The Phantom Plan in Your Employees’ Pockets

A couple years back, I got a panicked call from a benefits director at a midsize tech company. She’d just realized that the “wellness app” they’d rolled out to 3,000 employees - the one with the virtual doctor visits, mental health coaching, and a shiny medication reminder tool - had never been mentioned in any plan document. Not one. It didn’t exist in their BenAdmin system. Their COBRA administrator had never heard of it. And when an employee who’d recently quit used the app to talk to a therapist, it hit her: we might be running an invisible health plan.

She’s far from alone. Walk through any benefits conference expo floor and you’ll see dozens of “employee health engagement” apps promising to slash claims and make your workforce healthier. They bundle telemedicine, chronic condition coaching, mental health support, symptom checkers, and plenty more. HR teams roll them out like free lunch programs, rarely stopping to ask one critical question: is this thing a group health plan under ERISA? Almost always, the answer is yes - and almost nobody’s prepared for what that means.

How a “Wellness App” Becomes a Medical Plan

ERISA defines an employee welfare benefit plan as any plan, fund, or program established or maintained by an employer that provides medical, surgical, or hospital care or benefits. The word “medical” does a lot of heavy lifting here. If the app performs anything resembling diagnosis, treatment, or prevention of disease - and your company pays for it - you’ve crossed the line. Vendor labels like “engagement platform” or “digital front door” don’t matter one bit.

Let’s break down common features that create a plan:

• A telemedicine button connecting an employee to a board-certified physician for a sinus infection. That’s medical care, no question.
• Health coaching for diabetes or hypertension that adjusts recommendations based on biometric data. That’s disease management.
• Text therapy with licensed mental health counselors. Still medical care.
• Medication adherence tools that pull pharmacy data and push condition-specific nudges. That’s squarely in the “mitigation of disease” bucket.
• Even an AI symptom checker that escalates to a nurse line can trigger plan status if the employer sponsors it.

I’ve lost count of how many times I’ve heard “but it’s just a wellness tool.” Unfortunately, ERISA looks at function, not branding. If the employer pays a per-employee-per-month fee for a bundle that includes medical services, a plan exists. And a plan needs a written document, an SPD, fiduciary oversight, a claims procedure, likely a Form 5500, and - here’s the kicker - COBRA continuation coverage. The exceptions for employer-paid medical care are narrow, like the de minimis rule for on-site first aid, and a telemedicine button fits none of them. In the real world, I’d estimate one in twenty employers has properly addressed all of this.

Why the EAP Label Doesn’t Help

When a vendor wants one of these apps to look harmless, it gets rebranded as an employee assistance program. The label buys less than most people think. An EAP qualifies as an excepted benefit only if it does not provide significant benefits in the nature of medical care, and only if it meets the other conditions in the rules, including no premiums or cost sharing. A telemedicine button, a diabetes coach reading biometrics, or a licensed therapist chat is significant medical care. Put any of those in the bundle and the exception fails, leaving you back at a group health plan with its document, SPD, claims procedure, and COBRA obligations.

The label was never much of a shield to begin with. The DOL’s long-standing advisory opinions treat counseling by licensed professionals as medical care that establishes a plan, so an EAP with real counseling still carries Title I obligations like a plan document and, past 100 participants, a Form 5500. When a vendor hands you a compliance memo saying otherwise, check whether it was written for the medical features actually in your bundle or for a generic EAP description.

The Tech Stack Blind Spot

Most benefits administration systems are beautifully designed to handle the usual suspects: medical, dental, vision, HSAs, FSAs. They weren’t built for an app that lives on someone’s phone and gets licensed by the wellness team. So what happens? The app gets its eligibility file through a manual spreadsheet upload or a one-off SFTP feed, totally disconnected from the core system. It’s an orphan, and that creates a mess.

Here’s what I’ve seen firsthand:

• Eligibility breaks constantly. An employee leaves the company, their medical coverage ends, but they can still log into the app for weeks - sometimes months - and enjoy a free telemedicine visit. You’ve just provided a medical benefit to someone no longer entitled to it. That’s a fiduciary problem.
• COBRA goes completely overlooked. Nobody has figured out the “app-only” COBRA premium because no one recognized the app as a separate plan. A former employee with a chronic condition keeps using the care features and later argues they should have been offered continuation coverage. The DOL doesn’t find that amusing.
• HIPAA gets messy. A self-insured employer can only receive PHI for plan administration purposes, but these apps often ship aggregated dashboards straight to HR - showing things like “your workforce has elevated blood pressure” or “engagement in the diabetes program is low.” That’s still PHI, often easily re-identifiable. Without a properly drafted business associate agreement and a plan document that restricts employer use, you’re playing with fire under both HIPAA and the ADA.

