Free with a sign-in.
The best 14 min of 45
Writer's May Habib: Building The Enterprise AI That Anthropic And OpenAI Sell — Then ‘Leave’ Behind
May Habib argues enterprise AI scales when companies pair capable models with trusted workflows and measurable results.
Upstarts · Alex Konrad
- May Habib is cofounder and CEO of Writer.
The best minutes
- 40:15AI buyers should be held accountable for results4 min
- 35:09Writer's moat grows as enterprise AI gets more capable5 min
- 22:06May Habib describes bias in her VC fundraising5 min
Also worth watching
- 30:20Writer kept training models because enterprise adoption lags3 min
- 18:11Writer uses review policies to scale personal outreach3 min
- 9:16Internal politics determine how far enterprise AI projects go3 min
- 11:49Writer says workflow scaffolding cuts token use by 30–50%3 min
- 13:59A pharma briefing can link Outlook, CRM and spreadsheets3 min
Skip the opening setup and sponsor reads, the Cordoba backstory, and the closing exchange on hiring and retention.
Summary · 4 min read
May Habib is cofounder and CEO of Writer, an enterprise AI company valued at $1.9 billion, according to the YouTube description. She previously built Qordoba, a machine-translation company, before pivoting it toward Writer in 2020.
Habib says Writer builds AI agents for nontechnical business users, with brand and compliance controls and workflows that connect to companies’ systems. She argues that reaching production is not the same as scaling: companies need organizational change, employee adoption and measurable returns. Habib says Writer’s customer feedback loop and preference controls distinguish it from general-purpose AI labs, even as those models improve. She urges executives to hold AI spending accountable to results and says enterprise trust in autonomous agents has yet to catch up with coding.
Minute by minute
- 1:30Writer targets nontechnical business users
Habib describes Writer as an enterprise agent platform designed for business users who can think systematically but may not be technical. She says its built-in brand and compliance controls support shared workflows, and that the company can demonstrate agentic capabilities in 30 or 60 days; scaling the resulting value takes longer because it involves changing roles and organizational design.
- 6:29Qordoba's language work led to Writer
Habib traces Writer to Qordoba, a machine-translation company, saying she and cofounder Waseem had worked on language, NLP and machine learning before the current wave of LLMs. She says their vision progressed from AI-assisted writing to AI writing, while early transformer models handled tasks such as correcting language and writing haiku.
- 9:35Enterprise adoption depends on politics
Habib says executive reactions to AI depend partly on internal politics: sunk costs in home-built tools can slow decisions, while collaborative teams can set more ambitious plans. She describes customer councils as a way to share lessons and says Writer’s scaffolding and shared workflows use 30, 40 or 50% fewer tokens.
- 14:10Writer differentiates through workflows
Habib describes three market phases: resistance to AI writing before ChatGPT, pressure to adopt after its launch, and a later challenge as agentic products began to sound alike. She says Writer’s differentiation is the work underneath the chat interface, such as centrally building consistent, auditable briefings that connect to different systems.
- 16:55Brand controls make AI usable
Habib says brand voice, style and compliance controls helped Writer win in industries such as consumer goods, pharma and finance, where generic AI-generated messages could erode trust. She says administrators can require email review and prohibit automated one-to-one messages; in her own example, an agent drafted emails for 14 or 15 meeting attendees, and she reviewed them before sending.
- 21:16Habib recounts the pivot and VC bias
Habib says the shift from Qordoba to Writer was a recap and restructuring that effectively restarted the business, while she and Waseem kept the legal entity because their visas were tied to it. She recalls Vinod Khosla falling asleep in a meeting and a Sequoia rejection calling it inspiring that she pitched with two small children. Habib says women founders can face added skepticism about how hard they will work.
- 27:29Habib prioritizes ROI over hype
Habib says the key was to stay in the game and focus on customer outcomes, even when ROI is less exciting to discuss than AI adoption. She says customers use Writer to reduce costs such as agency, Salesforce and Adobe bills, and estimates that 10 to 15% of sales and marketing teams may eventually have AI-related titles.
