A week in AI is like a year in other industries. I hope these issues become your weekly source of AI information, inspiration, and ideas. If we haven’t met before, I’m Amanda Smith. I write about AI and the fascinating folks who are building in this brave new world.
Our latest founder has spent her career trying to give people a fair shot, first as a teacher and now in the benefits system. Her newest bet is an AI assistant that helps families make sense of and access public benefits.
This week in AI:
Anthropic files for a potentially record-breaking IPO
Putting the AI industry’s valuation frenzy to a public-market test.
Meta plans a $95B stock offering to fund AI
The giant is raising substantial capital to cover its AI efforts and data centers.
Mark Cuban on not treating AI as magic
He stated AI is “not smart,” it “doesn’t know the consequences of its actions” and “doesn’t have inherent ethics.”
🤝 This edition is kindly brought to you by Snowflake
Founders and tech leaders: put your data to work with AI. The Snowflake AI Data Cloud lets your team build data agents, run LLMs and analyze data on one fully managed platform that runs across clouds. Start a 30-day trial with $400 in free credits through AWS Marketplace.
Company background: MyFriendBen
Founded: 2025
Team size: 10
Funding to date: Spun out of Gary Community Ventures as an independent 501(c)(3) and secured $4M+ in philanthropic commitments from funders including Kauffman Foundation, the Annie E. Casey Foundation, CPAL, and Next Ladder.
Revenue model: The model is philanthropy plus a growing base of earned revenue from state agencies, 211s, and partner organizations that pay for co-branded screeners, integrations, and navigator tooling.
Laura Glaab: MyFriendBen
Laura Glaab started her career as a teacher, where many of her students came to school hungry, exhausted or from homeless shelters. She went on to lead a school and scale Amira Learning, an AI reading tutor, to statewide partnerships. There she learned that AI has to earn people's trust, prove itself with research and scale without losing the relationships at its center.
Now, as CEO of MyFriendBen, she's bringing that playbook to America's safety net. MyFriendBen's screener tells a family which of more than 40 programs they likely qualify for in about six minutes. It has helped more than 175,000 households in nine states.
The AI component
Benji started the way most things at MyFriendBen do: Users asked for it. Families kept asking for help with applications and knowing where to start.
"The hardest part of the safety net isn't learning that you're eligible. It's everything that comes after: The applications, documents, deadlines and renewals," Glaab said.
About a third of users screen at night, when navigators aren't available. Then H.R. 1 brought new work requirements and more frequent eligibility checks, so millions of families will have to prove again and again that they still qualify.
"There will never be enough caseworkers to meet that moment one family at a time. AI can," she said.
Benji is available 24/7, speaks plain language in the family's own language and knows each household's screener results. That frees caseworkers and 211 navigators for the human work only they can do.
The build process
Benji took about six months from prototype to launch. A small engineering team built it, led by Glaab’s past colleague at Amira. They started in spring 2026 and user-tested it hundreds of times over the summer, including two days in Dallas with First Step Staffing, where people facing homelessness and job barriers used it live. It shipped to every live state in September 2026.
The key design decision was what Benji isn't allowed to do. It doesn't decide eligibility. That comes from rules the team researches program-by-program and state-by-state. Benji sits on top of those results, explains them in plain language and helps families decide what to do first.
The team didn't train or fine-tune a model. They built context and guardrails instead: Curated content for each program, a persona modeled on good human navigators, instructions to point people to human help rather than guess, and a setup where the AI provider stores nothing.
The tooling is deliberately ordinary: GitHub and Linear, automated weekly quality reports on every conversation and a weekly human review of transcripts. The team also uses AI for code review and program research, which Glaab says is the only way a team their size keeps up across nine states.
"When a family is deciding whether to apply for food assistance, a wrong answer isn't a bug. It's a missed meal," she said.
Go-to-market
There was no big launch, on purpose. Benji went out with a regular Thursday deploy to every family already using MyFriendBen in its live states.
"We wanted real conversations before we wanted headlines," Glaab said.
In the first 24 hours, Benji had 16 conversations across five states about shut-off notices, evictions and stopped disability payments. The team shipped fixes within hours.
The wider rollout runs through partners. The week Benji went live, Glaab trained a few dozen navigators, and their one question was whether they could get it now. The team made a walkthrough video, updated its privacy policy and partner talking points, and is introducing Benji one partner at a time, from 211s to school districts to workforce programs.
Benji serves the same families as the screener. It also serves navigators and caseworkers at partner organizations, who use it to answer questions faster.
Scaling and growth
"We don't buy our users; we earn them through trust," Glaab said.
Hundreds of community partners, including 211s, school districts, housing providers and clinics, embed the screener in their own workflows, and Benji now comes with every deployment. That looks like co-branded screeners with 211 Metro Chicago and Denver Public Schools, an integration at the end of the Denver Preschool Program application, and events like the Dallas days, where the team screened more than 200 people in two days.
Each new state is its own launch. Kansas and Missouri went live this fall with United Way of Greater Kansas City, and New Jersey is launching with NJ 2-1-1.
Marketing backs that up.
This includes the Friends of Ben newsletter, LinkedIn and Instagram, an Access Initiative webinar series for nonprofits and 211s, co-hosted partner webinars and conferences. Organic search drives most of the traffic in North Carolina.
Their biggest lever is evidence. The Urban Institute found that 86% of users applied for a new benefit, and 85% of those were approved.
"Partners and funders in a sector that has learned to fear AI don't want a campaign; they want proof that it works and that it's safe," she said.
Advice for founders
Trust is everything. Ground your AI feature or product in eligibility rules and verified information, rather than what a model happens to know.
Overcoming fear. Many nonprofits worry that AI will replace caseworkers or put vulnerable people at risk. The best AI gives you greater efficiency and time.
The rules never stop moving. Use AI to keep up with regulatory movements and programs.
Nonprofits get great technology last, and that's the opportunity. The families with the least time and the most at stake deserve the best tools, not leftovers. AI is our chance to flip that.
How's the depth of today's edition?
If one of these stories stuck with you, I’d love to hear which one.
Speak soon,
Amanda
P.S. If you want to get a founder feature about your own story, reply to this email. If you’d like to reach our newsletter audience (founders, creators, and marketers), click the button below.
If you’re new here, I’m over the moon you’ve joined us! To help me craft content that’s actually useful (and not just noise in your inbox), I’d love it if you took 1 minute to answer this quick survey below. Your insights help shape everything I write.
✨ Insane Media is more than one voice
Dive into our other newsletters - where psychology meets the creator economy, e-commerce marketing, and Human resources.
