For credit unions
Amblee holds gambling apps, prediction markets and impulse spend apps shut on a member’s own phone until they have moved the amount they chose into their own savings. A credit union can sponsor it for its members. No credit union has run this yet, so year one is a measured pilot and not a deposit promise.
One thing decides whether this is relevant to you at all: online sports betting is legal in your state, or it is not.
There are no credit unions using Amblee today. You would be the first, and the pilot is scoped and priced as a first.
Term
6 months
A limited-scope pilot with a defined end and a pre-registered read-out. It converts to twelve months on written notice, or it stops.
Cost to the member
$0
The credit union sponsors the seat. The member is never billed, never asked for a card, and never required to open an account with you.
Member data from you
None
You send Amblee no member file, no names, no member numbers, no core access and no single sign-on, and the primary analysis runs on your data, on your side. An enrolled member does share things with Amblee directly, and the next card says what.
The member installs Amblee
Our own app, from the App Store. There is no credit union binary to build and nothing to embed in your banking app. Your brand would appear inside Amblee’s app; that is work to be done, not a screen you can look at today.
You sponsor the seat
The design is one shared Apple offer, with Amblee issuing the seat once the member is bound to your program. The member enters no card and is never charged. None of that binding exists yet: today Amblee can grant months by redeemable code, and nothing in it knows what a credit union is.
What the member shares with Amblee
Read-only bank access through Plaid, the amount they commit to, which categories they block, and anything they write in a hardship exit. That is the member’s own decision, made in the app, and none of it comes from you or goes back to you. It is set out in full on trust and privacy.
The money is the member’s
The destination is an account the member owns, at your credit union or anywhere else they choose. Amblee never holds, moves or accepts a dollar of it.
The fee is flat
A program fee. Nothing varies with deposits, balances, assets under management or the dividend rate you pay.
The product
Four steps, on the member’s own phone, under the member’s own consent. None of it touches your systems.
Sportsbooks, casinos, prediction markets, the impulse-spend apps. The member chooses them in Apple’s own Screen Time picker on their own phone. iOS hands Amblee an opaque token rather than a list, so we never learn what is in it, and neither do you.
The member sets the monthly amount, and nothing changes it until the member confirms. The app does offer a starting point, built from the member’s own answers: the goal field opens at the monthly spend the member has just entered, held inside the app’s minimum and maximum, and a member who names a savings target and a date is offered what is left to save divided by the months left. The spend answer itself can start from figures Amblee chose: the question offers three quick answers to tap beside a field for the member’s own figure, on betting today and on every money habit in the iPhone onboarding now being built. Amblee never raises a number on its own and never derives one from an account balance or an income. A floor and a ceiling would be yours to configure once there is a tenant to configure them on. A member’s own number would never be.
A read-only Plaid connection sees the deposit clear, or the member logs a screenshot that is read on the phone and never uploaded. Either counts. Both are meant to be money landing in an account the member owns, and a screenshot is taken on trust: the only person a made-up one fools is the member who made it.
For the rest of the month. Pull the money back out of a connected account within 72 hours of the unlock, enough to drop below the number, and the block returns. Only a connected account shows money leaving, so a screenshot never re-locks. Once the 72 hours have passed the apps stay open until the 1st, when it all starts over and the number is due again.
A card decline stops a payment, and a member who wants to bet routes around it. Amblee shields the apps themselves under Apple’s Screen Time, on the member’s own device and under their own consent, and it opens when a deposit into their own account clears.
iOS shields the apps on the phone Amblee is installed on. A second device, a desktop browser, or a counter in a betting shop are all outside it. Deleting Amblee is possible unless the member has switched on the commitment lock, which is opt-in and off by default, and a member determined enough can still erase the phone. A tool that tells you it is airtight is lying to you.
Inside the app, members get the short version: blocks cover that iPhone only, and the commitment lock can stop apps, Amblee included, from being deleted.