- 30:31Writer keeps training its own models
Habib says Writer started training models because third-party options were not available when it began in 2020, and kept doing so because it did not see the labs’ capital advantage as necessarily creating a capability gap. She says Palmyra powers 90%+ of Writer’s LLM calls, while customers can also connect outside or in-house models. Writer is not a lab, she says, because research, product and go-to-market teams work closely with customers.
- 33:35Customer adoption is part of Writer's moat
Habib says Writer’s moat is its ability to keep pace with model capabilities while bringing customers along. She argues that better models increase enterprises’ need for a preference layer covering brand, compliance and workflows, which Writer can productize across customers and teams.
- 37:38Habib rejects services-led deployment
Habib says enterprise customers want to build internal capability, not fill their offices with consultants, and argues that product teams can use customer feedback to improve implementation. She says Writer helps people onboard conversationally and gives teams shared account context, including agent activity that may not appear in the CRM.
- 40:15Habib calls for AI spend accountability
Habib says companies should hold people accountable for AI purchases without results, citing $3 million in overage over two weeks for an Excel plug-in and calling its ROI zero. She says enterprise adoption depends on auditable workflows with human oversight and trust, which have not yet reached the level seen in coding. Habib says the larger opportunity is still ahead and retaining top talent is a challenge.
Published Jun 25, 2026.
Transcript · 36 min read
Upstarts, hosted by Alex Konrad, with May Habib.
Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.
No one wants to talk about ROI, right? All the cool boys aren't talking about ROI. Why is me talking about ROI? Customer cares about ROI, right? Are they paying for software that does something? There are lots of cutting-edge AI tools helping engineers write faster code. But at the world's biggest companies, other roles from marketing to sales and even IT can benefit from on-brand AI output that doesn't break the bank. Writer is a startup unicorn that trained its own large language models and spent a decade on software to make that non-technical work flow faster. Our guest today is Ryder's co-founder and CEO, May Habib.
May pivoted her startup Qordoba in 2020 to leverage that new wave of AI tools. She now works with some of the world's largest businesses from Accenture to Hilton. It is really fun to be a builder in the enterprise right now because I don't think anyone has had this level of intimacy with the customer ever. Today on the podcast, we're going to talk about token maxing and solving for corporate AI sticker shock, what companies need to get value out of LLMs today, and the challenge of running a woman-led startup in AI's male-centric Wild West. I'm Alex Konrad, founder and editor of Upstarts Media, and this is the Upstarts Podcast, our weekly show where we talk to emerging startup founders who are punching above their weight to take on the status quo. May, welcome to the podcast. Thanks for having me, Alex.
This episode is brought to you by Rippling AI, the only AI built to give you full visibility into your startup and the ability to take action across every department. So Mae, I know what Ryder is, but for anyone who isn't familiar, give me the sort of elevator version of what makes Ryder exciting. Yeah, we are an agentic platform for the enterprise. And yes, the models are so infinitely capable today, but enterprises need built-in brand, built-in compliance. They need to scale workflows and share workflows, and Writer makes that possible. And so Writer is not necessarily intended for the engineers that we hear about using a lot of tokens to code new apps or anything, but maybe more other roles within a company. Is that right?
Read the whole transcript
It is primarily for business users. And so our big North Star from a product perspective is a normal person needs to be able to use this. And so if they're building an app and need a database or are doing a session and have shared deliverables, we're taking care of all of the infrastructure behind the scenes for them. In reality, you get the emergence of a builder class within it. team that is adopting AI. And so the kind of persona behind the business persona is a power user, but that is a non-technical person. It's a non-technical person who can think in systematic ways. And we really adopt that to kind of build the workflows that everybody else benefits from.
