Every PRIIPs KID, in every language, proven against the fund data before it can publish
Every figure in every language proven against the fund data before a KID can publish.
See one case, screen by screen ↓PRIIPs KID Production drafts and checks the PRIIPs KID, the key information document, for every share class in the language of every market where it is sold, and drafts the UK CCI Product Summary from the same data. Each KID is built from the fund data template, translated and back-translated, and every figure in every language, from the summary risk indicator and performance scenarios to the cost disclosure, is proven against the master before release. Manufacturers must publish a KID for each packaged product sold to retail investors in the EU, and for every UCITS share class since January 1, 2023. A named reviewer approves each language and the Head of Product Compliance signs each share class.
AI in asset management: a buyer's guide to agentic AI for investment managers →
Updated
Every PRIIPs KID has to say the same thing in every language. A share class sold to retail investors in the EU needs a key information document in the language of each market where it is registered, and UCITS have needed one since January 1, 2023. Quillfield Capital's European range runs to 612 share classes in 27 markets: 4,212 live documents, reissued after every fee, risk or prospectus change and at least every 12 months. The figures come from the fund administrator's European PRIIPs Template (EPT) file, a template engine builds the English, translators and a language vendor produce the rest, and a few reviewers spot-check a sample of PDFs.
The risk sits in the pages nobody reads back. Translation memory carries last February's transaction costs into this year's sentence. A vendor's post-edit swaps two digits, so the Polish version prints ongoing costs of 1,12% while the master says 1.21%. A locale breaks the thousands separator, or a risk indicator is copied from last year's version. The data was checked going in; the document going out was not. An inaccurate KID is a breach: member states must allow fines of at least €5 million or 3% of turnover, an investor who relied on it can claim damages, and distributors delist a share class whose document is missing or out of date.
Estimated split for a full reissue of a range this size, by hand and with the solution. With the solution the agents read every version back; people review, approve and sign.
Eight agents draft, translate, back-check and prove every figure; a reviewer per language approves and compliance signs.
Quillfield Global Funds SICAV's prospectus update takes effect on Monday, November 2: ongoing costs fall for 31 sub-funds after the September 24 fee review, and every document carries a new complaints address. All 4,212 versions are reissued for publication on Friday, October 30. Here is Monday, October 19, screen by screen, in the working solution.
Nadia's home shows the reissue on one line, from the prospectus approved on October 9 to the effective date, with 9 business days left to publication. 3,212 of 4,212 versions are signed and scheduled, 47 are blocked by a mismatch, 388 are in language review (12 past the 2-day service level) and 6 risk second opinions disagree with the calculation provider. Each item under “Needs you” has one next step: see the mismatch, regenerate 18, nudge reviewers, open fund data, retry the Nordic portal.
Overnight, the Numeric consistency agent: “Read 4,212 rendered versions back and checked every figure against its master and the data template. 47 mismatches in 47 share classes.”
Jonas Weber accepted the fund administrator's EPT v2.1 file at 6:05 p.m. on October 13: 612 records, 598 complete. The next day the Drafting agent built 598 English masters and held 14 for incomplete records. For Euro High Yield, Class A EUR Acc, the record explorer shows each field beside the previous KID: row 118, ongoing costs, now 1.21% against 1.30% before the fee review. The master's 22 key figures each point to their row, and it fits on three sides of A4.
“The fund data template (EPT v2.1) from the fund administrator: one record per share class, every figure in a KID comes from a row here.”
Checks groups the 47 mismatches by cause: stale figure 18, number format 11, transposed digits 9, meaning drift 5, risk indicator 4. In 18 versions the transaction-cost sentence still matches the February 2026 KID, because translation memory offered the old sentence. Nadia regenerates all 18 in one step, and the run shows itself: the plan, the sentence rebuilt from row 121, each language rewritten with its locked phrases, 18 pages re-rendered and read back — 18 of 18 match.
“Handed to 6 reviewers for approval · nothing publishes until each approves.”
Quillfield Euro High Yield UCITS, Class A EUR Acc (LU2471836598) opens as a wall of its 28 versions in 23 languages. Each page carries its key figures in the market's own format: SRI 3, ongoing costs “1,21 %”, the moderate scenario at four years “11.520 €” in Germany and “11'520 €” in Switzerland. One tile is red. The release gate across the top reads 27 of 28, release blocked, and the Sign share class button stays locked.
Back-translation checker: “Back-translated all 27: 1,107 sentence pairs, every one above the 0.92 meaning threshold (lowest 0.94, Finnish). The Polish cost sentence means what the master means; only its number differs.”
