Sovereign credit analysis in plain questions: every answer sentence beside its original-language page, the day a budget or IMF release lands
Every budget and IMF release read in its own language, cited to the page the day it lands.
See one case, screen by screen ↓Sovereign credit analysis, also called sovereign risk analysis, judges whether a government can keep servicing its debt, from its budget, fiscal rule, debt path and financing plans, as published in budgets, IMF Article IV reports and central bank and debt office documents. Sovereign Credit Analysis lets an EM sovereign credit research team ask that question in plain words, such as “Will Calvera meet its 2027 fiscal rule?”, and seven agents answer it from those documents in their own language. Each sentence sits beside the original-language quote, its English and the figure, cited to the page, with every translation back-checked. The analyst decides what goes into the country file, and a house-view change needs the senior sovereign strategist.
AI in asset management: a buyer's guide to agentic AI for investment managers →
Updated
Sovereign credit analysis on emerging markets starts in the original language. An analyst covering emerging market sovereign debt follows 15 to 25 countries whose budgets, medium-term fiscal frameworks, debt office bulletins and central bank reports are published in Spanish, Portuguese, French, Arabic, Turkish, Romanian, Russian or Bahasa Indonesia, and run 80 to 300 pages. Today they are read with a browser translation, or through a sell-side or rating-agency English summary that arrives days to weeks later, and only for the larger countries. The IMF's Article IV staff report, in English, comes months later.
The errors sit in the re-keying and the comparing. A decimal comma, % of GDP against % of revenue, nominal against structural balance, central against general government: each can flip a fiscal-rule verdict. The government's numbers are compared with the IMF's for the headline question only, so the gaps stay unexplained. Then the house country view changes by email, sometimes without a second person, and at audit nobody can find the page it rested on.
Estimated split for one fiscal framework release on a Tier 1 country, by hand and with the solution, in working days.
Seven agents read, verify and reconcile every release in its own language; the analyst decides, the strategist approves.
At Quillfield Capital, Inés Calderón covers 25 countries in 9 languages for the EM sovereign research team. On Wednesday, October 14, 2026, at 5:58 p.m., the Republic of Calvera's Finance Ministry publishes its Marco Fiscal de Mediano Plazo 2027–2030: 96 pages, in Spanish. Here is what happens next, screen by screen, in the working solution.
The Release Watcher watches 312 official publishers in 25 countries. It fetches the new fiscal framework, finds it supersedes the April 28 update and lists the five figures that moved, each with its page: the 2027 deficit target from 3.0% to 3.1% of GDP, copper from $9,900 to $10,400 a ton, growth from 3.0% to 3.2%, a new one-off revenue of 0.2% of GDP from the Litio Andino sale, and 2027 gross debt from 53.1% to 53.9%. Then it re-runs Calvera's three standing questions.
“Answer changed: ‘meets’ on the ministry's numbers, ‘likely miss’ on the house deck.”
Inés's Today screen opens on four numbers: 41 new releases this week, 17 not in English; 3 answers to review; 12 figure gaps explained, 1 unexplained; 1 view change waiting. Below, what needs her, one next step each, and what the agents did overnight, every line cited to its page. At the bottom, a box to ask the agents about any country, in any language.
“Explained 1.27 of 1.30 points between the ministry's 3.1% and the IMF's 4.4%: copper, growth and the Litio Andino sale.”
She types the question in plain words. The agents state their plan in one line, then run it step by step in view: the Release Watcher gathers 6 documents, 4 in Spanish and 2 in English; the Original-Language Reader reads 352 pages and takes 23 figures with page, unit and year; the Translation Verifier back-translates 23 passages, 22 match and 1 is held for her. The Reconciler explains the gap, and the Fiscal Rule Tester tests article 4 of Calvera's fiscal responsibility law, Ley 21.418, three ways. Her decisions stay locked until the run is done.
“Plan: gather 6 sources → read each in its own language → back-check every quote → explain the gaps → test Ley 21.418 three ways → write, then hand to you.”