Sponsorship
There is no credit union binary and no SDK inside your banking app. Apple grants the Family Controls entitlement per App ID and reviews the specific app, and Screen Time tokens are opaque and scoped to the app that created them, so an SDK could not be thin even if you wanted one. Your brand would appear inside Amblee’s app, and the entitlement stays with Amblee. The co-branding is a build, not a setting: there is nothing in the app today that knows your name.
Apple, Requesting the Family Controls entitlement, and Meet the Screen Time API (WWDC21). developer.apple.com
Your credit union would never distribute a code and never handle one. It distributes a link, and Amblee issues the seat after the member is bound to your program. The link goes out through channels you already own: a message in digital banking, an email to a defined frame, a card at the branch. The link and the binding behind it are both still to be built; the nearest thing that exists today is a redeemable code, which is the shape we are deliberately not using.
Not to the members your transaction analytics flag as bettors. Being singled out by your credit union for betting is a disclosure, and it is the one complaint most likely to end a program like this. It is branded as a savings commitment, never as a gambling program, and members self-select the categories they block. There is no offer through collections. Not governed. Cut.
The money path, in full. The member’s checking to the member’s savings, both of them yours, moved by your own rails on the member’s own instruction inside your own authenticated session. Amblee reads, read-only, through Plaid, that the credit cleared. Amblee never holds, accepts, originates or returns a dollar.
These are contract covenants rather than settings, and they are identical for every tenant. A credit union that needs one of them moved is not a credit union we can sell to.
A floor and a ceiling would be yours to configure. A member’s own number would not, and Amblee will not set, suggest or raise a member’s number for the credit union.
Three exits a month, 24 hours each, after a 10-minute delay, written down and audited. It is the release valve that makes the lock safe to wear, and it is identical for every tenant.
Money pulled back out of a designated account within 72 hours of the unlock, enough to drop below the member’s number, re-locks the apps. Same window for every member of every credit union.
Net new money into an account the member owns, and once that 72-hour window has passed, the apps stay open for the rest of the month. Not time served, not a discretionary override, and not something a support ticket can talk its way past.
Any qualifying account the member designates unlocks the apps, at any bank. Your incentive is to make your product the one they choose. It is not Amblee’s job to force it.
No screen in Amblee states an APY today, and none would state one for you. The design names your savings account and links to your own rate page, because a rate on our screen would be our advertisement, not yours.
12 CFR 707.1(c) applies the Truth in Savings advertising rules to any person who solicits an amount for placement in a credit union. law.cornell.edu
If the contract ends. Every seat converts to a 60-day no-charge grace and then to standard consumer pricing with clear notice. No member loses a lock they are relying on because two companies fell out. That clause is non-waivable.
What you never see
Not by report, not by export, and not on request.
All of it aggregate, and none of it built. The only aggregate reporting in Amblee today is the creator program’s, which bands every count down to a floor of three and is scoped to a creator, not to an institution. Organisation reporting would raise that floor to ten, which is the k-anonymity threshold in the plan, and would add cross-tab and complementary suppression, so a suppressed cell cannot be recovered by subtraction from a published total. A small cohort would then see mostly suppressed cells, and that belongs at signature rather than at report one.
Not at member level, not on request, and not as an escalation. There is no route to it.
Amblee is a budgeting app. Which habit a member picked, and each attempt the app blocked, say something personal about how that member spends money. The member gives that to Amblee to hold their own spending shut, not to their credit union to market to, price or lend against.
Amblee itself holds a member’s categories, because it has to in order to enforce the block. What does not exist is the other side: there is no organisation schema in Amblee today, so this is a commitment about what gets built rather than a description of a query that already fails. No per-member category field anywhere an organisation-scoped path can read, and a lint rule that fails the build if one appears.
Two data-minimisation rules already fail the build that way. One forbids a financial value in a URL or a query string, the other forces every raw ledger write through a sanitiser, and this would be the third, written the same way. A policy a support engineer can be talked out of at 11pm is not a policy.
No signal derived from blocked categories, block attempts, hardship usage or written reflections may be transmitted to, stored in, or used by your core, your marketing automation, your collections treatment, your pricing, or any underwriting decision. A flag saying that a member blocks gambling, if it touches a lending decision, ends the partnership.
