What Women Are Building With AI: 5 Real Projects

Ellie at Willa, Editorial lead, Willa · Published November 3, 2026 · Updated November 3, 2026

If you want to know what women are building with AI right now, the honest answer is: shipped products, not weekend experiments. At a recent Willa event, five members demoed what they had built, and the range was striking. A video networking app that costs 20 cents an hour to run. A bootstrapped healthcare platform built by two non-technical founders. A personal career knowledge base. A digital magazine produced in under 24 hours. An iPhone app that is live in the App Store today.

None of these women leads an engineering team. Most paid for little more than a Claude subscription and some hosting. Here is what each of them built, how they did it, and the patterns worth copying.

Can you really build a video app with AI?

Kelly Anlas, a founder and fractional CTO, got tired of watching a friend fight with breakout rooms in Google Meet. The rooms malfunctioned, grouped the same people together, and ignored everything attendees said about who they wanted to meet.

So she built Tether.video, a networking tool where hosts create an event, define guest profile fields (role, LinkedIn, industry, an "ask" and an "offer"), and let the software match people into breakout rooms using a points system. Someone looking for help with tech gets paired with someone offering it. Hover over any participant and their profile card appears, LinkedIn link included, so nobody has to paste links into the chat.

Her stack: open-source LiveKit for the video layer, a Fly.io server to host it, and Claude Code to build it. Total build time was roughly 8 hours. A one-hour session with about 10 people cost her around 20 cents to run.

Her theme for the whole talk: "We can just do things now."

Kelly's workflow is worth stealing even if you never touch video software. She starts in a regular Claude chat or Cowork session to think through the idea, the architecture, and the costs. Then she asks Claude to summarize all of that into instructions she hands to Claude Code. "If you have a really concrete idea of what you want the outcome of this to look like, the AI agents can do a very good job of providing it for you," she told the group. The unspoken step one is the thinking.

Can non-technical founders build a startup with AI?

Sophie, co-founder and CEO of Noor, is building a care discovery platform where women find doctors and practitioners recommended by women like them, matched on symptoms, life stage, age, and neighborhood. The beta is live in Los Angeles for in-person care and across California for virtual care.

Neither founder is technical beyond HTML and CSS. They are bootstrapped, so every tool choice is a cost decision. Their stack: Supabase for the database, Vercel for hosting, Brevo for transactional email, Google Forms for anonymous feedback, Figma plus Tailwind for design, GitHub for code, Linear for project management, and 1Password for keys, with Claude Code doing the building.

Three of her decisions stand out:

1. They spent about a month on specs before writing code. Sophie and her co-founder used Claude Cowork to produce a product scope, an engineering doc, and a database document, and to work out which features would keep their healthcare data handling in a safe zone during beta. Only then did they open Claude Code.
2. They locked Claude out of sensitive data. Claude Code has access to their Supabase database but is blocked from seeing member information, so user data stays private even from their own AI tooling.
3. They curate practitioner profiles by hand. AI scraping kept pulling outdated bios and photos, so they find accurate sources themselves and use AI only to help input the data. Slower, Sophie said, but "it puts us in more a place of integrity."

Her hard-won design lesson: AI is much better at coding than at design. "Claude is really good at functionality and database management," she said, but the design and UX work is where you differentiate, because so many AI-built products look the same. She recommends sketching screens, collecting inspiration, and going back and forth with Figma rather than accepting the first layout an AI produces.

How do you organize your whole career with AI?

Jackie Pines wanted to leave her job for consulting, which meant she needed every project, metric, and win from her career in one organized place. She built what she calls a career brain inside a Claude Project.

The system has a few connected parts:

• A discovery skill she developed with Claude that interviews her. She journals about each career experience, and Claude pushes back: do you have a metric for this, are you telling the whole story?
• A brain file, the giant source-of-truth document holding every job, volunteer role, and engagement in her own words, plus a short experience index she can pull from for resumes or proposals.
• A brand file, a tighter distillation of her voice and positioning that her other skills draw from, including one that maps her experience into consulting offerings.
• A changelog that records every change to her core documents, split into content changes and administrative ones.
• A weekly scheduled task (she calls it BrainSync) that runs every Friday, reads unsynced changelog entries, writes content changes into the brain file, and sends her a summary. Anything unclear comes to her to classify, so she stays the final check.

