AI-Powered Investment Intelligence
Where to invest.
And when.
Kilwa is the intelligence layer for Africa's next wave of investment. We turn live macroeconomic data, policy signals and multilingual news sentiment across all 54 African markets into two answers institutions can act on.
East Africa · Q2 2026
ISI score & METI signal
Scores from the published East Africa Investment Intelligence Brief, Q2 2026. Illustrative of platform output; live scores update as inputs move.
Read the brief- African markets covered
- 54African markets covered
- indicators per country score
- 90indicators per country score
- sources across 85 languages
- 416sources across 85 languages
- entry windows forecast
- 1–6moentry windows forecast
Partners & collaborators
Votee AIMongoDBElevenLabsThe assessment gap
Africa is not assessed the way it is invested in.
The continent pays a measurable premium to borrow, and a large part of it is attributed not to what the data says but to how the data is read. Whether that gap is bias or prudence is genuinely disputed. What is not disputed is the remedy.
$75bn
the estimated cost to African sovereigns of subjective assessment — $28bn in excess interest, $46bn in financing never accessed
UNDP, 2023
200–400bp
the spread African eurobonds have carried above similarly rated emerging markets elsewhere. The premium survives after the rating is controlled for
Brookings
4 of 55
African nations holding investment grade from the three agencies that rate over 95% of the world's debt
Moody's, S&P, Fitch
0.5–1
notches by which one econometric study found African sovereigns underrated. Moody's disputes it, citing 40 years of default alignment
KAS-Leibniz; Moody's
Chatham House, arguing against the African Union's answer of building a rival rating agency, named a different one: the best solution is a more dedicated effort to present reliable data, and talk to the market.
Why Kilwa exists
To close the evidence gap, market by market, in public. Kilwa is not a rating agency and does not grade sovereigns. It scores all 54 African markets on a published method, flags every input as verified or estimated, states what would prove each call wrong, and grades itself against that record where anyone can check.
We make no claim to be the largest provider of country risk data. We are not. What we do claim is testable: the method is published, the estimates are marked, and the misses are on the record with the hits.
Why Kilwa is different
Frontier data breaks the standard tools. We rebuilt them.
Most country-risk providers apply a global model to Africa and translate the news into English before reading it. Both choices discard exactly the information a frontier allocation turns on.
Conventional approach
One global risk model, applied to Africa as a region.
Kilwa
Built for thin data, not adapted to it.
Standard regression fails where series are sparse and non-stationary. ISI is a hierarchical gradient-boosted ensemble trained on clusters of comparable markets, so Ghana borrows statistical power from Nigeria instead of scoring poorly for want of coverage.
Conventional approach
Translate the news into English, then read the sentiment.
Kilwa
Read in the language it was published in.
Transformer models fine-tuned with triplet loss on a purpose-built corpus of African financial news, so ‘pression inflationniste’ sits beside ‘inflationary pressure’ in vector space. Translation is where policy nuance dies, and policy nuance is the signal.
Conventional approach
Tell you where the risk is. Leave the timing to you.
Kilwa
Two answers, not one.
Suitability and timing are modelled separately and published together. METI injects the sentiment delta into a time-series forecast as an exogenous variable, with the weighting learned rather than assumed, and returns a one-to-six month window.
Conventional approach
Coverage that thins out exactly when a market gets interesting.
Kilwa
Scored when the standard sources go quiet.
The IMF withholds projections for sovereigns in restructuring — Ethiopia and Zambia among them in its April 2026 outlook — which is precisely when an allocator needs a view. Kilwa scores them anyway, flags every estimate as an estimate, and says what would change the number.
Conventional approach
Judgements about the continent formed at distance from it.
Kilwa
Built on the ground, not adapted from abroad.
The African Union’s objection to how the continent is assessed is that the assessors hold no meaningful presence in the region. Kilwa’s modelling was built with the Northwestern MSAI programme and its language and delivery work with partners in-market — Digital Umuganda, QT Software, Rwanda’s Ministry of ICT.
Conventional approach
A composite score, with the workings held back.
Kilwa
Every number shows its provenance.
