AI & Model Assurance
Relying on AI? Make sure it's right.
If your organisation uses AI to forecast, price, analyse or decide, we can test whether it gets the numbers right, holds up over time and delivers what it promised. The services below are typical starting points, not a fixed menu. Tell us what your AI does and we'll design the test around it.
Who is this for?
Finance and FP&A teams relying on AI forecasts · Public bodies deploying AI in service decisions · Vendors who need independent evidence for a procurement · Insurers and lenders using AI in underwriting or credit decisions.
Output accuracy review
How often is your AI wrong, and where?
We test your AI tool against a reference set built for your use case and tell you how often it is wrong, where, and what to fix.
You get:
-
Error rates with confidence ranges
-
Where and why it fails
-
Fixes, ranked by priority
-
A signed opinion
Typical duration: 2-3 weeks
Model validation
Does your model hold up?
We test whether your forecasting or pricing model is well specified, stable over time and better than a simple benchmark.
You get:
-
Backtests and benchmarks
-
Stability and stress tests
-
Findings, ranked by severity
-
A signed report
Typical duration: 2-4 weeks
AI impact evaluation
Did your AI investment pay off?
We measure whether your AI investment delivered the savings or productivity it was bought for, using evaluation methods that separate the AI's effect from everything else going on.
You get:
-
What your AI really delivered
-
Separated from other factors
-
Ranges, not headlines
-
Evidence for the next decision
Typical duration: 3-5 weeks, depending on data
AI governance and monitoring
Keep your AI under control after go-live
We assess how your organisation inventories, tiers and oversees its AI against recognised frameworks, and we can re-test tools we have reviewed each quarter to catch drift.
You get:
-
An AI inventory, tiered by risk
-
Gaps against recognised frameworks
-
A prioritised plan
-
Quarterly drift checks
Typical duration: 2 weeks, monitoring is ongoing.
Why BlackPoint
We report against recognised frameworks including ISO/IEC 42001, the NIST AI RMF and DSIT's AI Management Essentials, and quality-assure our own work to Aqua Book principles.
01
Econometric rigour
Twenty years building and testing forecasting and risk models at Oxford Economics, the EIU, Dun & Bradstreet and Lloyds Bank.
02
Independent
No ties to AI vendors or developers, so our only interest is getting the answer right.
03
Accountable
Every report carries a named reviewer's signed opinion you can put before a board or regulator.