Insurance runs on documents: policy wordings, schedules, quotes, renewal terms, commission statements. Every placement, every renewal and every record in a broker's system starts with someone reading them. AI that gives a different answer each time it reads the same document cannot be adopted for regulated work, however good the demo looks.

Cluda is deterministic end to end: the pipeline that reads the documents, the broker workflows it feeds, the extraction that turns documents into structured data, and the policy administration system that holds the record. Policy analysis for employee-benefits brokers is where it starts.

The £2M seed takes Cluda from that first use case to the rest of the broker's work, and to leadership of the category in employee benefits.

Liability

A general model asked the same question twice can give two different answers. In insurance a hallucinated clause is an E&O claim: a missed exclusion is an uncovered client and a liable broker, and missed coverage gaps are a leading cause of broker E&O claims.

Accuracy

Before AI can compare, advise or fill in a record, it has to extract the data from policy PDFs: benefits, limits, exclusions and terms spread across tables, footnotes and endorsements. Extraction that is right most of the time is not usable at all, because every figure then has to be checked by hand. It has to be 100% accurate, and general models are not.

Evidence

The FCA's Consumer Duty requires firms to demonstrate, not merely claim, good outcomes, with fair-value analysis done at the time rather than reconstructed later. That needs a page reference behind every decision, and a chat window cannot give one.

So adoption has stalled. Under 5% of brokers use purpose-built AI; the rest paste client data into ChatGPT or Copilot, with zero traceability. The broker's systems (Acturis, SSP, Open GI) store and track policies. They don't read them. Marsh, with the capital and engineers to do it, built its own internal comparison tool. The other 400,000 brokers can't build their own.

Policy analysis

From fact find through quotes and analysis to the market review report, policy analysis is the centre of every placement and renewal. A typical renewal is 6 to 8 hours: getting the quotes, analysing the wordings, building the report.

Data entry

Once the business is placed, what the documents say is keyed again into the policy administration system: the scheme, the benefits, the members, the premiums, the renewal dates, every joiner and leaver. The same data, typed by hand, one more time.

Reconciliation

In the brokerage's finance team, commission statements arrive from each insurer in its own format and are checked against the policies and invoices behind them. Many documents, one answer that has to tie out.

Premiums are down around 20% across most UK commercial lines: brokers must do more with less. Each of these jobs can be handed to AI only if the answer is the same every time and traces to the page it came from.

Intelligence layer

Reads and understands the documents. Determinism comes from the ingestion pipeline, not from the model: every document is parsed, structured and checked the same way every time, so the same document always gives the same answer, with the page it came from.

Workflows

The broker's jobs: analysis, placement, renewal. Each is a chain of decisions a broker already makes; Cluda makes each one the same way every time and records why.

Processing layer

Deterministic extraction of structured data from any insurance document (benefits, limits, exclusions, terms, schedules), inside Cluda and, in phase 2, API access for broker PAS and insurer systems.

System of record

The employee-benefits policy administration system, where the record lives and stays consistent with the documents behind it, and where the finance team reconciles commission statements against the policies. Incorporate Benefits is the proof build, switching over in November 2026. Once a broker's records and Consumer Duty evidence run through Cluda, switching is structural, not a matter of convenience.

Determinism has to run through all four for AI to be trusted with the work. A deterministic model feeding an unchecked workflow, or clean extraction landing in a record nobody can trace, breaks the chain. That is why 40% of the raise goes to proprietary model development: the evaluation set and the deterministic pipeline first, then, over time, an insurance-trained model that beats general LLMs on that set at a lower cost per document.

The comparison runs on 70,260 product configurations and 8.27 million data points from nine insurer files, each traced to its source page.

Faster

Quote comparison from 30 minutes to 3: around 90% of the analysis time, measured in our broker ROI model. The saving is the by-product; the depth is the point.

Trust channel

AMII chose Cluda for its members. The Association of Medical Insurance Intermediaries is the UK trade body for the brokers who place private medical insurance. Its 106 member firms roll out from November 2026, with the PMI teams at five of the top ten UK brokers in the first wave, several already asking about additional licences. An ex-broker CEO and account manager onboard every firm in person.

The pattern is Harvey's in law, where trust is the product. Harvey won the hardest buyer first: Allen & Overy's 3,500 lawyers asked 40,000 questions before a firm-wide rollout, its first 50 enterprise customers came by referral, and it reached $75M ARR within two years of launch. Insurance has the same shape and nobody has done it. Our trust channel is a trade body, and the hardest firms go first.

