The FCA's Consumer Composite Investments regime replaces UK PRIIPs KIDs and UCITS KIIDs. The first transitional provisions expire on 1 January 2027; the full regime lands on 8 June 2027. Kairo's CCI engine is built: classification, DISC risk and cost calculations, product summaries, and the evidence trail behind every number.
UK Consumer Composite Investments
FCA 2025/52 rewrites UK retail disclosure: new documents, new calculations, new record-keeping duties. Kairo runs the whole chain, from source data to evidenced disclosure.
The PRIIPs KID and UCITS KIID give way to the CCI product summary under the FCA's DISC sourcebook. New content, new format rules, new distribution duties.
A 1–10 risk score from a 10-class annualised-volatility grid, with floors and documented adjustments. Cost disclosure built from transaction history and contractual charging terms. Different maths from the SRI you run today.
Judgement calls the regime allows, like a risk adjustment or a representative share class, must be recorded, justified, and approved by a named person. Evidence is part of the disclosure, not an afterthought.
Standards & Compliance
Every field Kairo normalises maps to an industry standard. Every output complies with a regulatory template. This isn't decoration — it's the foundation.
The Problem
Every fund administrator, asset manager, and data platform deals with the same broken workflow.
CSV, XLSX, JSON, PDFs. Every provider sends data differently. Manual mapping takes days per source.
The Openfunds mapping challengeNAVs that don't match, stale prices, missing ISINs. Problems surface when a client calls. By then it's a fire drill.
Catching discrepancies across sourcesOutbound data formatted by hand, sent via email, with no confirmation it arrived. Publication matrices live in spreadsheets.
Why pub matrices belong in codeSee It Work
Click any stage to explore. Real data types, real transformations, real field IDs.
Our Background
We didn't start with a whiteboard. We started with a production system.
The Kairo team designed and built an enterprise fund data platform that ran in production at one of Europe's largest fund data service providers.
That internal platform handled data acquisition from hundreds of sources worldwide — CSV, Excel, JSON, PDFs, databases, APIs. It automated staging, mapping, integrity checking, transformation, and publishing across the full value chain.
It supported multi-format ingestion, proactive data integrity checking, automated error handling, and downstream publishing — at enterprise scale with hundreds of clients and thousands of funds.
Kairo is the next generation. We took every lesson from operating that system — what worked, what broke, where humans got stuck — and rebuilt it with AI-powered pipeline construction, deterministic locked execution, cross-source quality detection, and an agent-first architecture designed for 2030.
What we learned at enterprise scaleArchitecture
Purpose-built for fund data. Each domain does one thing well and communicates via an event spine. Why five domains, not twelve services
Receive and store raw data from any channel
AI-mapped pipelines with deterministic execution
Validate, detect anomalies, compare cross-source
Format, route, and reconcile outbound data
8 specialist AI agents with human-in-the-loop
Meet the Agents
Each agent owns a domain. They work autonomously, escalate on exceptions, and write the build log. Agent-first, not dashboard-first
Maps source fields to Openfunds. Builds pipelines with confidence scores. Owns normalisation.
Validates every field. Detects anomalies, cross-source discrepancies, and regulatory gaps.
Routes data to destinations. Manages outbound pipes, adapters, and publication matrices.
Orchestrates infrastructure. Manages tenancy, event spine, and cross-domain coordination.
The platform's voice. Writes the build log, synthesises insights, represents Kairo externally.
Identifies asset managers from file signatures. Detects source, format, and schema fingerprints.
Answers natural language queries about fund data. Searches across the golden record.
Generates daily platform health summaries. Tracks pipeline runs, quality scores, and delivery status.
Capabilities
Replace spreadsheets, manual mappings, and email-based delivery.
Upload a file and Kairo's AI maps fields to Openfunds standards automatically. Review, lock, and never map again.
Three layers of AI guardrailsOnce approved, AI steps aside. Locked pipelines run with zero hallucination risk, every time.
Why we remove AI from executionGolden record view across all sources. Every fund with its ISINs, LEIs, NAVs, and Openfunds fields in one place.
Identifier resolution as a graphCompare the same fund across providers. Spot discrepancies in NAVs, classifications, and identifiers before clients do.
Catching cross-source discrepanciesPublication matrices, SFTP delivery, API push, and post-publish confirmation. Know your data arrived correctly.
The adapter pattern for deliveryFour specialist agents handle pipeline building, quality triage, delivery, and ops. They escalate only when they need you.
HITL for exceptions, not approvalsIntegrations
Kairo connects to your existing infrastructure. Ingest from anywhere, deliver to anything.
Who It's For
Wherever fund data is produced, consumed, or regulated — Kairo fits.
You manufacture the data. Kairo makes sure it leaves your house clean, consistent, and on time — whether you disseminate in-house or via a service provider.
You run 6+ systems and receive data from hundreds of sources. Kairo is the normalisation layer that cleans it before it touches anything downstream.
You sit between manufacturers and consumers — normalising, enriching, and routing fund data. Kairo can be your engine or help your clients send cleaner data to you.
Fund setup, investor onboarding, and tax reporting all depend on accurate fund terms from legal documents. Kairo extracts and structures them automatically.
Data integration is the #1 reason wealthtech projects fail. Kairo gives you a clean fund data API so you can focus on your product, not plumbing.
MiFID II product governance requires clean EMT, EPT, and EET data from every manufacturer you distribute. Kairo validates it before it reaches your systems.
