Grit Sep 21, 2026 1h 6m 43m saved
With Jesse Zhang, co-founder and CEO of Decagon
Decagon went from about 50 people to 500 in a year, which on the host's arithmetic means somewhere north of 2,000 interviews.
Most founders describe the search for an idea as a period of months. Jesse Zhang says his was three weeks, because he and his co-founder decided in advance that the search itself was the thing to avoid.
"I think most founders shy away from being as aggressive as you need to be in that phase because it feels like you're being too pushy or too salesy, but no one really cares about that."
Zhang, co-founder and CEO of Decagon, on Grit, sold his first company to Niantic and says it taught him almost nothing about running this one. Decagon builds AI agents for customer support across voice, chat, email and SMS, and the host, Joubin Mirzadegan, is a partner at Kleiner Perkins who now runs a company of his own.
The full interview is covered here so you can skip it. 66 minutes of audio, 23 minutes of reading.
Here are the 21 lessons that matter.
Key Takeaways
Product-market fit was six customers and $1M of revenue, with only the two founders selling
Three weeks in the idea maze, because avoiding a long one was the explicit goal The technique was sales discovery, not user research: budget, ROI, who signs
The team went from about 50 to 500 in a year, which is more than 2,000 interviews
Every round so far was preempted, roughly six months apart, with no actual fundraise
A round is a marketing event as much as a financing one
The seed came before the idea — a16z funded the pair before they had incorporated
Investors and founder friends called the idea too obvious and told them not to do it
Against Sierra, the difference he names is the deployment model, not the model quality Forward-deployed engineering against a product the customer operates itself
Marketing was the function he understood least and hired for late
"I would not describe it as fun" — the word he uses for the job is alive
He works every day except Saturday, and calls stability at home a prerequisite for the job
1. The Rectangle
The interview opens on an office move. Decagon had just left a space that held 70 people at a squeeze for one that takes 250 to 300 across two floors, and Zhang has a strong view about the shape of a room.
One big open space is underrated
I actually feel very strongly having one big open space is like very very undervalued
Jesse Zhang
The office they were sitting in is doughnut-shaped because it sits in a high-rise, and he dislikes it: from one side you cannot see the other side, or tell when people are eating. The new one is a rectangle.
Mirzadegan's own experience of moving into a bigger space was a kind of cognitive dissonance — pleased with it and quietly panicking about whether the company would fill it. Zhang said he has never had that, because they have never gone for very nice offices and always knew they would grow into them.
2. Hiring Behind The Demand
Asked whether hiring ahead of the demand was frightening, Zhang said they mostly did not do it.
They hired only when it already hurt
But we only hired someone when we actually felt the pain of okay like we need someone doing this and that was for sales as well
Jesse Zhang
He and Ashwin Sreenivas did all the selling at the start, and by the time the first salespeople arrived there were already deals waiting for them. In hindsight, he said, that may have been a mistake in a market this good. In the moment it did not feel like one, because everybody knows the stories of companies that hired ahead and had to reverse it.
That changed once the growth rate did
Now, of course, we do hire ahead just because you have to.
Jesse Zhang
Past a certain size, reactive hiring breaks things.
3. 50 To 500 In A Year
The scale-up number is the one the section turns on.
A tenfold year
We pretty much 10xed the team last year. We're 500 now.
Jesse Zhang
They were roughly 50 people 18 months in. The company is no longer only in San Francisco: sales is remote, and there are hubs in New York and London, about 50 people in Europe, and a new Australian office he flew out for. He goes to New York at least monthly and to Europe about quarterly.
He stopped knowing everyone's name at 200
I'm actually pretty good with names, but that probably stopped at around 200ish people.
Jesse Zhang
The first crack in the in-office rule was strategic salespeople, on the reasoning that they would be traveling most of the time anyway, so keeping them in one office defeats the point.
4. Bar Raisers
Mirzadegan put the arithmetic to him: going from 50 to 500 conservatively means about 2,000 interviews. Zhang said it was probably more, split across the team.
