It’s often the case that reconstructing a decision after a Cat event or claim can be difficult. Veriscopic are looking to enhance and improve that process;
Veriscopic, a UK insurance technology company, is developing a new approach to an often-overlooked operational problem across insurance: the cost and complexity of reconstructing consequential decisions after they have already been made.
The company calls this the reconstruction burden.
When an underwriting, claims, or delegated-authority decision is subsequently questioned through an audit, complaint, regulatory examination, litigation, or internal review, organisations may need to establish considerably more than the recorded outcome.
They may need to determine what information was available at the time, which policy and authority applied, which rules were active, what systems or AI recommended, which exceptions occurred, who reviewed the decision, and why the eventual judgement was considered reasonable.
Those elements can be distributed across numerous systems, documents, emails, and audit trails.
As AI becomes embedded within insurance workflows, Veriscopic believes the problem could become more significant.
Decision-making can increasingly be distributed between humans and systems while responsibility for the resulting decision remains with the organisation.
Veriscopic is developing the concept of decision-state replayability: preserving sufficient contemporaneous evidence of the conditions surrounding a consequential decision so that it can subsequently be replayed rather than reconstructed from fragmented evidence.
The company is also beginning work to measure reconstruction burden within individual insurance workflows, examining factors including reconstruction frequency, collective staff hours, number of people involved, systems and evidence sources consulted, and how much of that effort could potentially have been avoided through better preservation of the original decision-state.
Drew Young, Founder of Veriscopic, said:
“Insurance organisations are very good at recording outcomes. The harder problem can arise months or years later when somebody asks not simply what happened, but why the decision was reasonable given what was known at the time.”
Drew added:
“As AI becomes part of underwriting and claims, an audit trail showing that a system produced a recommendation and a human approved it may not always be enough. The organisation may need to demonstrate the state in which that judgement was actually exercised.”
Veriscopic is initially focused on consequential workflows across claims, underwriting, and delegated authority, with potential implications extending across the insurance and reinsurance risk chain.
One point I’d add since speaking with Insurance Edge:
The next step is measurement.
If reconstruction burden is a genuine operating cost, we should be able to observe it in real workflows, how often decisions need to be revisited, how many people and systems become involved, how long reconstruction takes, and how much of that effort could have been avoided.
That’s the question we’re now starting to test at Veriscopic.
Thanks to Alastair and Insurance Edge for covering the thinking behind it.