Published November 19, 2025 | Amy Bray | People Operations & HR Technology
🤔
The Problem I Kept Seeing (Over and Over...)
The Pattern: After years at The Mom Project and Upwork, I kept seeing the same painful scenario: people placed through marketplaces or EOR programs literally didn't know who employed them or who to ask about benefits, PTO, or payroll. The marketplace logo was in their inbox, the client name was on their badge, the EOR was on their paystub... and none of it felt clear.
You know how sometimes you see the same problem at different companies and think, "Surely SOMEONE has solved this by now"?
That was me with contingent worker confusion. Different business models, same core mess: people didn't know who actually employed them.
From the worker's perspective, it makes sense:
- Job posting was on the marketplace website
- Interview felt like it was with the client
- Onboarding emails came from the marketplace or EOR
- They signed documents from an EOR/staffing partner they'd never think about again
- Then they work every day at the client site or in the client's Slack
So when they need help, they contact the brand they remember. Usually the marketplace.
The conversation (repeated ~100 times):
Worker: "Hey, what are my benefits?"
Me (marketplace/talent rep): "Who's on your paystub?"
Worker: "Um... Employee Stand?"
Me: "They're your employer. Contact them."
Worker: "But I work at UnBungle and applied through YOUR website?"
Me: [internal screaming]
The whole W-2 contingent setup IS confusing:
- You interview with Company A (client)
- You sign paperwork with Company B (EOR)
- You apply on Company C's website (marketplace)
- You work at Company A every day
- Company B pays you and handles benefits
- Company C runs the program and answers emails
No wonder everyone was confused.
And 1099 contractors had their own version:
- Some thought they should get benefits from the client or marketplace
- Others asked client managers how to invoice or handle taxes
- A few weren't even sure if they were W-2 or 1099 – they just knew they got paid
💡
The Solution: Meet People Where They Are
The Breakthrough: Instead of making people learn employment law terminology, I built a chatbot that asks observable questions like "Do you receive a paycheck or send invoices?" and "Look at your paystub – is the company name different from where you work?" Then it explains what that means in plain language and routes them correctly.
So I built Aunikah – a chatbot that meets people where they are. You don't need to know the terminology. You just tell it what you want, and it figures out the rest.
📝 Smart Classification Reference
The "I'm not sure" flow asks about paychecks vs invoices, then checks if the paystub company differs from where they work. Instant clarification without legal jargon.
🎯 Plain Language Routing
Once classification determined, explains: "This is your employer of record. They handle pay, benefits, and PTO" vs "This is your client. They handle your day-to-day work."
👥 Manager Mode
Managers pick a team member's name, and Aunikah provides person-specific info: "Sarah Chen is W-2 through Employee Stand. Benefits include health/dental/vision, 401k with 4% match, 15 days PTO..."
🔍 Context-Aware Answers
Different responses for W-2 vs 1099, client-employed vs EOR-employed. The same question gets the right answer based on employment status.
🎨
The Design Decisions That Mattered
The "I'm Not Sure" Option Changed Everything
Instead of forcing people to choose between "W-2 Contingent" and "1099 Contractor" (terms half don't know), Aunikah asks:
"How do you get paid?"
💰 I receive a paycheck/paystub
📄 I send invoices
Then: "Look at your paystub. Is the company name different from where you work?"
BOOM. Instant clarification. No legal lecture, no jargon quiz – just observable reality.
The Paystub Trick is Magic
When people are confused about classification, asking them to check their paystub is magic. It's physical proof they can hold. No ambiguity. That company name is the answer.
For marketplaces, it's also a gentle way to say "I know you interacted with us, but legally, THIS is who employs you" without making them feel silly.
Language Matters So Much
Terminology Challenge: I kept calling everyone "employees" in the first draft. Then I realized:
- 1099 contractors AREN'T employees (that's literally the point)
- W-2 contingent workers ARE employees... but of the EOR, not the client
- "Worker" feels cold and transactional
- "Talent" is overloaded in marketplace branding
Solution: I use "team member" when talking about humans on a team, and switch to "employee," "contractor," or "W-2 contingent" only in legal/benefits territory when it actually matters.
For people who already feel "less than" because they're contingent, language is part of the experience. I didn't want the help tool to also make them feel othered.
✨
Real User Feedback (My Former Team Saved Me)
I sent the first version to my former team (shoutout to Abby, Mary, Claudia). They're the best! They actually tested it and gave honest feedback.
The feedback that changed everything: "The Quick Setup menu is confusing. I didn't know where to click."
