How Can AI Support Fair Value Monitoring in Delegated Authority?

Evidence about an insurance product can be spread across the organisations that manufacture, distribute and service it.

In delegated authority, an insurer may hold product and underwriting information while coverholders or MGAs hold details of the final price, remuneration, customer interactions, claims and complaints. Service providers may hold further evidence about how the product performs in practice.

For products within scope, FCA product-governance rules require relevant firms to assess and demonstrate fair value. The operational challenge is creating a reliable product-level view from information collected for different purposes and in different formats.

Why fair value evidence is difficult across a delegated chain

Fair value concerns the relationship between the overall price paid by the customer and the quality of the product and services provided. That requires more than a loss ratio or a review of premium alone.

Useful evidence may include product benefits and limitations, distribution charges, remuneration, claims acceptance and settlement experience, complaints, cancellations, service performance and information about customers in the target market.

The parties in a delegated chain may define and aggregate these measures differently. One coverholder may report complaints by product, another by binder, and another at business-unit level. Claims figures may cover different periods or use different status definitions. Combining them without resolving those differences can create a misleading comparison.

Regulatory scope must also be confirmed. FCA PROD 4 does not apply identically to every insurance product or market participant, and some contracts, including certain large risks and reinsurance, can fall outside relevant provisions.

The established product governance approach

A sound process starts with role clarity. Firms identify the manufacturer and distributors, agree what information is needed, define the product and target market, and set a cadence for review and challenge.

Teams then assess the evidence together. Pricing and remuneration are considered alongside the benefits, limitations and service actually delivered. Adverse indicators are investigated, documented and escalated, with changes made where the evidence shows that action is needed.

This approach relies on professional analysis from product, actuarial, conduct, claims and delegated authority specialists. Its practical weakness is the effort required to standardise submissions and examine narrative material before those professionals can perform the assessment.

Where AI can improve evidence preparation

AI can support the preparation and exploration of the evidence. It may help to:

• Map differently labelled fields into an agreed product-governance dataset.
• Classify complaints and claims narratives against controlled themes.
• Identify missing periods, inconsistent definitions and unusual movements.
• Compare outcomes between relevant customer or distribution segments.
• Prepare source-linked summaries for a product review meeting.

This can reveal questions that a single aggregate measure would miss. For example, stable overall claims experience could conceal a cluster of declined claims or complaints within one distribution segment.

The finding is a prompt for investigation. A difference may reflect product mix, reporting quality, a temporary event or another legitimate factor. Product-governance professionals need to validate the evidence and establish whether it indicates poor value or customer harm.

Keeping the assessment reliable and accountable

Comparable definitions are essential. Before analysis, teams should align product identifiers, reporting periods, claim statuses, complaint categories and the treatment of fees and remuneration.

Missing or low-quality data should remain visible. An AI workflow should not silently infer values that will be used in a regulatory assessment. Any grouping of narrative information should be testable against representative examples, including less common customer circumstances.

Outputs should link back to the source data and explain the basis of comparisons. Access should also be restricted because the evidence may contain personal and commercially sensitive information.

The product-governance record should show the evidence considered, limitations identified, challenge applied, conclusions reached and actions agreed. AI can make that record easier to assemble, while accountability remains with the firms and individuals responsible for the assessment.

Example

A hypothetical insurer manufactures a retail insurance product distributed through several coverholders. Each coverholder supplies pricing, claims, complaint and service information in a different format.

An AI-supported workflow maps the returns to agreed definitions and classifies complaint narratives into review themes. It flags that one distribution segment has a higher proportion of claims-related complaints and longer settlement times, while also identifying missing data for one reporting period.

The product governance lead asks claims, conduct, actuarial and delegated authority specialists to validate the data and investigate the segment. The committee considers the explanation, records the evidence limitations and decides whether changes or further monitoring are needed.

FAQs

• Which delegated authority products are subject to FCA fair value rules?

  Scope depends on the product, jurisdiction and the roles performed by the firms in the distribution chain. Relevant teams should confirm the current FCA rules and any exclusions before designing the assessment.

• Can an AI model produce a fair value score?

  A model can calculate indicators, but a single score can hide important assumptions and evidence. Fair value requires a documented assessment of price, benefits, service and customer outcomes by accountable professionals.

• What data is useful for ongoing fair value monitoring?

  Relevant evidence can include product benefits and limitations, price, fees, remuneration, claims outcomes, complaints, cancellations, service performance and information about the target market.