SHAP attribution per indicator, model cards naming known weaknesses, and a verified-or-estimate flag on every input — 42% of them flagged in the last flagship. Every pre-registered call is graded in public, misses included.
The modelling was built in collaboration with the Northwestern University MSAI programme. Every method is documented on the methodology page.
The two questions
Suitability without timing is half an answer.
Most research tells you which markets look attractive. None of it tells you which quarter to move. Kilwa runs both models simultaneously, across every market we cover.
Kilwa Research
Ten flagships. One scored continent.
Growth, tail risk, currency, exits, resource nationalism, regulation and energy, scored across all 54 African markets — every input provenance-flagged, every model published, and every pre-registered call in Africa Signal Check graded in public, misses included.
Flagship reports are sold individually from $1,500, with team and enterprise licences available. The Exit Door is free for a work email, and regional briefs and Africa Signal Check are published open.
The Kilwa Intelligence Platform
Six modules. One decision.
Every module feeds the same answer. Nothing here is a separate dashboard you have to reconcile.
Investment Suitability Index
A dynamic scoring engine that ingests structured economic data — GDP, FDI flows, inflation, governance and regulatory indicators — to generate a live investment readiness score for any country and sector.
Learn more 02Market Entry Timing Index
A predictive forecasting engine that fuses time-series analysis with NLP-driven sentiment tracking from news and policy statements to identify optimal entry windows.
Learn more 03Multilingual Sentiment Pulse
Trained on region-specific media to analyse risk, tone, policy shifts and leadership signals in French, Arabic and Swahili — capturing local context that English-only platforms miss entirely.
Learn more 04Zawadi AI
Ask a question in plain English and get an answer about the market — grounded in the same signals, with the same decomposition, as every score on the platform.
Learn more 05Explainable AI
Every score comes with an explanation. SHAP-based attribution and published model cards show exactly which factors drive a result — the transparency institutional risk committees require before capital moves.
Learn moreAsk anything
“Which of the 54 markets carry unhedgeable currency exposure, and why?”
Zawadi answers in plain language from Kilwa’s published scores, and every figure traces back to the input that produced it.
Meet ZawadiBuilt for
Different mandates. The same two questions.
Kilwa serves the institutions allocating capital into frontier markets — and the governments building the capability to attract it.
Sovereign Wealth Funds
Diversification intelligence that identifies high-potential markets beyond the traditional safe havens.
Multinational Corporations
Data-driven market expansion strategy that replaces analyst gut-feel with scored, explainable intelligence.
Governments & Public Institutions
The evidence layer a finance ministry, debt office or investment promotion agency needs to be read accurately by the market — and sovereign AI built to be owned locally.
Why Kilwa
Four things that make the score usable.
An unexplained number is not usable in an institutional process, however accurate it turns out to be.
Predictive, not descriptive
Competitors describe what happened. ISI scores where capital should go and METI forecasts when it should move — the two questions that actually gate a deployment decision.
Africa-first and multilingual
Purpose-built for data-scarce, multilingual, high-volatility markets. English, French, Arabic and Swahili processed natively, not translated into English first.
Synthesised and explainable
Not a data feed. Every score decomposes into the indicators that drove it, with provenance flags on every input and robustness testing published in full.
Built for institutional process
AES-256 encryption, role-based access control, and controls aligned to SOC 2 Type II — because the score has to survive a risk committee, not just impress an analyst.
Kilwa Sovereign AI
National-scale AI, built to be owned locally.
Our government vertical delivers sovereign AI capability — including language model work in national languages — built with the institutions that will own and operate it. Systems delivered without local capacity become dependencies, and dependencies expire.
Trust & transparency
Every score shows its workings.
We describe our scores as structured risk rankings, not validated predictive models, wherever that is the honest characterisation. Model cards document scope, training data, known weaknesses and what would invalidate the result.
1,000
draw Monte Carlo robustness tests published per flagship score
100%
of model inputs flagged verified or estimate
AES-256
encryption in transit and at rest
SOC 2
Type II controls, aligned
Request a demo
See the markets you follow, scored.
We'll walk your team through ISI and METI across our covered markets — and show you the decomposition behind every score.