Writer's May Habib On AI For Normals, Token Insanity, And Sexist VCs
A gendered rejection letter from a top VC firm. Shots at Anthropic and OpenAI token giveaways. On The Upstarts Podcast, Writer's CEO doesn't pull punches "People should be getting f***ing fired."
Writer CEO May Habib has a message for the parents out there in tech, and especially the women founders: if you think what you’re doing isn’t harder, you’re kidding yourself.
What started as an amusing anecdote about the recent discourse of founders sharing VC horror stories — “Vinod [Khosla] also fell asleep in my partner meeting at Khosla [Ventures],” Habib quips — has quickly gone somewhere more serious.
Reflecting on Writer’s Series A — ultimately a $21 million round led by Insight Partners in 2021 — Habib reflects that her startup would’ve had an easier time if she’d switched roles with co-founder and CTO Waseem Alshikh.
“If you’re a woman pitching VCs, we’re all fooling ourselves to think that in the back of their minds, they don’t think you’ll work as hard,” she says. “I think me and Waseem should have switched roles. Is that crazy? Have the technical man pitch.”
Sequoia, the famed VC firm we’ve mentioned a lot in Upstarts, rejected Writer during that Series A process. Sequoia’s rejection letter included a line about how the firm found it ‘inspiring’ that Habib was building her startup with two little kids at home, she recalls — a detail she thinks was a giveaway that the firm was concerned in a way it wouldn’t be with a CEO dad.
“Everyone wants to hear you say there is nothing more important to you than being a billionaire,” she says. “And the reality is: My kids are kinda more important than that. I think anyone would say their kids are more important than that.”
But don’t mistake that for a lack of ambition. Habib says she works 15 or 16 hours a day anyway. She’s got plenty.
And she believes that Writer, valued at $1.9 billion, represents the best chance for Fortune 500 boardrooms in weekend crisis meetings about AI adoption to see real results. Accenture, Cigna and Vanguard already use Writer’s AI agents to do everything from create morning sales briefings to personalized product test invites.
Writer isn’t as buzzy a partner as working with the big AI labs, but it’s building what Habib calls AI for “normal” people.
"Anthropic and OpenAI salespeople walk in like heroes, get a contract, and leave to literally never be seen again,” she argues.
And as those companies, well down the IPO path, gives away millions in tokens to secure business from each other — what Habib calls “absolute insanity” — she is calling for more accountability in what return on investment companies actually get from all their AI use.
There’s an adage in tech: nobody ever got fired for buying IBM.
Habib proposes an update: "People should be getting f***ing fired,” she says. “That’s the best way to infuse some sanity into this market.”
On this episode of The Upstarts Podcast, Habib goes into her journey starting and pivoting machine translation startup Qordoba into Writer; the VC who fell asleep in a pitch meeting; and how the real moat in enterprise is not leaving customers behind.
Plus, she shares her Upstart Moment: hiring top talent in the middle of a “gold rush.”
This whole episode is a must-listen, but our three biggest takeaways for busy builders can be found below.
A MESSAGE FROM OUR SPONSOR
Your AI is only as good as your data
Rippling AI is built on your live workforce data and permissions. So it gives you full visibility across your startup (on anything from headcount to burn rate) — and can take real action right inside the platform. Want to see how it can work for you?
Get Exclusive Access: rippling.com/ai
“Survive, don’t sell, don’t quit.”
The key to success in more competitive, harder startup categories isn’t any one decision, Habib says: it’s just staying in the game.
“This space rewards those teams that are physiologically set up to do hard things for a long period of time,” she says on the show.
“ The hype will pass, and that competitor that gets all the headlines, they haven't built the right thing for the right people, and [don’t] truly understand the customer. They may have their moment in the sun, but will not survive.
And for us, [it’s] the maniacal focus on customer outcomes, customer ROI. I know it's lame. No one wants to talk about ROI, right? All the cool boys aren't talking about ROI. ‘Why is May talking about ROI?’ Customer cares about ROI, right? Are they paying for software that does something?”
Writer is waiting for a “runaway market” that hasn’t fully materialized yet in the roles it sells to.
“We have a process that is really well structured to be able to turn human and AI collaboration into working code, right? Does that happen in sales? Does that happen in marketing? Does that happen in HR? We don't have the scaffolding to be able to take agents and get them into production, with human oversight that is well-known, well-established, and easy to audit.
That’s the scaffolding we are building for these teams. And that's the difference between production and scale.”