Designed, and unbuilt. The plan is a token the member generates in the app and reads to your representative: time-boxed, showing lock state, seat state and progress totals and nothing else, with no lookup by name or member number and an audit row written on every lookup. It is the piece that makes a refusal a non-event rather than a confrontation, so it ships before the first member is invited. Until it does there is no console for your staff at all, and a member with an app problem comes to Amblee directly.
Volume, as a number rather than a control. The pilot carries a monitored guardrail of fewer than 0.15 tickets a month for each enrolled member on your side, reported monthly with the other guardrails. App-side tickets come to Amblee directly. Crossing that line is our problem to fix and it is written down as one.
The arithmetic
A member moving $200 a month from their checking with you into their savings with you brings you no new funding and makes your cost of funds worse. FDIC national deposit rates as of 17 August 2026: interest checking 0.07%, savings 0.38%. Carry $1,200 of average balance across that gap and you are paying 31 basis points more on money you already had. At 400 enrolled members that is roughly $1,500 a year against you.
FDIC National Rates and Rate Caps, 17 August 2026. fdic.gov. $200 a month is an illustration, not a measured Amblee average.
Value only appears where the money is genuinely new: money that would otherwise have been staked, spent, or moved to an outside savings app. Nobody has measured what share of it is new. Not us, not anyone.
An analysis of 1.76 million members across 31 credit unions.
of checking accountholders had gambling or prediction-market transactions.
lower median deposit account balances among heavy bettors than among casual bettors.
Filene Research Institute, 22 June 2026, reported by CU Today, 13 August 2026 and CU Times, 25 June 2026. The member and credit union counts are from CU Today’s report. CU Today describes the 56% comparison as one with non-bettors; Filene’s own post says casual bettors, and this page follows Filene.
Three caveats, and they belong next to the numbers rather than under them. First, the 56% compares heavy bettors with casual bettors, not with members who do not bet, and Filene’s post does not say where it draws the line between the two. The post does say that once age and segment are controlled for, bettors’ and non-bettors’ checking balances look about the same, and that it is non-bettors who are building savings, money market and certificate balances, but it puts no figure on that gap.
Second, CU Today also reports that 4% of members account for half of all betting dollars. That is a share of members and the 9% is a share of checking accountholders, so the two do not reconcile into one concentrated-cohort story, and we do not stack them.
Third, 56% lower balances is a correlation. Filene did not establish that betting caused the lower balances, and we are not going to tell you it did.
Credit outcomes after legalization of sports betting.
credit delinquency overall.
card delinquency among under-40 borrowers.
auto delinquency among the same borrowers.
Goss and Mangrum, Liberty Street Economics, 25 March 2026. A population-level finding after legalization, not a finding about your members.
Your transaction data will tell you what share of your members are betting and what leaves for the sportsbooks. In the CU Today report above, Vertice AI’s chief executive puts that analysis first, before any new member education campaign.
Net income per member across the system runs about $140 a year. Keeping members who would otherwise have left is the largest line here in year one and the hardest to prove, which is the whole argument for the measurement design below.
Derived from NCUA, Quarterly Credit Union Data Summary 2026 Q1. ncua.gov
The average cash incentive on a new checking account in 2025 was $277, and the deposit incentives that source surveys run from $100 to $500. A competitor in the credit union wellness category publishes an industry account acquisition cost of $498.
ProSight Financial Association, and Debbie’s partner page, which is where the $498 figure is published. Their arithmetic, not ours.
The mechanism is in the New York Fed finding above. The size of it at your institution is not, and we are not going to populate it with invented loss rates. It needs two numbers that only you have, off your own 5300: net charge-offs and average balance on unsecured consumer loans.
Standing up a new vendor costs you internal time before member one: the marketing send, branch and contact centre training, compliance review of the campaign copy, the vendor diligence file, and support routing. Our estimate is 60 to 150 loaded staff hours. That is an assumption rather than a measurement, and the only place the real figure lives is your own project office. It belongs in your number, not in a footnote of ours.