Volume is not the constraint
The hard part is not so much hiring the people because there's a lot of people that want to work in, you know, hot AI companies, it's making sure you're you keep your bar high.
Jesse Zhang
The mechanism is borrowed, and he names the sources.
A final filter for culture and intensity
you kind of have these like filters that you add on to the interview process at the end. Bar raisors or whatever you want to call them where they're just mostly selecting for culture and you know intensity
Jesse Zhang
Revolut does a version of it and so does Amazon.
The founders were the first bar raisers
Well it was me and Ashwin for a long time. We met everyone up until maybe like 300 people.
Jesse Zhang
Now a group of long-tenured employees runs those interviews. Asked what share of candidates fail at that stage, Zhang did not have a number and said it is a clear majority who pass, which Mirzadegan agreed is the only sane answer — failing most people at the last step would mean something earlier is broken.
5. A New Company Each Qtr
Asked how often the internal processes have to be torn down and rebuilt, Zhang gave a rate.
The company reinvents itself two to four times a year
it does feel like kind of a new company every quarter or two roughly
Jesse Zhang
Neither founder has run a company this size before, and he treats that as a mixed blessing: mistakes get made, and the naivety also supplies the optimism, the same way it does when founding anything.
Mirzadegan's addition was that prior experience helps less than people assume. Arvind Jain had founded Rubrik before Glean and has still never run Glean at this size, in this technology wave, with this set of variables — and Marc Benioff has never done this in this moment. The conclusion he draws, and says he finds comforting, is that everybody is making it up, including the best-regarded competitor in Decagon's own market.
6. Why He Won't Solo Found
Zhang's first company had him as the only founder, and he gave two reasons he will not do it again.
The psychological one
you feel like you're like a single point of failure for a lot of things
Jesse Zhang
However large the team gets, only a founder carries the founder's context and responsibility.
And the operational one
you don't really have someone with the same level of like context and commitment to bounce ideas off of
Jesse Zhang
Talking out loud with someone who has the same context gets you to a decision, usually the right one, far faster than sitting in your own head. He added a caveat that did not exist when he was solo.
Part of that is now available from a model
Now you can actually do some of this with Claude, which is kind of interesting.
Jesse Zhang
It gives you somewhere to vocalize an idea and hear it back, which kills the bad ones quickly.
7. Wife Day
The founders are in the office every day, with one exception.
Six days, and a named seventh
we're in every day except for Saturday. Saturday is like wife day.
Jesse Zhang
Sunday is a working day for a self-selecting group, and he was careful about how it came to be that way. There was never a moment where anyone was told they had to come in, and no one is expected to answer anything.
Nobody is on call
no one's expecting you to respond on Sunday
Jesse Zhang
Why people come anyway
Sundays is a great time. There's no meetings. You can just like, you know, get your thoughts in order and catch up on stuff.
Jesse Zhang
He and Sreenivas live close to the office, which is most of why they are there.
8. Zero To One Is Awful
Asked whether the job is more or less enjoyable than it was, Zhang said the feeling barely changes — like being a child who gets taller without any single day feeling taller. The one real discontinuity is behind him.
The early phase is not romantic
I think no matter what anyone says, the early like zero to one phase is like way more way harder. It's awful.
Jesse Zhang
And nothing since has come close
every stage has its challenges, but the zero to one phase is like by far the hardest
Jesse Zhang
He spent about three years in that phase at his first company, starting at 20 as a new graduate, working very hard on things that a little more thought would have shown were bad ideas. Having done it, he says it is easy to appreciate not doing it now.
9. Three Weeks In The Maze
Then the number that gives the section its name. Asked how long the idea search took at Decagon, he said about three weeks.
Avoiding the search was the plan
our main goal when starting the company was like let's not be in the idea maze
Jesse Zhang
What they did instead was run something closer to a sales process than to customer research.
The mistake founders make is being too polite
I think most founders shy away from being as aggressive as you need to be in that phase because it feels like you're being too pushy or too salesy, but no one really cares about that.