What’s missing today, Habib argues, is trust. When teams trust agents as much as they do people, agents can take off in those roles the way they have in code.
“That explosion is incredibly latent, right? We haven't announced these things, but we've met a lot of the milestones that people brag about. I think for us, the thing that we will be bragging about is when we have the same kind of agentic autonomy in the enterprise that we all have in engineering and coding domains, and we're not there yet, because the customer is not there yet. The trust is not there yet.”
Bringing sanity to tokenomics
As noted above, Habib advocates for more buyers to face consequences for spending resources badly on AI — the new IBM.
“Almost every one of our customers has everything under the sun, and they don't get results. And our ability to come in and help them get results is just about thoughtful product meets thoughtful implementation, meets thoughtful rollout.
And the token maxing across all aspects of the business… I think on the coding side, you know, Cursor and others have built in the actual budget constraints, right, and that functionality. Anthropic hasn't.
And so I'll meet people where it's a $3 million overage on two weeks of the plugin in Excel. Like, what's the ROI on that? Zero, right? Excitement, maybe. But if you're not learning the right lessons, then that is not money that is well spent. And CEOs are just having to get involved. I think CFOs will have to get involved.”
If she’s right, startups should take heed: and find ways to be on the right side of ROI. Writer found a sweet spot as a “weapon” for its own buyers, CIOs and CMOs, to reduce their spend on more expensive software from Adobe, Salesforce, and Workday.
At the same time, those companies are all customers of Writer, too.
“Internally, everybody is trying to understand how to get this kind of leverage and get the form factors right.”
Writer benefits now from the tailwind of companies adding, or shifting, people to have “AI” in their title and remit.
“I think I agree with Aaron Levie [CEO of Box] who wrote about this recently. I don’t know if he gave a percentage, but I personally think 10% to 15% of most sales and marketing teams are going to be AI-related titles.
And us being able to bring that kind of aggregated data to an executive team who has two people in the business dedicated to AI, is what allows them to say, ‘Oh, we need to 10X the human investment here, not just the software investment.’
And so, yes, you’re giving people leverage over this cost structure, but you’re also helping them take that into retraining and in a lot of cases, rehiring, for the kind of capabilities and skill sets they want in the company.”
You don’t need to be a lab to own the stack
We heard a version of this insight from HeyGen CEO Joshua Xu in our story about how his video AI startup reached $200 million in ARR with minimal burn: controlling the whole infrastructure stack, you can avoid getting caught passing along revenue to others via API.
At Writer, the company’s internally-trained Palmyra models power 90% of their LLM calls, Habib says.
Because Writer started training those models before ChatGPT’s release, the big decision wasn’t so much to train models, but to continue training them in recent years, she adds.
“We started Writer to commercialize transformers. As we continued to train models, we saw that while the labs might be a few weeks ahead, everybody’s following the same research breadcrumbs.”
Writer started to use synthetic data as early as 2023 to help account for having fewer resources in its own training.
“We just never really felt the capital gap meant a capabilities gap. Sure, they may be ahead, but there are plenty of eras of this chessboard, where we were ahead.
The thing is, with the enterprise, it doesn't really matter. Because they're still trying to catch up to the innovation that you launched 24 months ago. And so we never felt like we needed to go use third-party models to be able to deliver for customers, and to be able to deliver frontier performance for customers.”
Customers will use Palmyra models when they’re faster and cheaper, but Writer’s position is that customers at big companies like Mars and Unilever shouldn’t be bound to any one model provider. Others will want to fine-tune their own.
“It gives folks the optionality, but we're never not going to be a company that builds and trains its own models,” says Habib.
One other distinction: even as Writer now tests its latest model, Palmyra 6, it doesn’t see itself as a lab.
That’s because Writer doesn’t operate research, product and go-to-market separately. “It’s everything from a single Slack to a single force in front of the customer,” adds Habib.
In one recent example, Cigna gave feedback to Writer’s head of research, Dan Bikel, that in a certain use case, Palmyra 5 performed better. “This is why I will be perpetually working 90 hours a week,” Habib says.
“It resulted in a deep amount of data I was able to provide to the research team, so it’s that continuous loop of product and research. And then our go-to-market teams who are literally on site with the customer saying, ‘This works great. This works shit. We need to improve this,’ right? ‘We need the model to be able to do X, Y, Z.’ And that is what makes us a product company, not a lab.”