Got it. Now, you get to work with some of the largest companies in the world, like a Clorox or Marriott. When these companies are coming to Rider for help, is there a business problem or a productivity gap that they are often trying to address? So, so much of what the inbound is about is, hey, I hear my peers are using you. Like, how do you actually fit into my landscape? All right. And the reality is when you serve the Global 2000, they're already using all your competitors. And that actually is excellent for us because they have now understood what they can or cannot get done with a co-pilot, a co-work, a chat, et cetera.
And so when we come in, we get to give them a proactive point of view on what the future of the front office could look like. So if it is wealth and asset management, that's a viewpoint. If it is pharma, that is a viewpoint. But essentially what it comes down to is sales and marketing working as one team. And so much of the shared workflows that we build for customers are about identifying those core go-to-market processes for them. It means a radical rethinking of their org design. And that really is honestly the biggest blocker to scale. We can get in and within 30 days, 60 days, show incredible agentic capabilities that is running on their workflows, connected to their systems, that the power users are very excited about.
But production and scale, we have learned, are two very different milestones in a company. And that initial production-grade workflow gets everybody excited. The number of times our stuff has been presented to a board, like hundreds of times. But then how long it takes for you to actually get the value, right? The 80 mil of savings because you've turned off all these agencies or the 100 mil of savings because you have rewired an organization. That takes time because it's about people's jobs and livelihoods and roles and responsibilities and leadership paths. And all of that is, unlike the coding market, painstakingly slow in the enterprise.
You just packed in a lot, but I think there were two pieces there that immediately popped for me. One was this idea. that customers have kind of tried everything. And I'm curious, what is the driver there? Is it experimentation that they just have FOMO and they're like, hey, we should be using these tools, but we're not sure which one? Or then to the point of the cost savings you mentioned sort of at the end of your answer, are they feeling pressure to save money or to find efficiencies in some specific way? Pressure is an understatement. If you are a Fortune 500 executive, like bold-faced names Alex are meeting on the weekend about AI.
Pressure is an understatement. They feel an incredible amount of stress actually trying to get the rewiring of their organization done, of getting agentic and AI use cases that are worth telling the street about, right? And that really is the North Star for a lot of these folks is, have I actually been able to get value? And from the meetings I had this morning that I told you about to literally just about every executive conversation, it's, you know, we've got adoption. I forced people to use this stuff, but I have no results to show for it. That's a huge gap that we step into because despite, you know, the trillions of dollars of value that has been created, right, you've got anthropic and OpenAI salespeople who walk in like heroes, get a contract and leave to literally never be seen again.
And it leaves so much to be done in these companies for startups like us. We first met because you had made the Cloud 100 Rising Stars list. And I co-created the Cloud 100 franchise for Forbes a decade ago. And so you were working on a company called Qordoba. Yes, machine translation. Yeah, using machine learning and sort of the pre-LLM AI tools that were exciting a decade ago. What was the through line of sort of the idea and the mission of Qordoba with those customers through to sort of what you're doing now with Rider? We've been working at the intersection of NLP, language, machine learning, the enterprise for, you know, 12, 13 years now, Waseem and I.
And, you know, through Qordoba is how we started to build transformers. Now, you know, those first transformers were writing haiku and correcting language. That was really kind of their highest use. And so for us it was, well, there's probably a very big business to be built correcting all of the English for brand compliance and ease of translation and the translation market itself. And so when we started Rider, you know, our vision statement was we're going to go from AI. assisted writing to AI writing to who knows what's beyond. And I think it's the who knows what's beyond that has really surprised us, just, you know, how far we've all been able to get with the brute scaling of these models.
And so, you know, I think we're very good at the enterprise because we have been doing it across two companies. And, you know, the core of our company are people who have been working together for a very long time. So it has been very fun to feel like, you know, I've been doing, in a lot of ways, working on the same mission for such a long period of time, because you do start to feel that mastery. I think the thing that has not changed is, you know, working late into the night and simply not being able to use software without filing Dura ticket reports and product design requests, right? Like I cannot use the product if I have not also figured out all the ways to correct it.
