How agentic AI is changing the infrastructure investment thesis in emerging markets
March 24, 2026
The volume of capital chasing AI as an investment theme has made it harder, not easier, to think clearly about where the genuine value will accrete. The narrative has collapsed into a small number of familiar positions: the foundation model providers, the hyperscale compute infrastructure, and the enterprise software layer being rebuilt around large language models. Those positions may be valid. They are also among the most contested, most expensively priced, and most geographically concentrated bets in the history of institutional investment.
There is a different question worth asking, one that is less visible precisely because it requires looking at markets most institutional investors are not watching closely. In high-growth economies, where financial infrastructure is being built from the ground up, where supply chains are still being formalised, and where digital commerce is expanding faster than in any mature market, agentic AI is not arriving as a productivity overlay on top of existing systems. It is becoming embedded in the foundational layer of how those systems operate. That distinction changes the investment thesis materially.
What agentic AI actually means in this context
The term needs to be defined carefully, because it is used loosely enough to be nearly meaningless in most investment conversations.
Agentic AI refers to systems capable of taking sequential, goal-directed actions autonomously, rather than simply generating a response to a prompt. An agentic system can receive an objective, break it into tasks, execute those tasks using available tools and data, adjust based on what it observes, and complete the objective without requiring human instruction at each step. The distinction from earlier AI applications is not one of sophistication alone. It is one of operational scope.
An agentic system does not assist a process. It runs one.
In the context of financial infrastructure and supply chain platforms in emerging markets, this matters in a very specific way. The operational challenge these platforms face is not primarily a technology challenge. It is an information and decision density challenge. Underwriting a loan to an agricultural operator or farming enterprise requires assessing dozens of variables across crop type, geography, weather exposure, market access, and repayment history, often without a formal credit file. Verifying inventory in a commodities supply chain requires continuous monitoring across multiple physical locations with imperfect documentation. Managing a portfolio of thousands of small business loans across multiple jurisdictions requires constant surveillance of exposure, concentration, and early warning signals.
These are tasks that, until recently, required either a large human workforce operating at significant cost, or a simplified rules-based approach that accepted meaningful accuracy loss. Agentic AI changes that equation. The same operational processes that previously required twenty analysts can be run continuously, at higher accuracy, with a system that costs a fraction of the headcount. For platforms already operating at the infrastructure layer of these economies, that is not a marginal improvement. It is a structural change in what the unit economics of the business can look like at scale.
Where the value actually accretes
The investment implication is not straightforward, and it is worth being precise about where value is created and where it is not.
Agentic AI does not create value uniformly across all technology businesses. It creates the most durable value for platforms that already hold a structural position in the data and transaction flows of a market, because those platforms have the raw material that agentic systems require to function at their best. A lending platform that has processed ten thousand loans across a specific agricultural market has accumulated the repayment data, seasonal patterns, and borrower behaviour profiles that make an agentic underwriting system materially more accurate than a generic model applied to the same problem. A supply chain platform that has managed commodity flows across a regional market has the logistics and pricing data that makes an agentic inventory and verification system far more reliable than one operating without that context.
The implication for investment selection is this: the businesses that compound most aggressively as agentic AI becomes operational are the ones that already have proprietary data positions and structural market roles. Agentic AI amplifies the value of what they already hold. It does not create value for businesses without it.
In mature markets, those proprietary data positions are hard to come by because the incumbents are large, established, and well-capitalised. In emerging markets, the infrastructure layer is still being built. The platforms that achieve early structural positions in financial infrastructure, agricultural data, and digital commerce flows are accumulating the data assets that will determine which businesses compound and which plateau over the next decade. That is the investment thesis. Agentic AI is the mechanism that converts those data positions into operational leverage.
The specific applications worth watching
Three areas within Kingson's investment themes are particularly affected by this shift, and understanding the specific mechanism in each case makes the thesis more concrete.
