You have the chance to pilot your system inside a bank or a defence body. Most teams treat it as a technical demo and set out to prove the system works. That is the least of what the institution cares about. What it wants to know is whether it will stand stronger in front of its regulator after the pilot, or whether things got more complicated.
At Horizon we set out what a pilot like this should measure, and what it should leave alone.
Six things to measure
- Every action on sensitive data leaves a complete record. The target is 100 percent, with no tolerance.
- No sensitive data reaches a system that is not allowed to receive it. The target is zero.
- When the regulator or internal audit asks for the records of a specific operation, they are delivered in under one working day.
- Anyone without the right to access is always refused, and people with the right are wrongly refused in fewer than 2 percent of attempts.
- The work runs no more than 10 percent slower than it did before the pilot.
- The institution rehearses ending the relationship: it gets its data back, confirms you deleted it, and checks that itself.
A pilot that meets four of the six is worth continuing. One that misses the first or the second has failed, whatever else it achieved.
The one that decides everything
The third: how fast records reach the regulator. Every other measure checks whether the system does what it was built to do. This one checks whether it is worth buying, because today the institution turns every such request into a small project. If you meet it and the institution doesn't care, the idea itself needs rethinking, and finding that out in the first pilot is far cheaper than finding it out with the second customer.
Three things not to measure
- Cost savings, unless you measured what the work cost before the pilot. A saving with no baseline is a number without evidence.
- User satisfaction. Sensitive work has few users and no alternative, so their satisfaction measures the interface more than the idea.
- Compliance. No pilot proves compliance, because the regulator decides it. What a pilot proves is that the evidence the regulator needs exists and can be produced.
How to choose the work to pilot on
Three conditions: it carries sensitive data, it needs regulator approvals, and the institution handles it today with custom code or by hand. And remember that central bank approval takes at least 15 working days for a domestic bank, so put it in your plan from day one.
- To put your product to the Horizon test: horizon.bishernp.com/founders
- For the sector brief: horizon.bishernp.com/capital
References
- Horizon Studio RDI Center, success measures for a first pilot of sensitive systems in regulated institutions, 2026.
- Saudi Central Bank, Rules on Outsourcing, circular 41027017, December 2019, article 23.
Kimi K3 from the inside. What it takes to run it yourself
Planning to run the Kimi K3 model? This report analyzes its MoE architecture, true operational costs, and data localization readiness in Saudi Arabia to guide your decision.
Your data is in the Kingdom. What changed in 2024?
Where must your data be kept in Saudi Arabia? In 2024 the localisation rule moved to SDAIA, but the real questions now are about regulators, people and contracts.
Who owns Saudi space? A map of the sector in 2026
Working in Saudi Arabia's space sector in 2026 requires understanding its structure. This guide maps out the key players: the regulators, builders, PIF funders, and service providers like Starlink.