Nadia asks the agents beside the wall. The Polish version prints ongoing costs of 1,12%; the master and the data template say 1.21%. The amount beside it, 117 EUR, still matches 1.21%, so the percentage is the error: two digits transposed in the language vendor's post-edit, job TR-88412. The other 61 numbers match. Each claim links to row 118, the master page and the Polish page.
The compare view lines up every sentence: the English master, the Polish as printed and a translation back to English on a second model, each number tagged with its data row. The checks strip reads numbers 61 of 62, meaning 41 of 41 sentences at or above 0.92, locked phrases 12 of 12, three sides of A4, paying agent and complaints text present. The Translation agent adds that its own draft of October 14 read 1,21% and proposes setting the figure back. Nadia cannot approve it.
“Language approvals belong to the Polish reviewer.”
Marta reviews the Polish versions for 146 share classes a cycle. She sees the one line that changes — 1,12% to 1,21% wartości Twojej inwestycji rocznie — with EPT row 118, master page 2 and the vendor's post-edit as its sources. She approves with reason code L2, Approved after correction, and signs through single sign-on.
“Matches the master and EPT row 118. The agent re-checks the whole version after you approve.”
The correction is applied on page 2, the Polish PDF is re-rendered on three sides of A4, all 62 numbers are read back against the master and row 118, the meaning is re-checked at 0.98, and the wall updates: 28 of 28 versions match. Marta's decision, its reason code and her signature are on the record.
With every version matched, the Sign share class button opens for Robert. The sign-off shows what he is signing: 28 of 28 versions match the master, 28 approved by their language reviewers, 1,736 numbers proven against the data template. He signs with S1, All versions match and reviewed. The Release agent schedules the share class for Friday, October 30, 6:00 a.m. CET, to six channels, from the fund documents site and distributor portals to the UK feed and the archive.
“Signing does not let you edit content.”
The Risk and scenario agent recomputed the risk indicator, scenarios and costs for all 612 share classes from NAV history and cost data: 606 agree with the calculation provider. For Emerging Markets Local Debt, Class A EUR Acc, it finds a VaR-equivalent volatility of 12.3% against 11.8% supplied, because the provider used weekly returns for January 2022 when nine daily NAVs were missing. That crosses the 12% line from risk class 3 to 4; accepting class 4 would reissue 41 versions across both of its share classes. Claire Fontaine, product governance, decides.
“Product governance decides; the agent never overrides the provider.”
The same reissue, seen by five of the people who carry it — what the cycle looked like, and what it looks like now.
Eight agents draft, translate, back-check and prove every figure; a reviewer per language approves and compliance signs.
Builds each English master KID from the EPT record, the product master and the prospectus, and checks it against the template and three sides of A4.
Writes each market version from the master with the approved term base, locked regulatory phrases and the market's mandatory text, and routes it to the language vendor where policy requires a human post-edit.
Translates every version back to English on a second model from a different provider and compares the meaning with the master, sentence by sentence.
Reads every rendered version, extracts every number, normalizes the market's number format and proves each figure against the master and its EPT row; it runs the release gate.
Recomputes the VaR-equivalent volatility, market risk class, performance scenarios and costs over time from NAV history and cost data, and runs the Article 15(2) checks.
Drafts the UK Product Summary and the Core Information file from the same data, and explains each difference from the EU KID.
Watches prospectus updates, board fee resolutions, EPT differences, new registrations and the 12-month clock, and proposes a reissue plan with a reason for every version.
Schedules and publishes signed share classes to the fund documents site, distributor portals, data vendors and feeds, verifies every published file's fingerprint and retries failures.
Approve, sign. The agents propose; a named person decides.
Ask in plain words, or tell the agents what to do. Every answer is cited, and the decision stays with a person.
Why is Euro High Yield, Class A EUR Acc blocked?
The Polish version prints ongoing costs of 1,12%; the master and the data template say 1.21% (row 118). The amount beside it, 117 EUR, still matches 1.21%, so two digits were transposed in the language vendor's post-edit (TR-88412). Marta Zielińska decides.
Which versions still show last February's transaction costs?
18 versions in 6 languages: German 4, French 3, Italian 3, Spanish 3, Dutch 3 and Swedish 2. Translation memory carried the February sentence; the data template has the October figure (row 121). The production manager can regenerate all 18 from Checks.
Is the Emerging Markets Local Debt risk class changing?
Possibly. The recomputed VaR-equivalent volatility is 12.3% against 11.8% supplied; 12% is the line between class 3 and 4 (Annex II). Claire Fontaine decides.
What is left before the October 30 release?
3,212 of 4,212 versions are signed and scheduled. 392 wait for compliance sign-off, 388 are in language review (12 overdue), 47 are blocked by a mismatch, 96 are being re-translated and 77 wait for data.