Year one
No vendor in this category has an independent outcome study. The published figures are self-reports.
The analysis plan is published before anyone is invited, and it says in advance that the answer may be null. If it is null, it gets published. That is the only thing that would make a second result worth believing.
You randomise who is invited, stratified on your own internal flag so subgroup analysis stays powered without a single member being singled out. The not-invited arm is invited at month seven. Everybody gets access in the end.
A savings-balance endpoint cannot see a member moving money out of their own checking, so it would report a large fake win. This is the most important line in the measurement plan and it is the one we would want you to check first.
No member data flows from you to Amblee for the primary analysis, and none flows back. Amblee would supply aggregate uptake counts and nothing else. You hold the rest and you run the numbers.
Complaints attributable to the program, with zero tolerance for a coercion complaint. Hardship exit rate, because a spike means the unlock is too tight. Contact centre load. Prohibited-field disclosures, where the tolerance is zero and any breach stops the line.
One credit union cannot power a credit outcome. Detecting a one point absolute shift from a 3% delinquency base at 80% power needs thousands of members in each arm on every convention we can compute it on, and an enrolled pilot cohort is a few hundred. So delinquency and charge-off are reported as underpowered, with intervals, and pooled across institutions later. Continuous outcomes are better served, but deposit balances are heavy-tailed, and the Filene finding is itself stated as a median, so the plan pre-specifies rank and quantile methods rather than means.
The only published benchmark for a savings product distributed by a financial institution is Pew’s finding that 17% of Michigan credit union members had access to Save to Win accounts in 2013, and that at the typical Michigan credit union offering it, only 1.3% of members opened one. Two caveats before anyone builds a forecast on it: that is a raffle account, and Amblee asks harder, because it asks a member to install a second app, hand it authority over their phone, and commit money. And the denominator in that figure is ambiguous enough that we will not turn it into an enrolment number for your membership.
Pew Charitable Trusts, November 2018.
The counterweight is what happens after adoption. Monzo’s gambling block reached more than 400,000 activations with fewer than 10% permanently switching it off. Different product, different country, different friction at sign-up. It is evidence that people keep a block once they have one, and it is not evidence about how many will take one from you.
A named number of member sends on a named schedule, a message in digital banking, branch collateral, and one named internal owner. It is written into the pilot as a condition of the refund, because a program nobody sends is how this fails. We would rather you decline the sends than sign for them and not make them.
Invite everyone. Among the members who opt in, randomise immediate access against a six-month waitlist. It is weaker on external validity and much easier to approve internally: nobody is denied something they asked for, and everyone who wanted it gets it. We will run it, and we will label it as the weaker design in every output it produces.
One randomisation list at the start, then one extract a month on a fixed schema, from your warehouse to your own analyst. Amblee writes the pre-registered analysis plan and supplies the code that runs it; you run it and you hold the output. We have never done this with a credit union, so we will not quote you an hours figure we have not measured. What we will do is agree the field list and the cadence in writing before you sign, so your team can price it rather than discover it in month two.
Suppose uptake and retention clear their thresholds and the deposit effect comes back at zero. The refund is keyed to enrolment and retention counts, so it does not trigger. You keep paying for the term, and you now own published evidence that this does not move deposits. That is the real year-two decision, so decide before you sign what you would do with that result. Our answer is that the evidence is worth having and we would publish it either way. Yours may be that the program only makes sense as a member wellbeing line, in which case say so at signature and scope it that way.
What it costs
We could be paid in basis points on the deposits this brings you. It is lawful with a credit union: the brokered deposit regime reaches insured depository institutions, and NCUA imposes no equivalent limit on federally insured credit unions. We are declining it, in writing, and the declination goes into the contract as a recital. The reason is not modesty. Amblee decides whether a person’s phone opens, for a population defined by impaired financial self-control. A fee that rises with the amount they deposit at one named institution is a conflict we would rather not have to explain to you, to an examiner, or to a member.
