Why a launch plan matters now
On 3 October 2026, Gameball's CEO opened a waitlist for founding brands to try a layer that passes each customer's relationship with a brand to the AI assistants shopping on their behalf. AI shopping is growing fast, but it is still early: AI-referred retail traffic grew 393% year on year in Q1 2026, while OpenAI scaled back ChatGPT's in-chat checkout in March 2026.
That makes the first cohort a learning exercise as much as a launch. The brands chosen, and how carefully they are onboarded, will decide what Gameball learns before going public.
Pick the right brands
Use this check for every brand on the waitlist. Tick what is true for the brand you are looking at. The criteria and weights are my proposal, built to favour brands that can give a clear result in six weeks.
75 and above: founding-ready
50 to 74: ready after preparation
Below 50: not yet
A balanced cohort
- 6 to 8 brands, so each one gets a weekly check-in.
- Mostly existing Gameball clients, to lower setup risk.
- At least two markets, for example Saudi Arabia and the UAE.
- At least three categories, and one brand with physical stores as well as online.
The weekly rhythm
The same routine for every brand keeps a small team on top of 6 to 8 pilots at once.
How we'll know it worked
One main measure, a few supporting ones, and guardrails that stop the pilot if something breaks. Targets are starting proposals, to be adjusted once real baselines are in.
| Measure | What it tells us | Proposed target |
|---|---|---|
| Main Repeat purchases through AI assistants | Do members keep buying from the brand when an assistant shops for them? | Higher than the holdout group |
| Recognition rate | Share of assistant-driven orders where the member's tier and points were passed on | 90% or more |
| Pick rate vs holdout | How often assistants choose the brand for recognised members, compared with the holdout | Positive lift |
| Time to live | Days from signing to switched on for all members | 21 days or fewer |
| Guardrail Wrong balances shown | Customers shown incorrect points or tier | Zero |
| Guardrail Reward cost per order | Whether lift is being bought with discounts | Within brand's agreed budget |
| Guardrail Brand team effort | Hours per week the brand spends on the pilot | 2 hours or fewer |
Go/no-go in week 6: expand if recognition is 90% or more, lift is positive and no wrong balances were shown. Continue for 4 more weeks if traffic is too small to read. Stop if a guardrail was broken twice.
Track the pilots
Add each brand, move it through the six weeks and enter the measures from the table above. The page applies the week 6 go/no-go rule for you and shows what to do next.
Pilots are saved in this browser only. Nothing is sent anywhere. Export a CSV to keep a copy or open it in a spreadsheet.
The week 6 rule used here is the one above: expand when recognition is 90% or more, lift is positive and no wrong balances were shown. Stop when a guardrail was broken twice. Otherwise keep going. Leave a measure empty until it has been measured.
Risks and what to do about them
| Risk | Plan |
|---|---|
| Too little AI-assistant traffic to read results in six weeks | Favour categories where AI shopping already happens, track recognition rate as an early signal, and extend to ten weeks where needed. |
| Wrong points or tier shown to a customer | Member match check in week 2, internal test accounts, and a soft launch before full rollout. |
| Brand team too busy | A named owner is an entry criterion; Gameball prepares templates so the brand only reviews and approves. |
| AI platforms change their rules | Stay independent of any single assistant and focus on recommendation, not checkout, which has already shifted once in 2026. |
| Privacy and consent | Share loyalty signals only, never more personal data than needed; legal sign-off under PDPL or GDPR in week 1. |
Lessons from today's onboarding
Gameball's current self-serve onboarding already uses AI assistants to set up programs. Public reviews and help documents point to three failure points that the founding cohort should be designed around.
- Repeated answers instead of escalationA merchant whose domain verification kept failing received the same copied steps several times. In the pilot, a second identical answer should trigger a hand-off to an engineer.
- Hidden prerequisitesOn Shopify, customers must be able to create accounts before they can join a loyalty program. The pilot's week 2 checklist should test every prerequisite before launch, not after.
- A learning curve at the startNew users found settings hard to navigate at first. Founding brands should get a guided first session rather than a self-serve start.
About me
I'm Shorouk, a customer success and implementation professional with 7+ years across SaaS and consumer platforms in the Middle East, including e-grocery SaaS at Instaleap, food delivery at Talabat, and AI and HR SaaS at Cercli. I co-founded Helpawi, an Egyptian Arabic conversational AI. I work AI-first and I'm based in Cairo.
I made this kit because launching a new product with its first customers is the work I enjoy most. I'd be glad to talk about running Gameball's founding cohort.