Jesse Zhang
Nobody minds a founder asking hard questions, he said, and the soft ones tell you nothing.
The questions a salesperson would ask
where's the budget coming from? Like how do you justify ROI internally? Like who are the people involved?
Jesse Zhang
Asking "would you use this" and "how useful would this be" is how founders end up unable to tell.
10. A Seed Before The Idea
The financing ran ahead of the company.
Funded before incorporation, with no idea
we raised our seed round like before Ashwin and I started working like we decided like hey let's be co-founders and then we had even incorporated yet and we had like zero ideas but then we just raised a small seed round from a16z
Jesse Zhang
He said the seed was clearly the right call for them, and that the value of it was credibility they could not otherwise have had at that point.
It was also, on the conversation's own telling, a room full of people who thought the plan was wrong: when the two were deciding what to build, most of their investors and a number of founders they knew told them not to do it, because the idea was too obvious.
11. What Early Money Is For
Mirzadegan's answer to what founders should optimize for in early investors starts with what they cannot do.
Nobody buys you product-market fit
no investor will help you find PMF faster like that's like not an expectation like no matter how good the investor is
Joubin Mirzadegan
What is left is not doing harm, emotional support, and hustling introductions to customers and hires. From that he draws a conclusion about seniority.
Mid-career investors have the time to grind
optimizing for mid-career investors in the early stages I think is like actually the sweet spot because they have the time and the sort of motivation to like really grind for you
Joubin Mirzadegan
Very senior partners, on his account, are better once you are scaling. Zhang said he took a different approach, signal first, with the firm's wider team doing the work rather than the board member, and that there is no single right one.
His own test for a prospective investor is temporal.
Their most helpful moment is before they are on the cap table
no investor will ever be as helpful as during that like phase where they're trying to close a deal
Jesse Zhang
If you ask for help while they are trying to win the deal and the help is thin, that is the answer. Doing this before every round is why he makes an unusual claim about the result.
He thinks they use their investors harder than anyone
I think I think we're probably one of the companies out there, if not the single company that like leverages their cap table the most.
Jesse Zhang
It compounds, he said: an investor who helps and then hears that the help closed a deal helps more.
12. PMF At Six Customers
The threshold he names is small and specific.
What product-market fit looked like
Six customers, a million in revenue.
Jesse Zhang
And it was still only the two of them
I think maybe we were like a million in revenue. It was just me and Ashwin then still, but I think it was repeatable enough
Jesse Zhang
There was no cliff, he said, just a gradual scale-up, and the read was that enough people wanted this and were signing real contracts.
On the category he picked
this space and coding maybe are like the two big ones for AI that are like very horizontal and just like very obvious
Jesse Zhang
Law and some healthcare uses are real, he added, but vertical. He was explicit that none of this came from market mapping, which he thinks is too noisy to be useful.
13. Every Round Preempted
The Series A followed almost immediately, and so did everything after it.
No fundraises, only term sheets
all of our rounds so far have been aggressively preempted
Jesse Zhang
At a rate he does not recommend
All the rounds were roughly like six months apart.
Jesse Zhang
Six months is too frequent all things equal, he said, but in heavy growth mode you raise to keep the momentum. His instinct has been the opposite, the same conservatism he applies to hiring, because a valuation raised too high is a problem that arrives later.
What the money actually buys after a point
raising gives you validation gives you sort of more people around the table
Jesse Zhang
Beyond that there are diminishing returns, and a higher valuation mostly helps recruiting, because candidates read the dollar value of the equity even though he thinks they should not.
The other function of a round
Yeah, it's a good marketing event.
Jesse Zhang
His suggestion is to raise smaller rounds more often, which Ramp does well, because every raise is an announcement people are guaranteed to read.
14. It Won't Stay Like This
The thing that makes him uncomfortable is how quickly good conditions come to feel like the baseline — customer budget falling out of the sky, fundraising staying easy.
