and a non-specialist is not 50% better.
Not 100% better.

20x better.

His team built AI PCR. Pre-call research. One specialist. Built once. Deployed to everyone.

Top outbound BDR: $174k closed-won in a single month. Cold.

The math on getting there is simple.

GTM engineer costs twice a BDR.
Take two BDR headcount.
Put it into one GTM engineer or a consultant.

Fix the system first.
Everything else gets better.

At seed, the founder owns this before it becomes a mess.
At Series A, the head of sales makes this call before adding the next two BDRs.

[Sparsh Gupta @sparshai]

Manual CRM updates killed your pipeline visibility.

Ask your head of sales how confident they are in the pipeline.
Really ask them. Not in a board meeting. In private.

Most will pause before answering.

Because the CRM shows what reps remembered to log. Not what is actually happening.

Call happened Tuesday.
Record updated Friday.
Notes are three bullet points that mean nothing.

Deal moved forward in the rep's head.
CRM still shows the old stage.
Forecast built on stale data.

New hire joins.
Inherits the same habits within a month.

The system was never designed to get updated in real time by people who are also trying to sell.

Here is what changes.

Meeting happens.
@meetgranola captures the full transcript automatically. @claudeai reads it and extracts what matters.
Summary written directly to the @attio contact record. Deal stage updates based on what was said.
Next steps logged before the rep closes their laptop.

The CRM reflects reality. Not memory.

Pipeline visibility becomes something you can actually trust.

No more guessing which deals are real and which ones just look real.

At seed, the founder finally has full context on every deal without being on every call.
At Series A, the head of sales stops rebuilding context in every pipeline review.

[Sparsh Gupta @sparshai]

Your reps should not be doing lead qualification.

Not because it is beneath them.
Because it is too slow and too inconsistent.

Here is what most Series A teams are doing today.

Inbound lead arrives.
Rep researches the company.
Checks LinkedIn.
Assesses ICP fit on gut feel.
Decides whether to reach out.

20 minutes per lead.
Happens differently for every rep, every time.

Here is what it looks like with an agent in the middle.

Lead arrives.

@clay_hq enriches it. Funding signals, headcount, tech stack.

@claudeai scores it against your ICP.

Not a filter. An actual assessment.
Does this company need what you sell?
Is the timing right?
Is there a signal that makes now the moment?

High-fit lands in @attio. Enriched, scored, prioritised. Rep gets a @SlackHQ notification with their top accounts for the day.

Low-fit goes into nurture automatically. No rep time spent. No leads lost.

Your reps open Attio and the work is already done.
They are not qualifying.
They are closing.

And they are reaching out while intent is still warm.
This is not a future state. Teams are running this today.

At seed, the founder stays close to every lead without drowning in research.
At Series A, the head of sales builds a consistent process before the next rep joins.

[Sparsh Gupta @sparshai]

@firefliesai @attio @meetgranola @fathom_ai I've seen Claude work well for this usecase. Also it needs to pull data from other sources(emails, CRM context etc) for coaching and qualification not just the transcripts. Hard for Fireflies to be that layer. Claude can.

[Sparsh Gupta @sparshai]

Your rep had a great call today.
Your @attio record has no idea.

@meetgranola. @firefliesai. @fathom_ai. Attio's native recorder. Whatever you use.

Transcripts are clean. Summaries are good.

But none of them live inside your CRM.

The coaching insight sits in one tool.
The deal lives in another.
The rep is the only bridge.

Which means most of it never crosses.

Rep finishes a strong call.
Notes are there.
Attio record looks exactly the same as before.

Qualification fields empty.
Economic buyer unconfirmed.
Next step vague.
Deal stage unchanged.

The call happened. The deal did not move.

Here is what changes when @claudeai sits in the middle.

Call ends.
Claude reads the full transcript.

It does not summarise. It evaluates.

Was economic impact quantified or assumed?
Was the real decision maker identified?
Was a next step agreed with a date?
Was pain described in the prospect's own words?
Which criteria are still open on this deal?

Everything written directly to the Attio deal record. Structured. Field by field.

Missing economic buyer?
Field stays empty until confirmed.
Visible in your pipeline.
Flagged in every review.

Deal cannot move stage until criteria are met.
Your methodology enforces itself through the CRM.
Not through manager reminders.

Rep gets a @SlackHQ message within minutes of hanging up.

What they did well.
What is still open.
What to find out next time.

Coaching in the moment. Not two weeks later.
Works with whatever framework your team runs.

MEDDIC. SPICED. BANT.
Or the playbook you built yourself.

At seed, founder close to every deal without being on every call.
At Series A, head of sales building process before the next hire joins.

You already paid for the notes.

Claude makes sure they change how the deal is managed.

[Sparsh Gupta @sparshai]

Your best rep is not logging calls.

You have noticed.
You have said something.
They nodded and kept not doing it.

Here is the thing.

It is not laziness.
It is feedback.

Top performers optimise for closing, not compliance.

When the CRM creates friction, they route around it.

They work off memory.
They track deals in @SlackHQ.
They keep a personal spreadsheet.
They batch update on Friday when the data is already stale.

And because they are your best reps, they still hit quota.

So the problem stays invisible.

Until a deal slips because nobody updated the record. Until you lose pipeline visibility before a board meeting. Until a new hire copies the habits of the person next to them.

The reps are not the problem. The CRM is.

A system your best people work around is not a system of record. It is a system of friction.

@attio gets used because logging feels like less work than not logging.

The UI does not fight you. The data model matches how deals actually move.

The best signal your CRM is working?

Your best rep uses it without being asked.

If that is not happening, the problem is the tool not the team.

[Sparsh Gupta @sparshai]

@saastr just graded @HubSpot and @attio the exact same for AI agent readiness.

Both scored 82.
Both got an A-.

But that number hides a much bigger story.

HubSpot took a decade to get there.
Thousands of developers. Years of iteration.

Attio got there in a fraction of the time.

That trajectory matters more than the snapshot.

Now here is what the grade does not capture.

The 6 criteria: rate limits, auth, webhooks, REST, SDKs, error handling.

Zero criteria for:
The data model underneath.
How flexible the objects are.
Whether agents can do meaningful work with what they retrieve.

That is where the real difference is.

HubSpot's API sits on top of a rigid data model.
Custom objects require Enterprise.
Relationships between records are limited.
Built for humans entering data. Not agents doing the work.

Attio was built assuming AI does the work.

Flexible objects.
Graph data model.
Every record relates to every other in any direction.

The kind of architecture that makes an agent genuinely useful.
Not just technically connected.

Attio is already the CRM of choice for @meetgranola, @modal and @replicate.

Not because of brand.
Because the architecture fits how they build.

Same score.
Completely different foundation underneath.

The question for 2026 is not which CRM has the most mature API.

It is which CRM was built for the world where agents do the work.

That answer is already clear.