The Institutions AI Built
On AI, institutional leverage, and where the real business opportunity sits — for practitioners, not engineers
I want to be upfront about something before this essay gets going.
I use AI. Heavily. Not as a novelty, and not in the way most people mean when they say it — the occasional query, the summarised email, the shortcut. I mean structurally. It is load-bearing in how I work. And because I think that context matters for everything that follows, I’ll tell you what that actually looks like before we get to the broader argument.
But first, the argument.
The conversation most people are having about AI is, at its core, a conversation about replacement. Which jobs survive. Which industries get hollowed out. Whether your particular skill set clears the bar. It’s an understandable conversation, and it isn’t entirely wrong — some roles will be compressed, some cost structures will collapse, some business models that depended on information asymmetry will not survive contact with a world where that asymmetry has been significantly eroded.
But it’s the wrong conversation for anyone trying to locate where the real opportunity sits.
The more consequential question — the one that keeps getting crowded out by the displacement narrative — is this: what has always been too expensive to have, and isn’t anymore?
Because AI is not primarily a story about automation. It is a story about the cost of professional capability, and what happens to markets when that cost drops by an order of magnitude.
Professional capability is a specific thing. It’s the kind of output that used to require a team, a firm, a retainer, a budget line: legal analysis, financial modelling, strategic intelligence, market research, quality writing at volume. The price of accessing that capability has historically been set by the markets where it was most abundant — London, New York, Johannesburg at a stretch — and calibrated for the budgets of the clients those markets served.
Which meant that for most of the world, and for most businesses operating outside those centres, serious professional support was either inaccessible or intermittently accessible at best.
This is where the AI story gets structurally interesting, and where most of the commentary misses the point. The people writing about AI’s economic impact are largely writing from inside markets that already had deep practitioner depth — established law firms, functioning capital markets, accessible advisory capacity across sectors. For them, AI is an efficiency story. Faster research. Cheaper first drafts. Reduced headcount at the margins.
For markets where that practitioner depth was thin or absent, AI is something categorically different. It makes viable what was previously unaffordable.
Anyone who has spent time in professional services on this continent will recognise a specific kind of conversation. It usually starts well. An SME owner — real business, real revenue, real complexity — comes in needing something substantive. A shareholder agreement before a new investor arrives. A financial model that can survive scrutiny. A market entry analysis for a new geography. They understand what they need. The conversation is good.
Then the fee comes up.
And the conversation changes register entirely. Not because the business owner is being unreasonable, and not because they don’t understand the value. The math simply doesn’t work for them. Professional services pricing was built for clients whose budgets were built for it. An SME doing $200k in annual revenue cannot absorb a $15,000 legal engagement, regardless of how necessary the underlying work is.
So they leave without the shareholder agreement. They build the financial model themselves, in a spreadsheet, with assumptions that won’t survive a serious investor’s due diligence. They make the market entry call on instinct and relationships, which sometimes works and often doesn’t.
That gap — between what African businesses need and what they could historically afford — is not a gap in ambition or in understanding. It is a pricing problem. And AI has changed that pricing equation. For good.
The way I think about where the opportunity sits breaks across three distinct layers.
The first is Capability Compression. AI has reduced the cost of producing previously expensive outputs — legal documents, financial models, strategic memos, market analysis, long-form content — to something approaching marginal zero. The opportunity here is direct: any service business that can now deliver ten times the volume with the same team, or the same volume with a fraction of the team, captures the margin that used to go to overhead. Goldman Sachs estimated in 2023 that generative AI could expose roughly 44% of legal work tasks to automation — not replacement, but compression. That cost curve has moved. It will not move back.
The second layer is where it gets more interesting: Domain Judgment Monetisation. Here’s what AI still can’t do: determine whether the output is actually right. Commercially, legally, strategically right. That judgment took years to build and it can’t be prompt-engineered — unless the person prompting already has the expertise to know what right looks like, to catch what’s wrong, and to push further where the output is merely adequate. Which means the expertise was always the product. AI just made it faster to deploy. A lawyer who understands how Nigerian courts actually interpret boilerplate force majeure clauses brings something no model currently replicates with confidence. A strategist who has watched enough African fund structures unwind under liquidity pressure knows where the real risk sits in ways that aren’t legible from training data alone.
But there’s a third thing AI does for practitioners that doesn’t get enough attention: it helps articulate ideas that were fully formed in someone’s head but never quite made it to the page. Most experienced practitioners carry frameworks, intuitions, and analytical models that took years to develop — and that sat unused, because translating complex thought into clean communicable form takes time most practitioners simply don’t have. AI compresses that translation cost significantly. The idea was always there. What was missing was the bandwidth to surface it properly.
The third layer is where this becomes specifically and acutely relevant for Africa: Institutional Gap Arbitrage. In markets where the capability always existed but the economics of accessing it never worked for most businesses, AI doesn’t just improve existing processes. It makes what was always present actually reachable. Contract review from practitioners who were always qualified. Just never affordable. Governance frameworks from counsel whose engagement models were built for clients three tiers up — not for the founder doing $200k who needed them just as badly. Credit underwriting that extends what skilled practitioners can deliver beyond the narrow band of businesses that could historically pay for it. The opportunity isn’t to replicate what Western professional services firms do. It’s to make what African practitioners already know how to do viable for the businesses that always needed it most.
I said at the start I’d tell you what this looks like in practice. Here is the honest version.
The L.U.M.I. Brief produces analysis — legal, strategic, financial — across a publication, a cross-platform content system, and client advisory work. The breadth is real: UK and Nigerian legal training, venture strategy, capital markets, governance architecture, creative economy IP. That multi-disciplinary depth took years to build. It is the actual product. AI did not build it and cannot replicate it.
What AI changed is the cost of deploying it.
Before AI became a serious operational tool in this workflow, the same quality of output took significantly longer to produce. A long-form essay that now moves from structured thinking to publish-ready draft in hours previously consumed most of a working day, sometimes more. A cross-platform content cycle that now runs in one concentrated session previously stretched across most of a week. The analytical work — the judgment calls, the structural decisions, the actual thinking — that part took the same time regardless. But the production work surrounding it, the drafting, structuring, formatting, adapting across platforms — that was extracting hours that the thinking itself doesn’t require.
AI removed that tax. It didn’t make the analysis sharper. It stopped making the deployment slow.
And then the inverse, which the optimists tend to skip: a generalist with no deep domain expertise and access to the same tools does not produce the same output. The tools are identical. The results are not. What AI amplifies, it amplifies in proportion to the expertise already present. The practitioners who built genuine depth — in law, in finance, in strategy, in specific markets — are the ones for whom AI creates the largest delta. Not because they needed help thinking. Because they no longer have to be slow. And because the ideas they’d been carrying but never quite had the bandwidth to fully articulate now have a vehicle.
The reframe I’d leave you with is this.
AI’s biggest economic opportunity was never going to be captured by the people who built the tools. It will be captured by practitioners — lawyers, strategists, analysts, writers, advisors — who understood early that AI had permanently restructured the cost of deploying expertise, and who moved into the spaces where that expertise was most needed and least accessible.
For Africa, that means the opportunity is not to build the next AI company. It is to become the practitioner class that AI makes economically viable — in markets that have been waiting for exactly that depth of capability for decades. The talent was always here. The judgment was always here. The ideas were always here. What was missing was the economic model that made deploying all three, consistently and at scale, actually feasible.
That model now exists.
This week’s premium note maps the specific opportunity zones — by sector, by market, and by business model archetype — with the financial logic of what each one actually looks like to build. If you’re advising, allocating, or building in this space, that’s the piece.