The bottom line comes first: likely miss. Calvera meets its 3.5% ceiling only on the ministry's copper price; on the IMF's numbers it misses by 0.9 points, on Quillfield's house deck by 0.6. A small chart puts the ministry's path against the IMF projection, the house deck and the rule ceiling. Below, 8 sentences with 13 citations, each beside the original-language quote, its English and the figure it uses, with a confidence per sentence.
“Breakeven copper on the ministry's own growth and accounting: $9,830 a ton ($10,330 on ours). No breach of the 60% debt anchor.”
Inés clicks “MFMP 2027–2030 · p. 34”. The page opens with the passage highlighted in the original and in the English, and the checks on the figure: the back-translation matches, the decimal comma “3,1” is read as 3.1, the unit and perimeter are % of GDP for the central government, and the year is the 2027 projection. The one passage the Verifier held, “ingresos tributarios” on page 42, waits for her to confirm: tax revenue, as the Verifier reads it, or revenue, as the Reader wrote it.
“Read by the Original-Language Reader v5.1 and checked by the Translation Verifier v2.4 on Oct 14 at 6:15 p.m. The page image and its hash are kept for 7 years.”
The reconciliation lays the 2027 deficit out by source: the ministry's 3.1%, the rule's 3.5%, the IMF's 4.4%, the house deck's estimated 4.1% and the superseded April 3.0%. The bridge from 3.1% to 4.4% is built from the sources' own numbers: +0.77 for copper ($1,100 a ton at 0.07% of GDP per $100, the ministry's own sensitivity), +0.30 for growth, and +0.20 because the ministry counts the Litio Andino stake sale as revenue while the IMF books it as financing under GFSM 2014. On the house copper price of $9,500 and growth of 2.9%, the deficit is 4.1%, a miss of 0.6 points.
The Reconciler never picks a “right” number; it explains and cites both sides. A residual above 0.1 point is marked unexplained.
She asks in the same thread, beside the answer. The Fiscal Rule Tester replies with the arithmetic and the page: on the ministry's own growth path and method the breakeven is about $9,830 a ton; on the house growth of 2.9%, with the asset sale booked as financing, about $10,330. Today's house deck has $9,500. Every reply in the thread cites its page and offers the next step, such as moving the copper price.
“These are estimates.”
Inés saves the answer to Calvera's file, her choice, and opens the View Drafter's memo: Neutral to Underweight, fiscal outlook stable to negative, 11 citations, with what would change the house's mind (copper above about $10,330 a ton, or new measures worth about 0.6% of GDP). The maker and checker steps are on the right: she proposes, Dr. Henrik Aaltonen decides.
“This memo changes the house view only; it does not recommend trades. PMs decide positions.”
Henrik reads the memo, its citations and the holdings, and approves with the reason code “Fiscal rule miss on house deck” and a note: house deck copper is $900 below the budget; the miss holds unless copper rallies. At 10:42 the note is published to the research management system, and six PMs and the CIO Fixed Income are notified, because the holdings are above $250 million.
“Client commentary that mentions this view must be pre-cleared by compliance (SEC Marketing Rule).”
Marcus opens Calvera's file: now Underweight, fiscal outlook negative. Quillfield holds $684.2 million in 14 funds, from portfolio accounting at the October 14 close. By bond, CALVER 5.375% 2031 is $212.4 million at a Z-spread of 318 bp, 14 bp wider since the release; by fund, his Quillfield Emerging Markets Debt Fund is 3.4% Calvera against a 2.9% benchmark weight. The position is his decision.
The View Drafter never places or suggests orders, and never contacts clients.
The same Calvera budget, seen by five of the people who carry it: what the week looked like, and what it looks like now.
Seven agents read, verify and reconcile every release in its own language; the analyst decides, the strategist approves.
Watches 312 official publishers in 25 countries, from finance ministries, budget offices, parliaments, debt offices and central banks to statistics offices and the IMF; fetches, versions and diffs each release and re-runs the country's standing questions.