The Design Questions Nobody Asks Until It’s Too Late

When an engagement app evolves beyond simple nudges and starts routing members to specific providers, scheduling appointments, and managing transitions of care, it quietly becomes part of your care coordination engine. That triggers fiduciary considerations most benefits committees never chew on.

For example:

• Are you vetting the app vendor with the same rigor you’d use for a TPA? The DOL’s cybersecurity guidance, first issued for retirement plans in 2021 and expanded in September 2024 to cover health and welfare plans, expects fiduciaries to ask service providers for SOC 2 reports and penetration-testing summaries. Most app sales reps look at you blankly when you bring that up.
• If the app steers employees to a particular telehealth provider or preferred coach network, are you accidentally exercising discretion over plan assets? You might be taking on a fiduciary responsibility you didn’t sign up for.
• If those preferred coaches form a de facto “network,” do they comply with mental health parity requirements? You might have just created a non-quantitative treatment limitation without realizing it.

These issues never surface in the vendor’s ROI deck because their world is sales, not plan governance. But as someone who’s cleaned up enough of these messes, I can tell you: if you wouldn’t add a new point solution to your plan document, SPD, and SBC without a thorough legal review, don’t launch the app without the same process.

The “Glue App” That Sticks Everything Together - Including Your Liability

A related phenomenon I’m seeing more often is the employer-branded “member experience” app that pulls together your medical plan ID card, EAP, wellness programs, fitness perks, and condition management tools into one pretty package. Total Rewards loves these because they feel seamless to the employee. The legal reality is that they blend ERISA and non-ERISA pieces so tightly that the whole thing might become a single ERISA plan, especially if the medical features aren’t just incidental.

Worse, these apps sometimes accidentally morph into the primary way employees access benefits information - effectively becoming an SPD delivery mechanism - without satisfying ERISA’s electronic disclosure rules. That area is moving: in July 2026 the DOL proposed a safe harbor that would let group health plans deliver required disclosures electronically as a default, modeled on the 2020 retirement-plan rule. Until a final rule lands, the 2002 safe harbor still governs, and most apps are nowhere near meeting it. A participant who gets a denial through the app’s chat feature might have a fully valid claim under ERISA Section 502(a) that your internal process isn’t ready to handle. The house of cards gets tall fast.

So What Do You Actually Do About It?

Don’t panic. You don’t need to rip the app out of people’s hands tomorrow. But you do need to face the fact that this isn’t a freebie perk - it’s a piece of your health plan architecture, and you’ve got to treat it like one. Here’s the roadmap I walk my clients through:

1. Sort every feature into three buckets.
   • Pure education/general wellness: step challenges, generic tips. Not ERISA.
   • Medical care: telemedicine, coaching for specific conditions, therapy, medication management. ERISA plan.
   • Administrative navigation: find-a-doc, cost estimator. Plan administration, but structure carefully.
   If Bucket 2 has anything in it, you’ve got a group health plan. I don’t care what marketing says.
2. Put a plan wrapper around it. Either fold the app into your major medical plan (amending your plan doc, SPD, and maybe SBC) or set up a standalone “Digital Health Plan.” A standalone plan gives you cleaner boundaries, but you’ll need a separate SPD, Form 5500 (if 100+ participants), and COBRA. Yes, you need to calculate an app-only premium for COBRA - get the vendor to give you a rate.
3. Plug the app into your benefits administration system. Eligibility should flow from the system of record, not a spreadsheet. Automatic termination feeds and life event triggers are non-negotiable. Loop in your COBRA TPA with the rules and rates.
4. Lock down the data. Get a HIPAA business associate agreement that restricts the employer’s access to PHI to plan administration only. De-identify any dashboard data before it hits HR’s inbox. If you’re self-insured, the plan - not the employer - should be the customer, and you might need a privacy officer to sign a firewall certification.
5. Audit the vendor like a fiduciary. Demand evidence of security compliance (SOC 2 Type II, HITRUST, pen tests). Make sure they’re following the DOL’s cybersecurity guidance for plan service providers. If they can’t produce it, they’re not ready for prime time.

The Bottom Line

The days of treating a health engagement app as a casual perk are over. These tools have grown into serious components of how your people access care, and the regulatory world - however slowly - is catching up. The sharpest benefits leaders I know are already bringing these apps inside the tent, wrapping them in proper governance, and integrating them into the enrollment and compliance infrastructure. The rest are sitting on a phantom plan time bomb. An unfiled Form 5500 runs up to $2,000 per return even through the DOL’s voluntary correction program, and welfare-plan non-filing is exactly the kind of gap the agency has said it is watching.

Don’t be the one who waits for an audit or a lawsuit to find out you’ve been running a phantom health plan in plain sight. Because when that notice lands, it won’t be on a shiny app. It’ll be on your desk.