In this note:
Why did Amazon Web Services back an early-stage brand perception company?
What did the support include?
What changed in the report engine?
What does this mean for brand teams and agencies?
Where can I read the full technical case study?
Why did Amazon Web Services back an early-stage brand perception company?
Every brand is now read twice: by people and by AI. PokPok measures both readings on one method of 17 brand-perception markers, and that takes serious computing. Each report is hundreds of small jobs: reading a brand’s website, its top social content, its Amazon pages and AI answers, then checking every finding.
Amazon looked at our code before it looked at a pitch deck. Then it chose to invest in that code. A cloud provider does not do that for a company it expects to disappear.
What did the support include?
Startup credits. Room to build and test the engine without watching every server bill.
An architecture review. Amazon funded SnapSoft, one of its top partners, to take our system apart and tell us honestly where it would crack.
A full migration. SnapSoft rebuilt the engine on Amazon Web Services in about 20 business days and moved more than 40 of our workflows automatically.
What changed in the report engine?
No ceiling. The old engine topped out at roughly 2,000 reports a month. Now the system grows when orders pile up and shrinks when they don’t.
Proof under load. In a 40-brand stress test, the core service used about a quarter of its capacity.
Freedom to choose AI models. Every AI call goes through one gateway, so we can switch models on cost, speed or quality without rewriting a workflow.
Data stays in Europe. Everything runs in Frankfurt, which matters for European enterprise brands.
What does this mean for brand teams and agencies?
Speed and reliability. A report arrives in about 12 minutes, and it stays that fast when a hundred brands order on the same morning. An agency can run a whole category before a pitch, and an enterprise team can check every market at once.
What changed in numbers when PokPok moved to Amazon Web Services?
Old engine | On Amazon Web Services | |
|---|---|---|
Reports a month | Roughly 2,000, the ceiling | 40 brands at once used about a quarter of capacity |
AI models | Wired into each workflow | One gateway, switch without rewriting |
Where it runs | A hosted service we did not control | Frankfurt, inside Europe |
SEE IT ON YOUR BRAND
Brand Perception Truth reads your website the way people and ChatGPT do.
Where can I read the full technical case study?
SnapSoft’s write-up covers every service and design choice: Scaling PokPok.ai’s AI brand-analysis pipeline on a model-agnostic AWS architecture. Thank you to the Amazon Web Services startup team, and to Caroline, Gábor, Cristian, Márk, Merdjan and Blanka at SnapSoft.
QUESTIONS
Where does a brand perception tool like PokPok process a European brand’s data?
In Amazon’s Frankfurt region. Data flow and AI processing both stay inside Europe.
Can PokPok switch the AI models that read my brand without changing my reports?
Yes. Every AI call goes through one gateway, so we can switch models on cost, speed or quality without rewriting a workflow.
Who built PokPok’s report engine on Amazon Web Services, and how long did it take?
SnapSoft, an Amazon Web Services partner, in about 20 business days, funded by Amazon.
SOURCES
SnapSoft, Scaling PokPok.ai’s AI brand-analysis pipeline on a model-agnostic AWS architecture, published 4 November 2026.
The 12-minute report time: measured by PokPok.
NEW // V5.3.5 // BRAND PERCEPTION TRUTH
See how people and AI read your brand.
Your website read twice, by people and by ChatGPT, on the same 17 markers. The gap, the root cause, what to fix first. €150 a report.
MORE FROM THE BLOG
//
Read by AI
Why does ChatGPT recommend a competitor instead of my brand, and how do I find out why?
//
Brand drift
Why do my website, my Amazon pages and ChatGPT tell three different stories about my brand?
//
For agencies