£79

Policy analysis. The comparison, six states and a page reference behind every cell. Takes the scan and the Excel sheet off the broker.

£149

Placement and renewal workflow. The analysis inside the whole job, from fact find through quotes to the market review report. Takes the renewal off the broker, not just the comparison.

£249

Policy administration system. The system of record with the comparison built in: data extracted from the documents lands in the record instead of being keyed in, and commission statements are checked against the policies behind them. Takes the data entry off the broker and the reconciliation off the finance team. From May 2027.

Prices are per user per month. The ladder extends across the whole employee-benefits package as each product line comes on in 2027: cash plan, life, dental and wellness alongside private medical insurance, so a broker compares, places and administers the package in one place. A firm starts where its need is and moves up as Cluda takes over more of the work; a firm with no policy administration system starts at the top.

The target: 500 active users at a blended £200 per user per month, £1.2M ARR by the end of 2027.

Bottom-up. The category is deterministic AI for employee-benefits brokers. The wedge is the UK's specialist employee-benefits and PMI intermediaries, an estimated several hundred firms. AMII's 106 member firms are the reachable core, with five of the top ten UK brokers in the first wave.

AMII is the commercial test case for broker networks. After the UK come the global employee-benefits networks, WBN and Brokerslink among them, then UK broker networks beyond AMII, where early-stage discussions are under way with Bravo Networks (450 broker firms) and Hedron Network for commercial lines. Growth inside each network is product-led: a broker who runs one real placement through Cluda does not go back to the scan. The aim is market leader of the category before widening to other lines.

Signed

AMII, £60k, a one-year contract, signed and paid, within a three-year exclusive partnership as AMII’s only AI tool for policy analysis; 106 firms rolling out from November. Incorporate Benefits, £60k, signed: an AI-native employee-benefits policy administration system, switching over on 27 November 2026.

Backed by

SFC Capital led the £320k pre-seed in March 2025. An existing investor came in on a £75k allocation in April 2026.

Targets

First paid seats come with the November rollout. In writing: 106 firms rolling out from November 2026, first paying platform users in Q4 2026, 500 active users and £1.2M ARR by the end of 2027.

The full milestone list, with what each needs and the capital line behind it, is in the roadmap document (section 04).

Meet the team: full bios →

How the team has changed, and the hiring plan, are in the team document (section 02).

An incremental ladder: a firm starts where its need is and moves up as the product takes over more of the workflow. The £1.2M target is 500 active users at a blended £200 across the three rungs. The time-saved basis is in the metrics document (section 05).

The £2M seed funds the plan in one round, and each hire in the model is tagged to the milestone it serves. Salaries move to market rate at close, which removes the retention risk of a team paid below market today.

12 months

The 106-firm AMII rollout from November 2026 · the employee-benefits policy administration system delivered 27 November 2026 · the whole benefits package comparable · first insurance fine-tune with published accuracy and determinism figures · first enterprise licence · ISO 27001.

Documents

The 24-month forecast from close in three funding cases, use of funds by capital line, the team by month and the filed accounts are in the financial plan (section 07).

AMII

The Association of Medical Insurance Intermediaries, the UK trade body for health and protection intermediaries: 106 member firms including five of the top ten UK brokers. A one-year contract, £60,000, signed and paid, within a three-year exclusive partnership: Cluda is AMII’s only AI tool for policy analysis. Members compared insurer wordings by hand on every case; AMII wanted one consistent, auditable comparison for the whole membership and chose Cluda to provide it. Member rollout from November 2026.

Incorporate Benefits

UK employee-benefits consultancy. A £60,000 contract, signed, to build an AI-native employee-benefits policy administration system with Cluda, delivered to switchover on 27 November 2026. Their administration ran on spreadsheets and re-keying; they wanted a system that reads the policy documents and keeps the record consistent with them, built by a partner who understands broking. Our phase 3 proof build.

What each customer bought and why is in the customers document (section 06).

History

Pre-seed £320k led by SFC Capital, March 2025. A £75k allocation from an existing investor in April 2026. The company is Simpli Automation Ltd, trading as Cluda.ai, registered in Northern Ireland (NI714368), founded April 2024.

Structure

Advance assurance for the seed shares is in progress with our accountants. The Articles, shareholders' agreement, latest confirmation statement and the shareholdings as registered are in the legal document (section 08).

Book a call with Mick30 minutes, any chapter.