The moat is worth $2.
The real cost is $18.

The deck is American. The company is Canadian.

Every figure in the pitch is a US figure: 10,000 Americans turning 65 daily, 58M → 78M, FBI fraud data, “1% of US 65+ ≈ $280M ARR.” But the moat is Fibernetics, a Canadian network, and the channel is Fongo, a Canadian app.

So in the market you sized, you don’t own a network. In the US you would be renting telecom exactly like everyone else. The moat slide and the market slide cancel each other out — the moat only exists where the market isn’t.

Canada has ~8.1 million people 65+. One percent of that at $35/month is roughly $34M/year, not $280M. That is a good business for a company Fibernetics’ size. Claiming $280M you cannot serve is worse than claiming $34M you can.

And Canada’s own data tells a better story than the American slide does: 7,621 Canadian seniors died from falls in 2022 — up 51% in five years; fall hospitalizations rose 47% in a decade; falls cost the health system ~$2B/year.

One thing to check before the meeting. Fongo’s pitch is free phone service — which attracts price-sensitive, tech-comfortable, younger users. That may be the worst possible list for a $35/month emotional-premium product. Pull the actual age distribution before you present it as “warm distribution,” because the CEO can pull it in five minutes.

What happens when Mom falls at 3 a.m.?

This is the buyer’s actual fear, it is the entire reason the pendant industry exists, and Evergreen’s answer is: nothing happens.

Margaret falls at 3 a.m. She can’t reach the phone. Iris doesn’t call until 9. Two missed attempts later — around 9:15 — the family is alerted. Mom has been on the floor for six hours. Tellingly, the deck’s own “day with Evergreen” contains no emergency.

The scale: 1 in 4 older adults falls each year; ~43,020 Americans 65+ died from falls in 2024. Falls are the leading cause of injury death in this group, and “long lie” — time on the floor — is what decides whether a fall is an inconvenience or a death.

Meanwhile the spec disclaims it outright: “not a medical-alert device… we recommend Evergreen alongside those, never instead.” You are charging $35/month and telling the customer to also buy the $30/month thing that solves their real problem.

Your buyer’s decision isn’t “companion vs. nothing.” It’s “I have $40/month for Mom.” Against a pendant, Evergreen loses that in thirty seconds of thought.

But a correction to the obvious fix. Real-world fall detection is far worse than the marketing: one study of 18 older adults over four months logged 84 fall alarms — 83 of them false, with one true positive — and three real falls that never triggered an alarm at all. Real-world sensitivity: 25%, against a lab-claimed 94%. FDA has already recalled a Philips pendant whose fall detection silently stopped working.

So this is exactly why Evergreen does not build or claim fall detection. Your likeliest lawsuit is the false negative — “your AI said it watches for distress; my mother lay on the floor for fourteen hours; it never noticed” — and the way you never face it is to never hold yourself out as the thing that catches the fall. Physical safety belongs to a licensed partner whose whole business is answering for it. You compete on the relationship and the scam screening, not on being a cheaper pendant.

Removing 911 doesn’t reduce your liability. It manufactures it.

The spec is proud of this: “911 is never in the chain… That keeps us a companion and early-warning system — not an emergency service, and not the liability of one.” This is exactly backwards, and it is the most dangerous sentence in the document.

Liability doesn’t attach because you called 911. It attaches because you held yourself out as the thing that watches over someone, they relied on you instead of buying the thing that would have worked, and then you failed. You have: advertised detection, advertised alerting, induced reliance — and then deliberately removed the only escalation path that saves the person, routing the emergency to a sleeping civilian in another city.

A pendant company that never received the signal has a causation defence. You will have a timestamped transcript, an AI distress flag, and a log entry proving you knew — and a written policy of not calling for help.

The proof is in your competitors’ contracts. ElliQ — the most-deployed AI companion in North America, in the homes of New York State’s own seniors — does not disclaim dispatch. It disclaims detection. In capitals, in its live Terms: “ELLIQ… MAY NOT BE RELIED UPON TO DETECT THREATS, RISKS OR ANY EMERGENCY EVENTS.” Amazon went the other way: Emergency Assist routes “call for help” to live human agents who can and do dispatch EMS.

ElliQ says “we don’t detect.” Amazon says “we detect and we dispatch.” Evergreen has invented a third quadrant — “we detect and we refuse to dispatch” — and it is empty because nobody’s lawyers would sign it.

And there is a Canadian statute. Quebec’s Charter of Human Rights, s. 2: “Every person must come to the aid of anyone whose life is in peril, either personally or by calling for aid.” Untested against a tech provider — but you are a Canadian company proposing a documented policy of not calling for aid when you know a life may be in peril. This does not exist in Ontario or BC.

