How AI Is Changing the Way Impact Investment Handles Information

The conversation about impact investing has historically begun with capital. There is good reason for that. The United Nations estimates that countries around the world require roughly $4 trillion annual investment to achieve the Sustainable Development Goals by 2030, while the Global Impact Investing Network estimates that the impact investing market now manages more than $1.5 trillion in assets. These two numbers are often presented as evidence of a financing gap, but they also point to a different challenge. Even when capital is available, investors still need to decide where it can create the greatest impact, and doing so depends on information as much as finance.

Impact investing has always had an information problem

Traditional finance benefits from a relatively common language based on revenue, margins, and cash flows that can be compared across firms with reasonable consistency. Impact investing operates differently in some ways. A healthcare project can report patients reached, an agricultural investment can focus on reporting water savings, while an education initiative measures school attendance. Many organizations also report under different frameworks such as IRIS+, the Global Reporting Initiative, or IFC Performance Standards.

These frameworks were built for different purposes, where IRIS+ was designed to give investors a shared taxonomy for outcomes, GRI was designed for corporate sustainability disclosure aimed at a broad set of stakeholders, and IFC's Performance Standards were designed to manage environmental and social risk in project finance. None of them set out to make one investment comparable to another. A metric like 'water saved' can mean withdrawal avoided, consumption reduced, or water recycled, each measured against a different baseline. Comparing investments across these frameworks means figuring out which definition sits behind each number, and that takes time and human judgment, not just access to the data.

For years, the industry's response was to improve measurement. Common taxonomies, reporting frameworks, and verification standards made impact data more credible, comparable and transparent. However, some have also highlighted that complete standardization is neither realistic nor always desirable. Especially when political and geographical contexts are very different, where for example a job created in an informal economy does not always mean the same thing as one created in a highly formal labor market.

AI changes how investors work with information

This is where AI enters the impact investing environment too, as a tool that helps work across frameworks rather than just replacing them. Large language models and other machine learning tools can process thousands of pages of reports, classify similar concepts expressed in different ways, and summarize unstructured documentation that previously required days of manual review.

One institutional example is the International Finance Corporation's MALENA platform, which was developed to help analysts extract environmental and social information from large collections of unstructured documents. The objective is not to automate investment decisions, but to reduce the time spent locating relevant evidence so analysts can spend more time evaluating what that evidence means.

This also changes sourcing. Investors are no longer limited to structured databases or polished company presentations. Public procurement records, local news, technical studies, government documents, and project reports become easier to search at scale. Opportunities that were previously difficult to identify can become visible without requiring every organization to produce identical reports.

The impact fund-of-funds, Better Society Capital, provides an example that shows what that shift looks like inside a due diligence process. This impact fund tested an AI research tool against its own analysts on 45 venture investments, scoring each across the same impact dimensions the team already used, then compared both sets of scores to a final, human-reviewed score. The AI-generated scores matched the final score 81 percent of the time, against 64 percent for the analysts’ own first-pass scores. But the AI also fabricated facts in an estimated 10 to 20 percent of its assessments, evidence that was not in the source documents at all. The lesson was that the tool applied a framework consistently and fast, but it was bad at knowing when it was wrong, which is why the output still needs someone checking it against the source before it becomes the basis for a decision.

From reporting to continuous learning

The same shift continues after capital has been deployed. Traditionally, impact measurement has relied on annual reporting cycles, where portfolio companies collect information throughout the year, compile it into reports, and investors review those reports months after activities occurred. Useful information often arrives too late to influence implementation.

AI makes a more continuous approach possible. Operational reports, disclosures, remote sensing data, and other digital sources can be reviewed as they become available. The value here is not simply faster reporting. Investors gain earlier signals that allow them to ask better questions, investigate unexpected trends, and even intervene before problems become embedded.

Technology cannot compensate for weak information

The enthusiasm surrounding AI should not make us forget its limits. AI reduces the cost of analysing information, but it does not reduce the cost of producing reliable information. If disclosures are incomplete, inconsistent, or inaccurate, algorithms simply process those weaknesses more quickly.

This challenge is particularly relevant in emerging and frontier markets. Many businesses operate with informal supply chains, fragmented administrative systems, or limited reporting capacity. Some of the enterprises creating the greatest development impact generate the least machine-readable information. 

This also raises governance questions. How should investors validate AI-generated assessments? Who is accountable when an algorithm misclassifies an outcome or overlooks an important local context? A recent OECD report stresses “transparency, human oversight, and accountability” throughout the deployment of AI systems. Those principles are especially relevant in impact investing, where credibility depends on evidence that can be explained and verified. When an algorithm influences which company gets funded or how a fund reports its results, the communities whose outcomes are being measured have a stake in understanding how that judgment was reached, not only what it concluded.

What this means for impact investing

The next competitive advantage in impact investing is unlikely to come from collecting more data as most organizations already have access to vast amounts of information. The advantage will come from turning fragmented information into better judgement without losing the context that gives that information meaning.

That is why AI should be viewed as an enabler rather than a substitute for expertise in this case. On one hand, it can help investors spend less time searching for evidence and more time understanding it. It can also strengthen due diligence, improve monitoring, and support better-informed decisions. However, for now, it cannot replace local knowledge, independent verification, or thoughtful governance. Practically, that changes what the work of impact investing looks like day to day: less time compiling and reconciling reports by hand, more time interpreting what the evidence means and deciding when to trust it.

The future of impact investing will depend not only on mobilizing capital, but also on building systems that make information trustworthy, comparable, and actionable. At The Bassiouni Group, this intersection sits at the core of the work carried out through TBG Capital and TBG Purpose.