Credit underwriting and portfolio surveillance. The highest-cost, highest-risk operational function in any lending platform is credit assessment. In emerging markets, where formal data is scarce and borrower profiles are non-standard, that cost and risk are amplified. Agentic systems capable of synthesising supply chain data, satellite imagery, mobile transaction history, and market price feeds into a continuous credit assessment change the economics of underwriting in markets where no bank has been able to operate profitably at small ticket sizes. The platforms that integrate this capability earliest will have a cost structure and loss ratio that is genuinely difficult to match. The competitive pressure runs in two directions. New entrants cannot easily build a proprietary data position from scratch. It accumulates through years of actual market operations, not through capital deployment alone. Large incumbent lenders face a different but equally constraining problem: they hold significant transaction history but it sits in legacy systems that are too siloed and too poorly structured for agentic deployment. Those incumbents hold data as an asset in name only. Platforms built natively on structured, machine-readable operational data hold it as a working capability. That is the distinction that compounds over time, and it is why the window for building a durable position at this infrastructure layer is narrower than it appears.
Supply chain verification and commodity finance. Post-harvest loss across sub-Saharan Africa exceeds 30% of total crop production, representing more than $4 billion in value annually according to FAO data cited by the World Food Programme, with the FAO and World Bank's own "Missing Food" report putting the broader supply chain loss figure at approximately 37% of total agricultural output. The primary driver is not poor farming practice. It is the inability to monitor, verify, and finance inventory as it moves through the supply chain. Agricultural operators and commodity producers lose product at every stage between harvest and market because the data infrastructure to track, verify, and finance that movement in real time does not exist at scale. Agentic systems with access to IoT sensor data, logistics milestones, and commodity pricing can run continuous verification of physical inventory as collateral, enabling commodity finance to operate at a speed and scale that manual processes cannot support. For the infrastructure platforms building these capabilities, commodity finance becomes a new revenue layer attached to the supply chain position they already hold.
Regulatory compliance and reporting. The compliance burden for financial platforms operating across multiple emerging market jurisdictions is substantial and growing. AML screening, KYC verification, transaction monitoring, and cross-border reporting requirements create operational overhead that scales poorly with headcount. Agentic compliance systems that run continuously across transaction flows, flag exceptions for human review, and generate regulatory reports automatically are already reducing compliance costs for platforms that have deployed them. In markets where regulatory complexity is a genuine barrier to entry for smaller operators, this capability advantage is also a competitive moat.
The risk of getting this wrong
The clearest risk in this thesis is category error: investing in AI capability rather than in the infrastructure businesses that AI capability serves.
A business built around an AI product, without a structural data position or a genuine market role that existed before the AI, is a different kind of investment from a financial infrastructure platform or a supply chain operator that is integrating agentic capability into an established operational foundation. The first is a technology bet with AI market competition as its primary risk. The second is an infrastructure bet with AI as the operational amplifier.
The distinction matters more in emerging markets than in mature ones, because the infrastructure layer in high-growth economies is still being established. The businesses that achieve structural positions now, before the market has consolidated, will have a compounding advantage that AI amplifies rather than creates. Investing in the AI layer above that infrastructure, rather than the infrastructure itself, is to mistake the tool for the position.
The second risk is timing. Agentic AI capability is real and operational today in specific applications, but the timeline for widespread deployment across the operational complexity of emerging market infrastructure is not uniform. Investors who underwrite this thesis need to be realistic about the build time required to integrate agentic systems into the operational fabric of businesses that are simultaneously managing growth, regulatory relationships, and market expansion. The value is not speculative. The path to it requires patience.
What this means for how we look at new investments
Kingson's view on AI has always been applied rather than abstract. We are drawn to founders using AI and agentic systems to solve infrastructure problems, not to founders building AI for its own sake. That distinction has sharpened considerably over the last twelve months as agentic capability has moved from research into production.
What we look for now in a financial infrastructure or supply chain business is not simply whether AI is present in the product. It is whether the business is accumulating the kind of proprietary data position that makes agentic systems substantially more valuable than they would be without it. That accumulation is happening in specific markets, at specific infrastructure layers, faster than most institutional investors are tracking. The gap between where the value is forming and where institutional capital is looking is, in our view, where the investment opportunity sits.
Kingson publishes perspectives to help allocators think more carefully about capital deployment in high-growth markets. This piece represents our considered view, not investment advice. For allocator enquiries, contact advisory@kingsoncapital.com