Second Chair

AI Advocate Opinion

Two sides. All the law. One clear opinion.

Legal AI for insurance claims handlers - balanced liability analysis grounded in South African case law.

Built for claims

Clear counsel when liability is contested

Second Chair sits beside your claims team as an AI advocate - reading the case pack, testing both sides, and delivering a reasoned opinion with an indicative liability split.

Case pack to facts

Claim forms, statements, and supporting documents are extracted into a structured, neutral factual record before any opinion is formed.

Liability pattern guidance

Collision patterns are matched to a curated South African motor liability catalogue - indicative settlements, reasons, and seed authorities.

Handler-ready opinions

Polished opinions with issue analysis, verified authorities, and an insured / other-party liability split your team can review and act on.

Deep Dive

Opinion the way careful counsel works

Facts before law. Law before advocacy. Advocacy before judgment - with a red-team audit before delivery.

01

Extract the facts

Build a neutral record of agreed and disputed facts from the claim pack - without deciding liability.

02

Frame the legal issues

Identify the questions that must be answered: relative fault, duties of care, contributory negligence, and more.

03

Research & verify authorities

Ground the analysis in South African case law, then check every citation so unreliable authorities never reach the opinion.

04

Hear both sides

Parallel advocate briefs for the insured and the other party surface the strongest arguments before judgment.

05

Independent opinion & audit

Form a reasoned finding with an indicative split, then red-team the draft and revise when defects remain.

Capabilities

Everything a claims desk needs from an AI second chair

Two-sided advocacy

Adversarial testing for Party A (Insured) and Party B (Other party) so conclusions are stress-tested, not one-sided.

Case-law grounded

Opinions lean on South African authorities and precedent - not model memory dressed up as law.

Authority verification

Each candidate citation is checked before advocacy and opinion drafting. Unverified or rejected sources stay out.

Red-team quality gate

A separate audit stage challenges facts, law, and reasoning - and can force limited revision loops.

Document intelligence

Structured claim capture, document assessment, and quantum signals keep the case file ready for analysis.

Organisation-ready platform

Secure multi-user access for insurer teams - cases, opinions, feedback, and operational controls in one place.

Two modes

Match depth to the matter

Handlers choose the path that fits urgency and complexity - without leaving the case.

Quick take

One-shot opinion

A focused liability view when you need a fast, structured reading of the pack - still informed by pattern guidance and the case facts.

Second chair

Deep Dive harness

The full multi-stage process: facts, issues, research, verification, dual advocacy, independent judgment, and red-team audit.

The Bench

When the other insurer is ready to be heard

Deep Dive gives each claims team its own opinion. When both insurers agree, they link those case packs and opinions. Second Chair then reviews the combined record — facts, merit, and documents — attacks its own reading, and returns a determination both sides can use.

Insurer A

Case pack and opinion Facts, documents, and the opinion that team already holds.

Insurer B

Case pack and opinion The other file, kept separate until that organisation accepts the link.
Red team

Determination

One result, both teams The apportionment, the reasons, and what each file established.
01

Link by consent

Each insurer keeps its own case. Nothing crosses until both organisations accept the link. Either side can decline.

02

Read the joint record

Second Chair reviews both factual records, both opinions, and the underlying documents. Disagreements between the files stay visible.

03

Attack the reading

A separate red team challenges missing documents, one-sided weighting, weak authorities, and any apportionment the papers do not support.

04

Return a determination

Both teams receive the same reasoned result: the apportionment, the findings each side made out, and what remains for the handlers to decide.

The determination is for both claims teams to adopt. It does not close either claim, and it is not an award.

Brand pillars

Designed for trust in high-stakes claims

Balanced

Considers both sides with fairness before stating an indicative apportionment.

Second Chair

Your AI advocate, by your side - supporting handlers, not replacing professional judgment.

AI Powered

Advanced reasoning for stronger, more consistent outcomes across the claims desk.

Focus

Purpose-built for South African motor liability

From junctions and lane changes to rear-end chains and special conditions - Second Chair is tuned to the patterns claims teams see every day.

Both sides Insured and other party argued before judgment
Verified law Authorities checked before they are relied on
SA-trained Engine schooled on South African cases across every motor accident category
Human review Opinions support - never replace - legal judgment

For administrators

What we mean by AI contamination

AI contamination is claim material ending up somewhere your firm did not choose.

Your people already use AI of their own. A handler can paste a claim into an account that belongs to them, not to the organisation. Your firm does not choose the controls on that account, cannot see what was sent, and has no record of it. The material can be kept, copied, or reused. That leaves the organisation, and the people named in the claim, open to abuse of that data.

Second Chair is the opinion path your organisation provides. It writes an independent AI legal opinion for a handler to review. It is not legal advice and it does not decide the claim. Your administrator chooses the controls below. They apply to work done in Second Chair.

What your administrator can set

A person decides

The opinion does not close the case. A person records the verdict and what happened. Your firm can require a rating, a liability split, or both, before the case can close.

A published judgment

An opinion will not be generated without authoritative judgments cited from our corpus and training data.

Personal details

We strip all personal information before sending any information to an AI model.

How long the opinion is kept

Your firm chooses how long the opinion and the text from the documents are kept after the case is closed.

An activity record

An administrator can see and download this month's record of documents read and opinions requested. It stays after the opinion material is removed.

The model service

Document reading and opinions go to a model service. The material is not used as training data. We also use services that prevent any misuse of your firm's data.

Put a second chair beside every handler

See how balanced, law-grounded opinions can tighten liability decisions across your claims organisation.