Her project instructions give the AI a role as an executive coach, require it to refer to itself as "AI coach" rather than "I," and tell it to push back instead of agreeing with her.

For anyone who feels disorganized in Claude, her advice was simple: ask Claude to help design a structure for you, and ask it to summarize each chat so the next conversation can pick up the context. Even a basic log of "I did this at work and here is how it went" pays off the next time you need a resume, a raise, or a career change. (For a related approach, see our guide to building a brag document with AI.)

Can you publish a magazine with AI in 24 hours?

Nancy Shenker started her career in publishing, ran marketing at the company behind Comic-Con, wrote a book about AI in 2017 called Embrace the Machine, and is now, at 70, publishing digital magazines in under 24 hours.

Her Halloween issue, a comic-book-style magazine about the monsters in our lives (tech bros, difficult clients, the ones in our own heads), went from concept to published in less than a day. Her process: she keeps a running outline in a Google Doc, adds thoughts as they come, and has Perplexity pull the whole thing together from that doc. Because she spent decades as a writer, her prompts carry most of the editorial direction.

Two rules keep her output sharp. Every piece of content has to educate, entertain, or inspire someone to do something different, or she will not write it. And she begins every AI conversation the same way: "What is my objective?" Without that, she said, you get a spew of tactics instead of work that moves your business.

Her motivation is blunt. Gendered ageism is real, she argues, and she intends to stay relevant and financially solvent for decades to come. Thought leadership, produced fast with AI, is how she is doing it.

Can you ship an iPhone app with AI?

Janiré Castillo, a product designer with more than 20 years of experience, built Marco, an iPhone app for freelancers that captures why a project decision changed. What changed, why, and what it affects in price, timeline, or scope, logged in under 30 seconds, with a one-tap read-only share for clients. It is in the App Store now at version 1.1.3.

Her process looked more like disciplined product work than vibe coding, and that is the point:

• Every build started with a document. Competitive analysis (which found nobody was building for solo freelancers), feature audits, jobs-to-be-done, and the state of every screen. "They're how you know the AI built the right thing, and not just a thing," she said.
• 63 automated checks, each required to prove it can fail. "A test that can't fail looks like safety," she warned, and it is not.
• Real users over scores. She demoed Marco at a UX conference, collected feedback from other designers, went home, and shipped the improvements the next day. "AI can review, but real people found things that no check did."

She also set scope rules up front, like a rule that photo features wait until the app has 10 real users. The discipline keeps an AI-assisted build from sprawling.

What do these five builders have in common?

Five different products, one repeated playbook:

1. Thinking comes before prompting. Kelly brainstorms in Cowork before Claude Code sees anything. Sophie spent a month on specs. Janiré wrote documents for every build. The clearer the idea, the better the output.
2. The costs are small. Twenty cents a session, a $20 Vercel seat, a Claude subscription. The expensive part of building has become the clarity, not the code.
3. Humans stay the final check. Jackie approves every sync. Sophie curates data by hand. Janiré tests with real people. Nobody ships whatever the AI produces unreviewed.
4. Design and accuracy are the differentiators. AI handles functionality well, so the builders win on UX, data integrity, and voice.
5. A background in anything helps. Journalism, publishing, product management, tutoring. Every presenter turned prior experience into better instructions for the AI.

If you are deciding what to build first, start with a problem you personally have, write down exactly what the outcome should look like, and hand that clarity to the tools. For a step-by-step walkthrough, read how to build an app with Claude, start to finish.

Key takeaways

• Women in the Willa community are shipping real products with AI: a video networking tool, a healthcare platform, a career knowledge base, a digital magazine, and a live App Store app.
• Most builds used Claude Code plus inexpensive infrastructure, with running costs measured in cents or tens of dollars.
• Every successful build started with clear thinking: specs, documents, or a stated objective before any code.
• AI is strong at functionality and weak at design, so hands-on UX work is where solo builders stand out.
• Keep yourself in the loop as the final reviewer, whether that is approving syncs, curating data, or testing with real users.

Want to see builds like these live, ask questions, and demo your own work someday? Willa members attend every event for free, from show-and-tells to hands-on build sessions. Check out upcoming events and join Willa.

• women in AI
• Claude Code
• building with AI
• AI projects
• Willa events

Join Willa, the community where women learn AI together.

All posts