A line they repeat internally
we say internally all the time that it's not always going to be like this
Jesse Zhang
He is not confident saying it works. Companies growing fast now, he pointed out, grew fast through the zero-rate era, went through a couple of rough years, and are growing again; the pattern is cyclical, and the quarterly recap says so out loud.
What he tells the company after a good quarter
when we recap the quarter we tell everyone like hey obviously it's great right now. Like everyone should be ready that it's not always going to be so easy.
Jesse Zhang
The instruction attached to it
So, you got to build the toughness right now. You got to build the grit right now.
Jesse Zhang
Very few people have left, though some have gone to start their own companies, and he does not expect a downturn to change that.
Starting a company is too easy for that to stop
it's also really easy to be a founder.
Jesse Zhang
Plenty of people at Decagon were founders or intend to be, and with the barrier that low they will go whether or not they have an idea or the timing is right.
The one upside of a downturn
I think the only good thing about a downturn may be that it might be easier to just recruit.
Jesse Zhang
There are too many companies right now, which he called good for the world and hard for hiring.
15. Generalists, Then Experts
Mirzadegan raised the tax that comes with hiring very ambitious people: more maintenance, and more of them wanting to leave and found something. His observation from a candidate debrief was that there is something restful about an engineer who simply wants to be better at engineering in three years, and that you cannot build a whole team of those but should have some.
Zhang's version runs by stage.
The early team cannot be replicated at scale
you cannot find like hundreds of those people
Jesse Zhang
The first sales team was smart, creative and self-starting, and there is a ceiling on how many of those exist.
So the shape of the hire changes
As you scale you kind of start hiring more specialists cuz you know now that now the functions are defined
Jesse Zhang
The intelligence bar stays where it was; the personalities widen. Some of the newer, more experienced hires want a playbook to execute against, which he treats as normal rather than a dilution.
16. Sierra And The Deployment
This is the competitive section, and Zhang prefaced it by saying Decagon historically does not discuss competition, because the space was too noisy for it to mean anything, and that both companies are established enough now for the comparison to be worth drawing.
The company sees itself as the main character
we just we view the world as like we're the main character
Jesse Zhang
On who he is up against
if I could hand select like a competitor would I have chosen Brett Taylor? Probably not.
Jesse Zhang
The former CEO of Salesforce and CTO of Facebook is not the draw anyone would pick, he said, with respect attached — and that is not how it works.
His read on how Sierra operates is that they are strong at getting into enterprises and have scaled a team quickly, with a delivery model built on forward-deployed engineering.
Their model puts the work on their own people
they're really focused on you know forward deployed engineering
Jesse Zhang
Decagon's founding thesis was that this makes less sense as the models improve. The case for doing it for the customer is that the customer lacks the internal talent to move at that speed — and the thing agentic AI is best at is removing that constraint, the same way coding agents let non-technical people build.
What the AI is for
One of the main benefits of AI, especially Agentic AI, is that it really empowers people on the team to do more.
Jesse Zhang
So Decagon still staffs people onto enterprise deployments, because it is never just handing over software, but the object of the exercise is different.
Everything gets built inside the product, not in code
Our whole thesis is like how can we empower your team and how can we deliver a end result where the AI is fully in there. It gets built in the product. It's tested in the product. All the conversations that it's having you can monitor in the product.
Jesse Zhang
And that is the claim he is making
We think that is probably our big differentiation right now. This deployment model where ultimately the customer owns the results and the process.
Jesse Zhang
In a market whose defaults are Salesforce and Google, the only way to take large deals from a well-liked competitor, he said, is the product itself.
No relationship substitutes for it
you have to build a product that is superior
Jesse Zhang
17. Marketing Was The Gap
Asked which function he most misunderstood, Zhang answered immediately.
The one he got most wrong
The first one that comes to mind is probably marketing.
Jesse Zhang
It is broad and fuzzy, and there is a hiring hazard built into it.
The people are good at selling themselves
And the people in marketing, just like in sales, are very good at marketing or selling themselves.