So something's never changed in the development side. From the capability side, though, you know, there was this wave of companies like Qordoba. Grammarly starts around this time. There are a bunch of these sort of pre-LLM tools. What do you think you guys were able to get the most right or deliver the most? And where did the technology not just be ready yet, you know, so that LLMs maybe really changed the game? Like if we think about sort of that was 1.0, maybe now you're working on some version of 2.0. Well, I think it's 4.0. I think, you know, writer versus Grammarly in that first year, we'd always been marketing the fact that Grammarly is rules-based. We're this, like, new thing-based, right? Now, LLM wasn't there, but Transformer was there. And we described it as encoder-decoders for folks who really got under the hood.
We've always sold to people who just wanted the thing to work, right? Even today, Palmyra powers 90% plus of our LLM calls, our own frontier models.
“And so much of the home-built stuff is accumulating tech debt as soon as it is launched.”
hope for the best. Got it. And that's more the saving their mood, keeping them from despairing
about the future. Yeah. And it's not just us. I mean, we've got a real set of executive programs that pairs everybody together. We have a CIO council, we have a CISO council, we have a CMO council,
“the way that we scaffold and set preferences means that you're using 30, 40, 50% fewer tokens each time.”
“If I am building a, you know, Monday morning briefing for my pharma salespeople, they better all look the same. They better be all audited the same, right?”
Was there a use case or a skill that you guys were really good at early on that was sort of the killer door opener with some of these really big companies? Yeah, I mean, especially in CPG, pharma, FINS. If you sell REITs or if you sell financial products, right, or financial advice, you simply cannot email somebody with LLM voice. Your client knows you. They'll be like, Alex, get out of my inbox. What's the shit you're sending me? I don't care. I don't trust it, right? And so from literally the very beginning, our ability to build in brand, voice, style, compliance has meant the world to these customers because it means I can plug into Alex's inbox.
And now I know when he's on the road, it's like tons of typos. It's two spaces after a period. Well, no, you're a crimson guy, so I know it's not two spaces after a period. But you know what I mean. It's you, right? And that has always been a huge calling card for us because customers want to scale AI, but they got to do it while scaling brand, while scaling authenticity, while reducing brand risk. Like, nobody is going to take the trade of productivity, right, versus brand dilution or risk. And in the enterprise, especially in the regulated industries that we serve, writer is for the stuff you have to get right, that you cannot afford to get wrong.
And that's always been a huge differentiator.
“I'm literally already responding to the responses faster than me even drafting a single email.”
And it works out of the box. And the thing that they'd spent two years building doesn't even come close to it. What was a moment that you were glad you fought through and executed on that you would not want to do all over again? Yeah. A lot of amazing moments. I would love to relive. The thing I would never want to relive is really the pivot from Qordoba to Ryder because it was not nearly as well understood as it is now what this technology could do.
And if you could have been in those rooms where we were trying to explain encoder decoders to people and why they would be so massive, you know, we would all be pretty amused. Was it really just a clear pivot or was it more of a— Yeah, it was a recap restructuring. I mean, we would have started a brand new cap table, but both mine and Waseem's visas were tied to the old cap table. So, you know, it was basically a complete restart except for the legal entity. But then the mission and the product, it does feel like there are traces, like there is a through line from Qordoba through Rider. Absolutely, yeah.
“In the series A, the Sequoia rejection included a line about how inspiring it was that I had two little kids at home while I was pitching them.”
That's because Rippling AI is built on your live workforce data that gives you full visibility into your startup and the ability to take action across every department. Say you're planning your next few hires. Just ask Rippling AI, what would it cost to add two more engineers this quarter? and you'll instantly see a breakdown of comp ranges by level and location. So you can make the smartest hiring decision based on where your team has gaps and where the talent is. But it doesn't stop there.
Rippling AI can draft the job wrecks, route them for approval and kick off the hiring workflow. All you have to do is tap confirm and then get back to building. Don't settle for AI that's all talk. Head to rippling.ai slash upstarts and get AI that turns insights into action. that's r-i-p-p-l-i-n-g dot a-i slash upstarts sign up for exclusive access today it sounds like you feel pretty good about where things are right now, when you think about sort of the most important decision that you guys strategically made to get to this point where maybe you do feel very proud of the product, what comes to mind first? Survive, don't sell, don't quit. This is a hard space. And I think it rewards those teams that are physiologically set up to do hard things for a long period of time.
