Project Kilimanjaro, explained.

Project Kilimanjaro is AIA’s Africa-first decision-intelligence operation — built on governed infrastructure, so every number we put in front of a board is defensible.

The reframe that makes it work: a model is a function, and its inputs are not the model’s private property — they are firm-level shared knowledge. When two AIA models running “the same scenario” disagree, the firm is voting against itself. So there is one governed source; every model reads from it. That single change is what makes stewardship, audit and reproducibility possible at all.

Four pillars. Every artifact is testable against them.

Each pillar maps to a specific gap in the African decision-support market. If a piece of Project Kilimanjaro work doesn’t speak to at least one, it’s drifting.

The problem it kills.

Today every model team carries its own copy of the inputs. They are usually close, rarely identical — and nobody can prove which number fed which board paper.

If a model needs an input the firm has a view on, it reads it from one governed place — versioned in git, owned by a named steward, stamped with the commit that produced it. That is the whole idea.

What we sell — four offerings, one funnel.

Intelligence widens the funnel; Research deepens a specific issue; Data & Modelling provides technical certainty; Decision Support takes it to the boardroom. A single client can buy one, two, all four, or the combination across cycles — same firm, four different buyers, four different budgets.

The buyer decision map.

The questions buyers actually pay to answer — mapped to the assets that answer them and the offer that sells the answer. This is the bridge from "what we sell" to "what we've built".

Mirrors narrative/insights_source_map.md — the single source of structure.

The infrastructure underneath.

Two productized assets sit under all four offerings. Neither alone is sufficient; together they are what makes the offerings productized rather than bespoke — and they are invisible to the client, who only ever sees the deck, the model and the data feed.

Mozambique gas monetisation.

The clearest proof the operation is real and not aspirational: the registry already governs the inputs underneath delivered, partner-approved paid work for a Mozambique gas value-chain client — not a separate product the client buys, the operating layer beneath the engagement. Here is the whole motion in four steps.

Transnet & TNPA corridor economics.

The second flagship, in a completely different value chain — the proof the substrate isn’t a one-domain trick. The authority-published tariff schedules that decide whether a corridor pays are governed in the registry and read by the corridor financial model, with the integration verified end-to-end against production. The Boegoebaai S9 corridor (Northern Cape) is the showcase.

The ExxonMobil gas mass balance.

The first two flagships prove the data layer — the registry governing inputs underneath delivery. This one proves the method layer: a live South African gas mass-balance engagement for ExxonMobil, run on the frontier where the registry doesn’t reach yet. Most of its entities — Mozambique upstream, LNG terminals, gas pipelines, demand zones — don’t exist in the registry today, so registry-first wasn’t possible. The Modelling SOP held anyway. That is the point.

Rail manufacturing — the next CoE taking shape. Emerging

Build-in-public means showing the work before it’s finished. Rail manufacturing isn’t registry-wired end-to-end yet — but it already carries two live engagements that show both legs of the offer: the model and the strategy. The rolling_stock domain is the substrate being stood up to tie them together.

Pull the governed data into your tools.

One source, every tool. Copy-paste and you have a live, governed dataset in Excel or a notebook in under a minute — no re-keying, no stale copies. This is the data layer, in your hands.

let
    Source = Csv.Document(
        Web.Contents("https://www.africaia.com/insights/assumptions/series.csv?key=macro.gdp.timeseries_index"),
        [Delimiter=",", Encoding=65001]),
    Table = Table.PromoteHeaders(Source)
in
    Table

import pandas as pd
df = pd.read_csv("https://www.africaia.com/insights/assumptions/series.csv?key=macro.gdp.timeseries_index")
df.plot(x="period", y="value")

# or the governed API (scenario-aware):
# requests.get("https://master-assumptions.digitalnyika.com"
#   "/v1/series/macro/gdp.timeseries_index?scenario=medium",
#   headers={"X-API-Key": "<your-key>"})

What is live now.

This is the board the wider team watches the operation take shape on. It reflects today, not the pitch — edited as each piece of wiring lands.

As of 21 June 2026 — updated as wiring lands.

Portfolio health — auto-scanned

Live from scripts/build_estate.py — the engineering reality behind the board above: what's versioned and backed up.

At risk: team_recommender, aia_power_tools, gtm-operating-system, swasap_growth_strategy, diverzify_freight_analysis — not backed up to a remote. (Fix: git init + push to the AIA org.)

Open honestly — the gaps are the opportunity.

Today the bench is the wider AIA analytical practice; the dedicated seats are mostly open. We’d rather show you the real org than a finished one.

We grade on one thing — producing the weave — on a five-band ladder from Analyst to Partner, the same spine for everyone. See where you’d land, and what the next band asks of you. The door in is a conversation with Nigel.

Where to next.

“Bullish on the destination; ruthless on the evidence.”