NCUA, Brokered and Reciprocal Deposits FAQ. ncua.gov
The definition that keeps it clean. An Enrolled Member is a member who, on the last day of a calendar month, would be bound to the credit union’s Amblee tenant and has an active lock configuration with a non-zero monthly commitment.
Note what is absent from that definition. Any reference to a deposit.
There is no price on this page. The fee has a platform component and a per-enrolled-member component, an enrolment cap so a credit union that markets hard is not punished for it, a pilot rate credited against year one, and a refund condition tied to pre-registered endpoints, and a number without those is a number you would have to unlearn. Ask and we will send the whole card, including the pilot terms and the refund condition.
Two things that cost nothing to state and that a CEO asks first. The pilot is exclusive in your market for its term: no second credit union whose field of membership materially overlaps yours while it runs. And the fee is invoiced, not taken out of anything, because there is nothing of yours for Amblee to take it out of.
The conflict, stated rather than managed. Your return on this is highest when a member’s money lands in the cheapest account you offer, and the member’s return is highest when it does not. Our fee is the same either way. That does not resolve the conflict. It removes us from it.
NCUA’s Community Development Revolving Loan Fund awarded $3.47 million to 97 low-income-designated credit unions in fiscal 2026, with a maximum award of $100,000. The Technology, Cybersecurity and AI category made 35 grants averaging about $21,500. Around 2,379 credit unions hold the low-income designation, per NCUA’s own quarterly data. If you are one of them, we will help write the application.
We have no relationship with NCUA, no view on your eligibility, and no idea whether an application would succeed.
NCUA, 29 June 2026, and the designation count from NCUA’s Quarterly Data Summary, 2026 Q1. ncua.gov
Diligence
The gaps, in one place, before the questionnaire asks for them.
This is the headline of the section and it belongs first. Amblee today is a consumer app with a consumer backend, on iPhone with Android in testing, and it has no notion of an organisation at all: no tenant, no credit union administrator, no sponsored-seat binding, no co-branding, no aggregate reporting, no support token. Every mechanism this page describes on your side of the line is a design and a commitment, not a screen you can be shown. The consumer product underneath it is real and shipping on iPhone, and the Screen Time shield, the deposit check, hardship and the reversal relock all work there today exactly as described above.
None. Nothing writes to your core, nothing runs inside your online banking, and no plugin ships in year one.
No core or digital banking platform lets a third-party app open a deposit account, and we are not going to spec around that. Later, and only after evidence, Amblee can deep link into your own opening flow with a referral parameter. Origination, identification and eligibility stay entirely yours.
Amblee does not move money and does not intend to. In year one the member sets the transfer up in your own digital banking and Amblee reads that the credit cleared. There is no wallet, no pooled account, no ACH file and no transfer button anywhere in the app.
Not in year one. It converts better, and it puts member data in our hands, which changes your diligence file. Year two, with a report in hand.
Never. Apple grants the Family Controls entitlement per App ID and reviews the specific app, and there is no published policy on granting it to a bank-branded binary, so nobody can pre-clear it. We will not sell you a roadmap we cannot promise.
Apple has not been asked in writing yet, and the request goes in before you sign. If the answer is no, the pilot-scale fallback is that the member subscribes and you reimburse. That costs more and changes nothing about the product.
None today. Type I is on the critical path, and its delivery date goes in the contract with a fee credit if it is missed. Type II follows, and the observation window is the one part of it that cannot be compressed. A pilot that receives no member data, touches no funds and has no core access is the only honest shape for a first deal without one.
Amblee is built by John Harzan as Amblee LLC. There is no team behind the name. Key-person risk is real, and three answers are offered here without being asked: source code escrow with release triggers, a data export on 30 days’ notice, and a 90-day wind-down obligation. No escrow agent has been chosen yet.
Amblee never accepts, holds or receives member funds. Not a pooled account, not a sweep, not briefly. That one sentence is why there is no state licensing, no FinCEN registration and no BSA/AML program in this file.