Reads each document in its own language, with OCR for scans, and extracts figures with page, table, unit, year and basis; translates the passages it quotes.
Back-translates every quoted passage and checks numbers, decimal commas, units, negations and perimeter, central against general government.
Compares the same figure across the government, the IMF, the central bank, the debt office and the house deck, and explains each gap from the sources' own sensitivity tables and GFSM 2014 classification.
Encodes each country's fiscal rules from the law text, cited, and tests them on the government's, the IMF's and the house's assumptions; computes breakevens.
Writes the answer sentence by sentence, each beside its original-language quote, its English and its figure, with a confidence per sentence.
Drafts the house-view change memo with its evidence and the holdings it touches; after approval publishes the note to the research management system and notifies the PMs.
Maker and checker on every view change. The agents propose; a named person decides.
Ask in the thread beside the answer, or from the Today box: the agent that holds the answer replies, and every reply points to its page.
Why does the IMF see Calvera's 2027 deficit at 4.4%?
Three things, each on the record. Copper: the budget uses $10,400 a ton, the IMF $9,300, worth 0.77 points on the ministry's own sensitivity. Growth: 3.2% against 2.6%, 0.30. The Litio Andino sale: revenue for the ministry, financing for the IMF, 0.20. 1.27 of 1.30 points explained.
What copper price meets the rule?
On the ministry's own growth path and method the breakeven is about $9,830 a ton. On the house growth of 2.9%, with the sale booked as financing, it is about $10,330. Today's house deck has $9,500. These are estimates.
Who holds Calvera?
$684.2M in 14 funds (portfolio accounting, Oct 14 close). Largest: Quillfield Emerging Markets Debt Fund, $238.5M, 3.4% of the fund against a 2.9% benchmark weight. By bond: 2031 USD $212.4M, local BTC 2030 $206.6M, 2036 USD $168.9M, 2033 EUR $96.3M.
Compare with the April framework
Five figures moved since the April 28 update: deficit target 3.0% → 3.1%, copper $9,900 → $10,400, growth 3.0% → 3.2%, a new one-off sale worth 0.2% of GDP, and 2027 debt 53.1% → 53.9%.
13 screens from the working solution, on its sample data. Pick one to see it large.
Where the agents stop for a person is a setting. Every change waits for a second administrator and is on the audit trail.
| Setting | Default | Choose from |
|---|---|---|
| Sentence confidence | 0.85 | Below it, a sentence shows as “check” and cannot be saved until the analyst accepts it |
| Scan (OCR) confidence | 0.80 | Below it, figures from a scanned page wait for a person |
| Reconciler residual tolerance | 0.1 point of GDP | A larger gap is marked unexplained and sent to the analyst |
| SLA, Tier 1 / Tier 2 | 1 / 3 business days | From publication to a reviewed, cited answer; Tier 1 = holdings above $250M |
| Escalate to the strategist after | 4 hours past SLA | Past it, an open item goes to the senior sovereign strategist |
| When the View Drafter drafts a change | Fiscal-rule miss on the house deck · unexplained gap on a rule figure | Agency action on a Tier 1 country: off by default, because agency actions are read but a house view is ours |
| Every view change: maker and checker | Always | Cannot be switched off |
| Primary region and retention | US-East (New York) · 7 years | US-East (New York) · EU-West (Dublin); London read replica, read only |
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 data · “target” = the design goal, measured in the live solution · “estimated” = our estimate (where an analyst's days go after a budget lands) · “typical” = published figures (Aguilar Pérez, Gatekeepers of the Market? IMF Surveillance and Sovereign Access to Global Finance, EconomiX Working Paper 2025-32: Article IV statements narrow EMBI sovereign spreads by about 15 bp on average, 67 emerging markets, 2004–2023; CERDI research seminar “The SuperIMFeconomist” (IMF), December 16, 2025: Article IV assessments for nearly 190 member countries; Article 8a of the EU CRA Regulation, as amended by Regulation (EU) No 462/2013: unsolicited sovereign rating actions on a calendar of at most three dates a year). Rules named as context: SEC Rule 204-2 under the Advisers Act (books and records) and the SEC Marketing Rule. Quillfield Capital and the people, countries, laws, documents, funds, bonds, identifiers and figures named on this page are fictional: characters and sample data in the working solution.