I had built this fancy collapsible menu at the top where you could set your classification. Very efficient. Very... hidden.
Meanwhile, the whole point is to help people who are ALREADY confused. Making them hunt for a setup menu is the opposite of that.
The Fix: Redesigned the whole onboarding to be conversational. Now it walks you through setup step-by-step in the chat itself. It feels like "Let's figure this out together," not "Fill out this form before I'll talk to you."
This is why you test with real people. I was so deep in building mode I couldn't see the obvious UX issue.
👨💼
The Manager Side (aka The Database Trick)
One of my favorite features is manager mode. Managers often don't remember which contingent worker is W-2 through which EOR vs 1099. Why would they? It's not their job to track that.
But it IS their job to answer: "Can I approve this PTO?" or "Where should I send them for benefits?" Without context, they guess or forward people around.
The Solution: In Aunikah, managers just pick the team member's name from a list. The system "knows" (from the database) whether that person is W-2 or 1099, which EOR employs them, and which client they're at.
Then when the manager asks "What are Sarah's benefits?" Aunikah responds with Sarah-specific info:
"Sarah Chen is a W-2 Contingent Worker employed by Employee Stand and working at UnBungle. Benefits include health/dental/vision through Employee Stand's plan, 401(k) with 4% match, and 15 days of PTO..." but reminds to route to the EOR for individual answers.
No more "uh, which one is she again?" The system handles it. The manager just has to care and ask.
In production, this would be an API call to your HRIS. In the demo, it's a JavaScript object. But the concept is the same: make it easy for managers to support their teams without memorizing everyone's employment status.
🎓
What I Learned (The Messy Bits)
1. Scope creep is real and I almost went global
At one point: "Ooh, I should add international classifications! W-8BEN forms! VAT!"
Then I caught myself. This is a DEMO. US-only is fine. The point is to show I understand the problem, not to solve every global employment scenario.
Sometimes done is better than perfect. (Still learning this one.)
2. The paystub trick is major
When people are confused, asking them to look at their paystub is magic. Physical proof they can hold. No ambiguity. That company name IS the answer.
For marketplaces, it's also a gentle way to say "I know you interacted with us, but legally, THIS is who employs you" without making anyone feel silly.
3. Personalization makes it feel real
When Aunikah says "Benefits for Sarah Chen are handled by Employee Stand" instead of "Benefits are handled by their EOR," it feels custom-built. Like the system actually knows them.
That's what makes demos compelling – it feels like it could actually work in production.
⚙️
Tech Stack (I Kept It Simple)
This is a single HTML file. Yes, really.
Frontend
HTML/CSS/JavaScript (vanilla, no frameworks)
Data
JavaScript objects (in production: API calls to HRIS and vendor systems)
AI Assist
Claude Sonnet 4 helped design and iterate (the chatbot itself uses pattern matching, not live LLM calls)
I built this in one weekend after iterating on the concept with Claude. Could I have used React? Sure. Did I need to? Nope.
Sometimes the simplest solution is the best one.
📊
The Stats (What This Could Do)
These are directional, based on patterns I've seen over the years – not a formal study, but a realistic picture of impact if you wire this into actual systems.
40-60%
Reduction in HR support tickets about "who do I ask?"
100%
Benefit enrollment confusion routed correctly
2 min
Average time to get clear, personalized answer
🚀
What's Next
Honestly? I'm proud of this thing. It solves a real problem I saw across 13 years in People Ops and marketplace/EOR programs.
If I were to take it to production, here's what would need to happen:
- Integration with HRIS systems (Workday, BambooHR, etc.)
- Real EOR/vendor data feeds
- SSO authentication and role-aware views (team member vs manager vs HR)
- Actual AI/NLP for better intent detection (currently pattern matching)
- Company-specific customization (benefits plans, portal links, escalation paths)
- Analytics dashboard for HR teams (what people are confused about, where they're stuck)
But as a portfolio piece? As proof that I understand contingent workforce operations and can build solutions instead of just writing docs about them?
It does exactly what it needs to do.
Try It Yourself
The demo is live. Try the "I'm not sure" flow – it's my favorite part, because it mirrors so many real conversations I've had with people who just wanted a straight answer.
Launch Demo →
📧 Email: [email protected]
💼 LinkedIn: linkedin.com/in/amybrayphr
🖼️ Project Page: AuniConnect Platform
Note: All company names, EORs, and benefits in the demo are fictional examples. The confusion, emotional labor, and operational pain points are very real.