Compensation does not vary, directly or indirectly, with deposits, balances, assets under management or the rate you pay, and Amblee is paid by nobody in connection with the placement of member funds. That sentence is a recital in the contract, not a sales line.
Amblee states no rate on any screen and receives nothing from any adviser, custodian or brokerage. It does recommend one provider and shows a grid of brokerages, and both belong here rather than in your diligence findings. In the iPhone app, a member saving by investing taps the brokerage they already use in that grid, so the app can open it when it is time to deposit. A member who says they have no account to put the money in is shown a single card for Wealthfront, a cash account for a saver or an investing account for a member who chose investing, labelled “Recommended, not affiliated. Any account works.” Amblee is not paid for either, and nothing requires a member to use them. In a sponsored program the design puts your savings account in that place instead, and that is not built.
A structure where Amblee holds member funds in an account at your credit union works on paper. Under the model money transmission act it needs the credit union to assume all risk of loss for funds in Amblee’s hands, and the long-standing precedent for that structure still ships a screen telling consumers the funds may not be share-insured. A product people use to protect their own money cannot ship that screen. The analysis goes in the vendor pack.
On being examined. We will not sell you this as unregulated. NCUA has no authority to examine third-party vendors, so your examiner cannot look at us and leans harder on your file about us instead. A missing regulator means more diligence, not less.
NCUA, Third-Party Vendor Authority, March 2022. ncua.gov
There is no trust centre at amblee.co yet, so this is the honest inventory. Sent on request, in one file, before a call: the information security policy, the data retention and disposal policy, our answers to Plaid’s security questionnaire, the subprocessor inventory, and the never-list as a draft contract schedule. Not yet in existence: SOC 2, a completed SIG Lite or CAIQ, a tested business continuity plan, and a penetration test summary. Those, and a trust centre to host them, are committed before a first pilot signature, and the SOC 2 Type I date goes in the contract with a fee credit if it is missed.
Limitation of liability, stated before it is asked for. Our opening position is two to three times annual fees generally, with a separate and higher cap for data-security claims backed by the cyber policy limit. That is a position, not a prediction about where a negotiation lands. Technology errors and omissions and cyber cover are being quoted now, and we are not going to print limits on a public page that we have not yet bound.
The person who answers, negotiates and signs is John Harzan, who builds Amblee as Amblee LLC. You can read the rest of that on the about page, and the member-facing version of the data posture on trust and privacy.
Getting started
A stop condition, not a formality. Without a reliable read of member deposits there is no observation layer and no deposit endpoint, and screenshots alone cannot carry one. Two ways to settle it before you write: send the name and routing number and Amblee checks coverage and tells you either way, including when the answer is no; or ask your own digital banking team whether members can already link your accounts inside Venmo, Robinhood or Chime, because that is the same coverage.
Your vendor file and our answers. The never-list in front of your compliance officer as a non-negotiable schedule. Campaign copy through your own advertising review. Apple’s written guidance request on the sponsored seat goes in during this window too.
You define the invited population and randomise it, or you choose the waitlist design. The analysis plan is pre-registered publicly before anything is sent. Nothing is built that depends on your core.
Invitations go out through your channels, the funnel is reviewed weekly, and you get a PDF a month with a version number and a generation date, so it can go in a board pack without being rebuilt.
An honest calendar. Six weeks after signature is the floor. For a first credit union, 16 to 24 weeks is realistic, and diligence rather than engineering is what fills it. A promised four weeks that takes fourteen is how a relationship starts badly.
Contact
Four things in the first email make the first call worth having.
If you would rather send your vendor questionnaire before you send an introduction, send the questionnaire. We would rather answer a hard file in week one than reach it in month three.
One person reads it, and you get a reply either way.
Nothing on this page is legal, investment, tax or accounting advice. The regulatory descriptions are our reading of the public sources linked beside each one, and they are not a substitute for your own counsel. Amblee is a self-restriction tool, not treatment, and nothing here claims a clinical outcome. Every figure attributed to a third party carries its source and date; every figure that is our own arithmetic says so, and every assumption is labelled as one.