An AI that phones a senior every morning is a robocall

In February 2024 the FCC ruled that AI-generated voices are “artificial” under the TCPA, requiring prior express consent of the called party. The called party is the parent. The buyer is the child. Your entire go-to-market — “the adult child sets it up remotely in ten minutes” — is the exact fact pattern that generates liability, and appellate courts have held only the subscriber can consent.

And there is no exemption to fall back on. Every artificial-voice exemption caps at three calls per 30 days. A daily check-in is thirty. The product is lawful only on a consent theory. Damages: $500–$1,500 per call, no cap, class-actionable.

Your dignity feature is now evidence. The FCC holds that “stop,” “cancel,” “quit” are per se reasonable revocations. Now re-read the spec: “Iris never hangs up first… only ends the call when they say goodbye.” An AI that warmly talks past “stop calling me” is generating willful-violation evidence at $1,500 a call.

And your flagship feature is illegal on weekends in Canada. CRTC rules set automated-call hours at 9:00 a.m.–9:30 p.m. weekdays but 10:00 a.m.–6:00 p.m. weekends. “Every day at 9, Iris calls” — the most-repeated sentence in the pitch — is a rule violation every Saturday and Sunday.

Separately: “wellness trends” makes you a medical device. The FDA’s decision-support carve-out is limited by statute to software supporting health care professionals. Family-caregiver-facing output is categorically outside it. “Sounded more confused three mornings this week,” shown to a daughter, cannot be defended as clinical decision support.

The missing crisis protocol

You are shipping an AI that talks daily and intimately to socially isolated elderly people — and the spec contains no protocol for suicidal ideation, self-harm, or abuse disclosure.

Men aged 75+ have the highest suicide rate of any group in the United States: 40.7 per 100,000, against a national average of 14.1. Loneliness and isolation are the primary risk factors. You have built a product that selects precisely for that population and gives them a sympathetic voice to confide in — and you have not written down what Iris says when a widower tells her he doesn’t want to be here anymore.

This is also now legally required. California’s SB 243 and New York’s AI-companion law both define “companion chatbot” in language that describes Iris exactly, and neither is limited to minors. California requires a published crisis protocol with referral to a crisis line, plus annual reporting of referral counts — with a private right of action at $1,000 minimum per violation. New York requires a verbal notice, every three hours, to adults: “This artificial intelligence is not a human being. It is unable to feel human emotion.”

Say that out loud: telling a lonely 84-year-old widow, every three hours, that the voice she has come to trust cannot feel. That is not a compliance checkbox you bolt on in Q4 — it is a constraint that shapes what Iris can be.

Also missing: elder-abuse disclosure. Iris will be told, by someone, that their son takes their money. There is no protocol — and note that Evergreen currently routes every alert to the family. If the family is the danger, Evergreen is the abuser’s dashboard.

And the unresolved contradiction: the spec says “if a parent can’t meaningfully consent to a feature, it simply isn’t available to them.” But ~10% of over-65s have dementia and another 15–22% have mild cognitive impairment. You have written a consent rule that excludes your highest-need cohort from the features they most need. That isn’t a policy — it’s an unshipped decision, and it will get made badly by a support agent at 2 a.m. if you don’t make it now.

The $35 price point is already dead

The AI-companion market is in a race to zero, and Evergreen is priced at the top of it.

You are priced at 1.75× the nearest venture-backed competitor — and that competitor just halved its price. Meanwhile Amazon prices its flagship AI assistant at $19.99/month, above Prime itself — which means effectively nobody pays it. Amazon, with the largest install base on earth, does not believe consumers will pay $20/month for an AI voice.

And there is already a Canadian telecom in this market. TELUS Health Medical Alert: $60/month for home, $80 with a fall-detection pendant — discounted to $35 and $55 for existing TELUS customers. Read that structure: it is exactly the “we own the network so we bundle” play the deck says only nucleus can run. A telecom fifty times your size is already running it — and it charges more for fall detection than for companionship.

Note also the free-scam-blocking problem: T-Mobile, Verizon and AT&T all give network-level scam blocking away at $0. You propose to charge $12/month for it à la carte. Your version is better — but the category name is priced at zero in the customer’s head.

Six pillars is not a product. It’s a features list in a product’s coat.

Scam Shield is the wedge. It is the only pillar that is an acute named fear, that the parent can’t get free elsewhere at carrier level, that is technically defensible via Fibernetics, and that demos in thirty seconds. It is also the only pillar where the “why nucleus wins” argument survives contact. Lead with it.