Sovereign credit analysis, also called sovereign risk analysis, judges whether a government can keep servicing its debt, from its budget, fiscal rule, debt path and financing plans, as published in budgets, IMF Article IV reports and central bank and debt office documents. Sovereign Credit Analysis lets an EM sovereign credit research team ask that question in plain words, such as “Will Calvera meet its 2027 fiscal rule?”, and seven agents answer it from those documents in their own language. Each sentence sits beside the original-language quote, its English and the figure, cited to the page, with every translation back-checked. The analyst decides what goes into the country file, and a house-view change needs the senior sovereign strategist.
Sovereign debt analysis turns on why the government's numbers and the IMF's differ. The Reconciler lays the same figure side by side from the government, the IMF Article IV staff report, the central bank, the debt office and the house deck, and builds a bridge between them from the sources' own sensitivity tables. In the working solution, 1.27 of the 1.30-point gap between Calvera's 3.1% and the IMF's 4.4% is explained by copper, growth and an asset sale the IMF books as financing under GFSM 2014. The Fiscal Rule Tester then tests the rule on each set of numbers.
The Original-Language Reader reads each document in its own language and quotes the passage it uses; the Translation Verifier back-translates every quote and checks numbers, decimal commas, units and perimeter before the sentence can appear. A mismatch holds the sentence for the analyst, and figures from a scanned page read below 0.80 confidence wait for a person. The working solution reads Spanish, Portuguese, French, Arabic, English, Turkish, Romanian, Russian and Bahasa Indonesia.
The Fiscal Rule Tester encodes each rule from the law text, cited, and tests it on the government's, the IMF's and the house's assumptions, with breakevens. It uses only the economist's approved house deck, and its results are labeled estimated. In the demo, Calvera meets its 3.5% ceiling on the ministry's numbers, misses by 0.9 points on the IMF's and by 0.6 on the house deck.
The Release Watcher fetches it from the official publisher, diffs it against the last version and re-runs the country's standing questions, so the analyst starts with the answer and what changed. The same holds for a central bank decision or the IMF's press release concluding an Article IV consultation. Each release carries an SLA, one business day for Tier 1 countries and three for Tier 2. A failed fetch, a restated report or a moved page goes to a person with one next step.
Yes. Name any of the 25 countries in the analyst's book in plain words, in any spelling, such as Itaporã or Itapora, and the agents answer with the latest cited, original-language figure beside each sentence, so country risk analysis no longer stops at the countries with an English summary. If a question names no country, the agents ask which one; a restricted country is refused for anyone who is not wall-crossed.
No. The analyst accepts, edits or rejects each sentence with a reason and decides what is saved. A house-view change is drafted by the View Drafter, proposed by the analyst and approved or rejected by the senior sovereign strategist with a reason code, never by the proposer. The agents never trade, never contact clients and never change a view; PMs decide positions.
The Agentic Solution Engine builds and deploys it from your requirements and documents (your countries and languages, publishers, fiscal rules, house deck and systems of record), and it goes live once every quality gate has passed. A four-step first-time setup covers countries, languages and sources, people and roles, and the systems it connects to.
Every credit question answered in minutes, each sentence pinned to its tranche and page.
1-min demoBond Covenant AnalysisEvery basket in dollars and every leakage route found, before books close.
1-min demoProspectus ReviewEvery deal that fits a fund briefed and decided before books close, every line cited.
1-min demoWe'll run Sovereign Credit Analysis on your own countries' budgets, IMF reports and central bank releases.