Jesse Zhang
He thinks he hired for it slightly late, and that the reason is the difficulty of naming the pain point. Product marketing, demand generation, brand and social are, on his account, four different jobs with very little systematic overlap and a real element of art.
The first marketer arrived in the first 30 people
It was first 30 people.
Jesse Zhang
It was a product marketer, and his VP of marketing came from Glean.
Why that order worked
the thing with product marketing is because it's more amorphous that actually like fits better in your early early company cuz your early company is like very amorphous
Jesse Zhang
Growth and performance marketing people come later, once there is something systematic to run.
18. Finding A Co-Founder
Zhang met Sreenivas through mutual friends, having known one of Sreenivas' previous co-founders, now a founder of Cognition. The two first talked about building something together at an Andreessen Horowitz offsite.
He thinks it is mostly luck
it is very luck based I would say but it's not that dissimilar to like finding a spouse
Jesse Zhang
Right person, right time, enough opportunities. Mirzadegan pointed out that this contradicts the standard advice, which is to work with someone you have known for five years.
He does not hold with it
If that is, then I don't agree with it.
Jesse Zhang
What he thinks actually predicts it
I think the thing that matter a lot more are, hey, do you want the same things out of life? I think that's like the most important thing.
Jesse Zhang
Both equally committed to being a founder, and the same ambition for the company — which he thinks is not aligned for most founding pairs. Operating at a similar level comes second. Long friendship may make those things more likely, but it is not the variable.
19. Advice Is A Data Point
Asked what founders come to him for, Zhang said it is everything, and the useful source is always someone slightly ahead.
A year ahead is the right distance
anyone that is like maybe a year ahead in terms of progress has a lot to offer
Jesse Zhang
Older companies he admires, like Ramp, are past the point of relevance to his problems.
Who has been useful
I think Harvey's been really gracious with their time just because they're exactly a year older than us
Jesse Zhang
The rule he applies to himself is to talk only about his own experience, because opinions about someone else's situation are dangerous without real depth on it. Mirzadegan said that is the point of the show — a specific answer to what number your first marketing hire was is a data point rather than a prescription.
And the supply runs out
there's fewer and fewer data points that you can ask. So each one is very valuable.
Jesse Zhang
Fewer and fewer companies have been at this stage in this environment in the last six months, which is why the relationships have to exist before you need them.
20. The Truth Sets Us Free
On when internal politics starts, Zhang's answer was: as soon as there is a team.
They treat it as something to stamp out
Ashwin and I have always been very like allergic to stuff like that. So, we try to stamp it out aggressively as soon as possible.
Jesse Zhang
Someone arguing they should own what a colleague owns gets asked why they are not busy with their own work. He does not think it can be avoided, only suppressed.
Two days before the recording an early engineer came to him with feedback about another person.
What he said
What are you telling me for? I don't want to hear any of this. Go tell the person.
Jesse Zhang
The alternative, he said, is sitting on it and letting it fester. Mirzadegan's framing was that this is what makes a cultural value useful: telling people to give feedback directly is actionable in a way that valuing transparency is not.
The value it maps to
one of our values is the truth sets us free
Jesse Zhang
And what he did afterwards
after he did it I texted him the truth set you free.
Jesse Zhang
21. Not Fun, Alive
Asked what has surprised him most, Zhang started with the commercial answer.
Demand arrived before the product did
we basically had PMF without the P like from incorporation and I know how rare that is
Jesse Zhang
Starting in the enterprise is hard mode in most respects, he said, and the market was the exception. The thing that has moved against them since is recruiting.
Offer acceptance has slowed
And now it's like much slower and it's because we're trying to scale the company and so we're just in more fights.
Jesse Zhang
The profile of candidate is different too, and the economics are different: the first ten people have ownership that the hundredth cannot be given.
Then the answer to the question he says he gets most, which is whether he is having fun.
He would not use that word
I would not describe it as fun. It wouldn't be the first word I would use. I feel like I feel alive.
Jesse Zhang
That is the surprise
That's like the most surprising feeling that I've had is like how alive I feel.