13 screens from the working solution, on its sample data. Pick one to see it large.
The thresholds every version is checked against are versioned settings: the administrator proposes a change, product governance approves it, and the agents are re-evaluated before it applies.
| Setting | Default | Choose from |
|---|---|---|
| Numeric tolerance | 0 | Every figure must equal the master after the locale's number format is applied |
| Meaning threshold | 0.92 | Back-translation similarity per sentence; below it, the language's reviewer decides |
| Agent confidence | 90% | Below this, the agent hands the item to a person |
| Length | 3 sides of A4 | Every version |
| Risk recompute tolerance | ± 0.3 points | Volatility difference, or any change of risk class |
| Moderate scenario check | 5 points | Mean return moved since the last KID |
| Vendor post-edit | 9 markets | Markets where a human linguist edits the agent's draft |
| Language review service level | 2 business days | Escalation after 1 day overdue; flagged lines shown first |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample fund range · “target” = the design goal, to be measured in a pilot · “estimated” = our estimate · “typical” = published rules (Regulation (EU) No 1286/2014, Articles 5, 7, 11 and 24 — fines of at least €5,000,000 or 3% of turnover; Delegated Regulation (EU) 2017/653, Article 15 — review at least every 12 months; FCA PS25/20, Consumer Composite Investments, December 8, 2025). Quillfield Capital and every fund, ISIN, vendor and person named on this page are fictional — characters and sample data in the working solution.
PRIIPs KID Production drafts and checks the PRIIPs KID, the key information document, for every share class in the language of every market where it is sold, and drafts the UK CCI Product Summary from the same data. Each KID is built from the fund data template, translated and back-translated, and every figure in every language, from the summary risk indicator and performance scenarios to the cost disclosure, is proven against the master before release. Manufacturers must publish a KID for each packaged product sold to retail investors in the EU, and for every UCITS share class since January 1, 2023. A named reviewer approves each language and the Head of Product Compliance signs each share class.
In the EU, UCITS used the UCITS KIID until their exemption from PRIIPs ended on December 31, 2022; since January 1, 2023 they need a PRIIPs KID, with its 1 to 7 risk indicator, four performance scenarios and cost tables. In the UK, the UK PRIIPs KID and the UCITS KIID both give way to the CCI Product Summary by June 8, 2027.
At least every 12 months, and whenever a change significantly affects it, including when the moderate scenario's mean return has moved by more than five percentage points (Delegated Regulation 2017/653, Article 15). The Change monitor reads prospectus updates, fee resolutions, EPT differences, new registrations and the 12-month clock, and proposes which versions to reissue and why. Product governance approves the plan before the agents act.
Two ways. The Back-translation checker translates each version back to English on a second model and compares its meaning with the master sentence by sentence, against a 0.92 threshold. The Numeric consistency agent reads every rendered page, normalizes the market's number format and proves each figure against the master and its data row. A version that disagrees blocks the share class until its reviewer decides.
Yes, as a second opinion. The Risk and scenario agent recomputes the VaR-equivalent volatility, market risk class, scenarios and costs over time from NAV history and cost data, and runs the Article 15(2) checks. It never overrides the calculation provider: in the sample cycle 606 of 612 share classes agree, and the six that differ go to product governance with the working.
In the UK, the Consumer Composite Investments (CCI) Product Summary replaces the UK PRIIPs KID and the UCITS KIID: manufacturers may use it from April 6, 2026 and must from June 8, 2027. The UK summary agent drafts the Product Summary and the Core Information file from the same data as the EU KID: a risk and return score of 1 to 10 from 10 years of volatility, the ongoing charge as the headline in percent and pounds, and explicit transaction costs only. It explains why, for example, the UK score is 4 of 10 while the EU indicator is 3 of 7. Compliance approves it as a financial promotion.
Yes. The agents draft, check and propose; they never approve, sign, publish or clear their own flags. Each language's reviewer approves it with a reason code, product governance decides risk-class changes and reissue plans, the Head of Product Compliance signs each share class, and nothing with an open mismatch can be signed or released.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your KID procedure, your data template feed, your registration matrix and a sample of published KIDs — and it goes live once every quality gate has passed. We will walk you through it on your own share classes first.
Every 15(c) answer cited and checked against last year, ready 10 days before the deadline.
1-min demoFee BenchmarkingEvery peer fee your board sees, cited to the peer’s own fee table and signed.
1-min demoFund Launch ManagementFirst NAV on the planned date: every slip caught early, re-planned and signed off.
1-min demoWe'll run PRIIPs KID Production on a sample of your own data template records and published KIDs.