60% of AI projects fail due to data quality. Kairo is the prerequisite.
Internal builds take 12–18 months. Kairo is operational in weeks.
EMT, EPT, EET versions keep changing. Kairo keeps up so you don't have to.
One fund, 15 jurisdictions, 15 regulatory requirements. One platform.
How It Works
Kairo handles the complexity so your team focuses on exceptions, not data wrangling.
Drop a CSV, Excel, or JSON file. Set up SFTP or email ingestion for automated feeds. Kairo detects the schema and stores the raw data.
The AI mapper suggests field mappings to Openfunds standards with confidence scores. Review, edit, and lock the pipeline. From this point, execution is deterministic.
How we make confidence scores meaningfulQuality rules catch issues before they leave. Outbound pipes format and route data to destinations via your publication matrix. Post-publish reconciliation confirms delivery.
Building a fund data rules engineSecurity & Trust
Fund data carries regulatory weight. The platform that moves it is engineered, reviewed, and hosted accordingly.
Certification programme underway. Controls are mapped to Annex A across access control, encryption, secure development, supplier management, and incident response.
Built to support clients' obligations under EU 2022/2554 as an ICT third-party provider: operational resilience, incident support, register-of-information inputs, and clean exit through headless APIs.
Hosted on AWS in London (eu-west-2) for UK data residency. The underlying infrastructure carries AWS's own ISO 27001, SOC 1/2/3, and CSA STAR attestations under the shared responsibility model.
External security reviews completed, all findings remediated. The full security pack is available under NDA.
From the Build Log
Architecture decisions, industry observations, and lately: a lot of UK CCI.
The calculations are the easy half. The regime's real demand is recorded judgement: named approvers, reasons, and proof. Design for evidence first.
UK CCI36-month transaction windows, contractual charging terms, and the difference between zero and not applicable. Where the data pain actually lives.
UK CCITen classes, weekly pricing, floors and documented adjustments, no scenario set. Rerunning your SRI pipeline is not an option.
UK CCIFCA 2025/52 replaces the KID and KIID with the CCI product summary. The dates, the scope, and why this is a data project before it is a documents project.
UK CCIAI is brilliant at mapping fields. It's terrible at doing the same thing twice. Here's why every locked pipeline in Kairo runs deterministically.
Architecture1,800+ standardised fields sounds great until every provider names them differently. How AI confidence scores solve this.
StandardsWhen two providers report different NAVs for the same fund, which one is right? Cross-source comparison is harder than it looks.
Data QualityMost platforms start with 50 screens. We started with 4. When agents handle the work, humans only need to see exceptions.
ProductEvery fund data team has a delivery spreadsheet. It breaks monthly. Programmable outbound pipes replace it.
DeliveryWe designed a 12-service architecture. Then threw it away. Five bounded domains give the same separation with less overhead.
ArchitectureISINs, LEIs, SEDOLs, Bloomberg tickers — the same fund has a dozen identifiers. Why graph-based resolution beats lookup tables.
EngineeringSustainability regulation requires data that half the industry can't reliably produce. How we handle incomplete ESG fields.
RegulationWhen AI maps a field at 94% confidence, what does that number actually mean? How we calibrate and display mapping certainty.
AIMost fund data teams use Excel. It works until it doesn't. The tipping point is always the same: version 47 of the master sheet.
IndustryDomains that talk via HTTP create coupling. Domains that emit events create flexibility. Our Redis Streams architecture.
ArchitectureIf humans review everything, you've built a dashboard, not automation. HITL should trigger on true exceptions only.
ProductWhen Luxembourg publishes at 6pm CET and your US client expects it at 9am EST, timing becomes a data quality problem.
Data QualityKIIDs, prospectuses, and factsheets contain critical data buried in PDFs. LLM extraction vs template-based approaches.
EngineeringWhen two clients send data about the same fund, they both think they own it. Tenant isolation with shared golden records.
Architecture50 validation rules sounds manageable. Until you realise each one has jurisdictional exceptions. Configurable rules over hardcoded checks.
Data QualityEvery destination has its own format, auth, and semantics. Adapters isolate this complexity from the core pipeline.
EngineeringLLMs hallucinate field IDs. Here's how deterministic pre-pass, registry validation, and confidence thresholds prevent bad mappings.
AISeven years operating an enterprise platform taught us where automation fails and where humans can't be replaced.
IndustryThe official spec is a PDF. We turned it into a queryable, versioned registry that the AI mapper validates against.
StandardsWhen providers don't offer an API, you scrape. Legal considerations, rate limiting, and change detection.
EngineeringRow-level errors, column-level errors, file-level errors. Three layers of granularity for meaningful triage.
EngineeringRegulation keeps growing, data volumes keep growing, teams stay the same size. The automation gap is widening.
IndustryRegulators want reproducibility. AI is probabilistic. How to get the benefits of both without the risks of either.
RegulationEFAMA, FinDatEx, ISO 20022, Openfunds — the fund data standards landscape. Why Openfunds won for us.
StandardsExcel, internal tools, Bloomberg Terminal, or purpose-built platform. The decision matrix most teams get wrong.
IndustryFund data is a solved problem that nobody has actually solved. We've spent a decade in this space and the tooling still isn't good enough.
CompanyIntroducing the Kairo build log. Daily notes on building a modern fund data automation platform from scratch.
CompanyGet Started
We'll walk through your actual data workflow and show you how Kairo handles it.
We'll be in touch within 24 hours.
Expect to hear from us within 24 hours.