Jesse Zhang
Most of it is not the pleasant kind of alive — whether the thing ships, whether the customer who bought will actually deploy.
The surface area is what gets him
I think that's like the scope of things that I now care about is insane. I just can't believe how many things I have to care about.
Jesse Zhang
And some of it he would rather not
And there's some things that I just don't want to care about honestly. Like finance.
Jesse Zhang
Decagon has no finance leader and needs one; the hire keeps not making his top three. Mirzadegan's version of the same trap is that hiring is the only exit from it and the search itself takes the time you were trying to free up, which everyone in the organization feels at once.
The last thing Zhang volunteered was about the seven-day weeks he used to work.
Home stability is load-bearing
my ability to have stability and like bliss at home is like almost a prerequisite for me to be good in my job
Jesse Zhang
There is too much in the job to do it without peace of mind, and that requires deposits made over time. He had expected to redline straight through, and instead started giving himself and his partner Saturdays — after which, he said, the team followed.
Bonus Insights
His last company taught him nothing about this one
Zhang sold his first company to Niantic and rates its relevance at zero. What it did give him were preferences: he was a solo founder and disliked it, the team was remote because of the pandemic and he disliked that, and it was a consumer app, which he found he cared less about than business software. On how to run a company, he says, almost nothing transferred.
He does not like traveling, at all
I've never like traveling ever. I don't like traveling for fun.
Jesse Zhang
He does not like planes, cars or long walks, and prefers an air-conditioned room and his own work. The trips are his wife's choosing; they had been in the Dolomites over the July 4th weekend. He no longer feels guilty about going, partly because the team is large enough that nothing stops.
Remember this office
you should always remember this office because you'll think back at this as like the good old days
Jesse Zhang
He told the company that about the old, worse office, when 200 people could still eat dinner together. Catering is now the way to get everyone eating at the same time, and they have run break-even math on at what headcount it starts to pay; the first attempt drew complaints about the food.
The offsite problem
Getting the whole company together is now impractical. Mirzadegan described his own company's answer: a glamping trip at Auto Camp about three and a half hours away, reachable by bus, and probably the last time the whole team will fit. He said people were more giddy about it than about the largest deals they had closed. His reasoning is that colleagues spend enormous amounts of time together and almost all of it talking about work.
The show is the host's own data-gathering
Mirzadegan said he has recorded more than 200 episodes, used to do them weekly, and would have these conversations without a microphone. There is no question list; his selfish interest is asking founders what number their first marketing hire was.
Zhang's bottom line is that the AI companies competing for the same enterprise budget will not be separated by model quality, because the models are converging — they will be separated by who leaves the customer able to run the thing themselves.
Products, Companies & Tools Mentioned
Decagon (His company: AI support agents across voice, chat, email and SMS, 500 people and six customers at the point of product-market fit)
Sierra (The competitor, run by Brett Taylor; strong in enterprise sales, with a forward-deployed engineering delivery model)
Claude (What he now uses to do some of the thinking-out-loud a co-founder does)
Andreessen Horowitz (Seeded the pair before they had an idea or a company, and hosted the offsite where the two first talked)
Kleiner Perkins (The host's firm, where Decagon was working out of when it outgrew its first office)
Glean (Where Decagon's VP of marketing came from; Arvind Jain's company, cited as proof that prior experience does not transfer)
Cognition (One of its founders was a previous co-founder of Ashwin Sreenivas, and the connection that led to the introduction)
Ramp (Raises small rounds frequently, which he thinks is the better pattern)
Harvey (A year further down the same path, and the company he says has been most generous with its time)
Revolut and Amazon (Where the bar-raiser interview stage comes from)
Salesforce and Google (The default platforms whose solutions Decagon has to beat on product)
Niantic (Bought his first company; the experience he rates at zero relevance to this one)
WeWork (His view of what a first office should be)
Auto Camp (The glamping site for the host's company offsite)
Listen to the full episode
🔴 YouTube | 🔗 Episode page
Watch the full episode:
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: