Vision 2028
Decisioning in the future
Edvin Maehre
In two years, one view of each customer drives every decision
The customer
Sees their credit limit in real time, and how to raise it.
6/9/12: the first product built on it
Best customers: longest terms, largest baskets. Riskiest: shortest terms, smallest baskets.
Scalability
Every product, market and channel plugs into the same base.
Performance
Every rule and data source keeps proving its worth, or we remove it.
One view of each customer
One risk score per customer, updated on every significant event.
2
The base: one model gives every customer and every transaction a risk score
One customer over time: each significant event updates the score
Checkout
The transaction gets its own risk score.
Instalment paid
The customer score improves.
Instalment missed
The customer score falls.
New data shared
The model uses it, for example bank data.
Scalapay decides
Which data the model pulls in, and which events count as significant.
The case data shows the start
The internal score ranks risk. The bureau adds to the score through credit-history depth.
3
4
Scalability: every product problem becomes a data problem
Each block plugs into the same base
Shorter plans
Already build the view of each customer.
6/9/12
Reuses that view from day one.
New markets
Start from what we already know.
AI agent orders
New data: the metadata of every agent request.
One view of each customer
Agent shopping today
- Visa ran live agent payments with 30+ European issuers, including Nexi and Klarna (July 2026).
- Klarna and Affirm are in US agent checkout through Stripe and Google. They still check credit on each purchase.
- No European BNPL provider is live in agent checkout yet.
- Under CCD II, human review and explanation happen after the decision.
Position: guardrails that the customer sets in advance let the automatic decision go through. How to support agentic shopping is a business decision; risk says which data it needs, and when.
Performance: tend the garden
1. Propose
An analyst, or an AI agent that looks for better rules, proposes a rule on a low-code platform.
2. Test
The platform prices the rule on the control group, also beyond the risk frontier.
It stays only if it raises NPV per decision.
3. Decide
The analyst understands the result, then keeps, changes or removes the rule.
4. Monitor
Every live rule and data source is benchmarked against the quality we expect.
Then the loop starts again.
Every change is tested against a control group
A random holdout shows what each rule change really does, before we roll it out.
Rules earn their place
Rules run in parallel. Each is monitored on its own; a rule below its performance floor is removed.
Data providers earn their keep
Short contracts with several providers. Score with all of them, then keep, renegotiate or drop each one.
A human stays in the loop where it is necessary and appropriate.
5
The customer side: what one view makes possible
Customers understand their credit
The customer sees their credit limit in real time, and what shapes it.
A reason to share data
The app shows how to raise the limit, for example by sharing bank data, and what sharing is worth.
An explanation on demand
We tell the customer everything we can within our regulatory duty: their situation and how to improve it.
First step: test open banking data on a subset of our customers. If it performs, we productise and incentivise sharing data directly in the app.
6
How we get there
We align on our vision
We know where we're going. We find the quickest way there together.
Execution delivers value early
We maximise for speed and impact, building the plane while flying it.
Our first steps
For the business: we analyse our current setup and find the highest leverage improvements we can make right now.
For 6/9/12: we grow our portfolio by pushing volume to our best customers.
6/9/12 is step one.
Every step after it plugs into the same view of each customer, so each one starts from what we already know.
7