It's not one decision other than stay in the game. The hype will pass, and that competitor that, like, gets all the headlines, they haven't built the right thing for the right people and truly understand the customer, like, may have their moment in the sun, but will not survive. And for us, 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. Wives may talk about ROI. Customer care is about ROI, right? Are they paying for software that does something? And there isn't a CIO I talk to, CMO I talk to, who doesn't want leverage over their cost structure.
I know I'm a weapon for them, right? We know that we are being, like, used. I'm happy to be used, right? What are they using us for? Reduce their workday bill, Salesforce bill, Adobe bill. agency build, you are lucky. And we know we're lucky to find somebody who wants to use us for that. Adobe's a customer, Workday's a customer, Salesforce is a customer. And so, you know, internally as well, right, everybody is trying to understand, right, how to get this kind of leverage and get the form factors right. And when we look at, let's take our champions program. So we've got hundreds of people in the champions program. These are folks who, depending on the customer, right, usually are reaching 500 mil to a bill a month of tokens, business people, right?
Now, you can be invited to the Champions Program if you're doing really important things. It doesn't have to just be about token maxing, et cetera. We're not about like, you know, the vanity of the token. But most of those people have AI in their title now. They have invented new roles for themselves. And I think, I agree with Aaron. Aaron Levy wrote about this recently. I don't know if you gave a percentage, but I personally think 10 to 15 percent 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 agency, executive team who has two people in the business dedicated to AI, right, 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 a lot of cases rehiring, right, for the kind of capabilities and skill
“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.”
When you think about what gives Rider a moat or sort of real differentiation, is it how the technology is packaged together? Is it more a user experience for that end customer? What is the sort of framework you use today? The idea of a moat is just incredibly different today than it was in the era of SaaS.
Today, we think about our moat as, yes, all of the above, but the ability to do that at pace while driving customer adoption. Because it's actually our North Star for the past five and a half years has been productizing to the capabilities of what the models can do. And if you leave behind the customers that have joined you at each era of capability, then they don't believe that you are there for them anymore. And the kind of organization you need where, you know, we're almost 500 people, where every single role is in a constant state of up-leveling to meet that most ambitious customer with the most ambitious product, but also empathetic enough to bring forward the customer who hasn't yet passed a security review to turn on any connectors or doesn't have any power users yet doing really advanced stuff or needs a lot more hand-holding to kind of rewire a workflow end-to-end.
That, I think, is our moat because we can go into an organization and no org is a monolith, right? Nobody is all in or all power users or, you know, all wired up. It's within the same customer, multiple generations of capabilities, ambition, et cetera. And I think taking that and making sure our product reflects that. and our EPD team and research team are empathetic to that, I would consider them out. Another kind of interesting framework that I've seen is,
“And in so many ways, whether it's built-in brand, built-in compliance, or ability to structure these workflows, scaffold them, et cetera, is essentially a preferences stack.”
“The answer to you never, no one got fired for buying IBM should be people fucking getting fired for buying IBM.”
And I am curious with you whether you feel like your upstart moment would have been more in the early days, you know, getting those first customers to sign up or today when, you know, these companies are getting trillion dollar market caps and there is a lot of noise in the market.
I see it in the future. I mean, I don't know if that's like a cop-out answer, but it does feel harder now, frankly. And I can't wait for these companies to go public and just like, you know, have everybody really understand how much space there is and how much room there is for everybody else. anyone who's building in this space right now, it's just the biggest challenge, right? Hiring, retaining employees. And I think, you know, when it comes to folks who their goal is to be a founder, their goal is to build, they're not going anywhere, right? But it's a gold rush for everybody else. And, you know, really holding on to that top talent is it's a challenge for every builder right now.