Cut Good News. Nobody in the history of commerce has bought a $35/month product because of an uplifting news read. It adds an editorial pipeline, a positivity filter, a human spot-check, and a local-ad sales motion — an entire media business bolted to a care product. You price it at +$5/month, which is you telling the customer it’s worth almost nothing.

Cut StoryLine from the pitch. It is the most beautiful thing in the document and it is a beautiful distraction: the most expensive inference minutes in the product, the most sensitive data you will ever hold, plus a printing and fulfillment business. It is a gift product with a gift purchase cycle, grafted onto a safety subscription with a fear purchase cycle. And it does not sell the product — no adult child at 2 a.m. googles “AI memoir for mom.” They google “how do I know if my mom fell.” Keep it on the roadmap. Take it off the deck.

Demote Patience-as-a-Service — OpenAI already gives it away. 1-800-ChatGPT is a free phone number that works from Canada, on a flip phone and a rotary landline, with no account. It is endlessly patient. It will walk Margaret through the printer. Your “zero-adoption front door” was shipped by OpenAI, for free, in December 2024. Your differentiator isn’t the patient voice — it’s the memory of this family and the digest to the daughter.

Keep CareCircle. It is the real asset (see Finding 09).

The sharpened product: Scam Shield + Daily Check-In + CareCircle, with a licensed monitoring partner standing behind the emergency. One price. Three things Evergreen owns and one it delegates — all pointed at the same parent, a product the CEO can repeat to his board from memory.

One more omission a senior-care expert would flag immediately: hearing loss. Roughly 1 in 3 over-65s. The deck’s answer is “Iris speaks lower and slower” plus captions. But if the parent cannot hear Iris, every single pillar fails at once — check-in, tech help, scam screening, all of it. That is a single point of failure for the entire product, currently treated as a typography note. Test it. Budget for hearing-aid pairing and a text fallback.

And the device reality: only 17% of Americans 80+ own a smartphone. Two of your three surfaces — Evergreen Phone and Evergreen Tablet — and nearly all of the mockup serve a minority of the target user. Evergreen Line, the no-app flip-phone product, is the real product — and it has the least design work in the deck. Invert that.

What the moat should actually be

Telecom cost isn’t a moat — it’s a $2 discount that Google erases with a rounding error. When OpenAI or Google ships a free AI companion that calls your mother, here is what they cannot trivially copy:

• The family graph and the escalation contract. Who to call, in what order, at 3 a.m., and what they’re authorized to do. Nobody re-enrols their whole family and re-negotiates who’s the primary contact to save $10.
• The longitudinal baseline. Six months of this person’s voice, cadence, word-finding, latency, mood. The value isn’t “Iris talks.” It’s “Iris knows what normal sounds like for Margaret, and this morning wasn’t normal.” That compounds daily and is worthless to a competitor starting from zero — and the deck mentions it in one bullet and then forgets it.
• Carrier-native call control. Screening a call before the phone rings, in the signalling path. Genuinely hard to copy without being a carrier. This is the real Fibernetics advantage and it is worth more than the per-minute rate.
• Trust — which is the whole category. Ask Papa: 1,200+ complaint reports, allegations of assault and theft, and Humana, Aetna, Molina, Cigna and ~36 other payers gone in a single renewal cycle after one Bloomberg story. A $1.4B company with locked-in distribution lost the category in a news cycle.

The moat slide should say: “Not the network — the relationship. We own the family’s escalation contract, and six months of knowing what your mother sounds like when she’s fine. Both compound. Neither can be bought.” And then, as one line, not a slide: “and yes, we own the network, which is why we can screen the call before it rings.”

Nobody has ever made “adult child pays” work

The CEO’s real question is not “is this a good idea.” It is “if this is such a good idea, why is the field littered with corpses?” You must answer it before he asks.

The pattern is unmistakable. The dead were consumer-pay. The living are institutionally paid — ElliQ (state agencies), Rendever (NIH grants), Sensi.AI (home-care agencies), DUOS ($130M, payer-only), Cera ($150M, NHS). The venture money in this category has looked at consumer-pay and walked away from it.

Meanwhile your deck asserts “≈0% churn — nobody cancels the thing keeping their mother safe” with zero evidence, and no published churn figure exists anywhere in this category. Worse, the claim has a structural flaw nobody said out loud: your customer is 78. The churn event is death. The second-biggest is moving into assisted living. Putting “≈0%” in front of a telecom CEO who models churn for a living is how you lose the room.