What three IPAs, and the evidence so far, suggest about where the real change lies
Over the last few years, AI has started to change the daily work of many industries, and investment promotion is no exception.
In my experience, a typical FDI professional once spent roughly half their time on groundwork: building, cleaning and maintaining contact databases, then shortlisting companies by reviewing financials, past interactions, news and references, before engaging anyone at all. Much of that work was routine, and AI is now beginning to take parts of it over.
But that groundwork was never the hard part of investment promotion. It was what stood between the professional and the hard part. The harder question has always been which signals matter. A company may be growing, hiring in Europe, or acquiring a business. Is it actually considering a new investment? And if so, why now?
So I looked at what IPAs are actually doing with AI, and what the evidence says so far.
Where adoption stands
UNCTAD’s 2026 IPA Observer on AI is the most useful source I found. It describes AI use in four functions: automating routine operations, synthesising information, predicting which investors are most likely to invest, and generating content such as draft proposals and briefs.
It also shows how uneven adoption is. According to the report, 82% of IPAs reporting AI use are in high- or upper-middle-income countries. Among 76 agencies reviewed in least developed countries and small island developing states, only 13% have deployed visible AI tools, mostly basic chatbots. UNCTAD points to limited digital infrastructure, skills gaps and weak data governance as the main reasons, and estimates that entry-level tools can cost roughly $300 to $6,000 a year.
From finding companies to finding signals
Costa Rica’s CINDE is one of the earlier experimenters. In a 2023 interview, Pilar Madrigal, the agency’s director of investment advisory, said CINDE had been running an AI trial programme for about five years, mainly to use predictive analysis to work out where investors are looking to invest, with a model of more than 150 data points tracking more than 800,000 companies worldwide.
She also made a point I found striking. Companies will have more access to data about locations too, so she expects investment promotion to get harder. “The race gets faster,” she said.
An IPA traditionally starts with a list: which companies should we approach? AI lets it ask a different question: which companies are showing signs that something may be changing?
Take the example of an Indian engineering company that starts hiring in Germany, acquires a small European business, and sees its European revenues grow. None of these proves it is considering a new European facility. Together, they may be worth investigating. AI is good at finding those combinations at a scale no human team can maintain.
It can widen the radar. But a wider radar doesn’t necessarily mean better investment promotion.
The first conversation still matters
Invest Estonia is among the best-documented examples. Its chatbot Suve answers questions from official sources. Its e-consulting assistant Eia asks prospective investors a series of questions and provides a tailored report, and can create investment offers and, with some human assistance, assign leads to advisers. Every enquirer also gets a personal human adviser for what Eia can’t handle. The agency says Eia has helped more than 5,000 potential investors since 2019 and had a part in facilitating €139 million of investment, and that about 85% of enquiries were being handled automatically at the time of that statement.
Two cautions. These are the agency’s own figures. And the tools date from 2019 and spring 2020, before today’s generative AI; the agency describes them as smart automation and machine learning.
But the practical benefit is clear. It is not that a machine has replaced the investment adviser. It is that the adviser can know more before the conversation begins. The conversation can start from:
“You already have customers in this market and have started building a local team. Are you considering a more permanent European presence?”
instead of:
“Let me tell you why we are an attractive investment destination.”
AI hasn’t replaced the relationship. It has improved the preparation for it.
The opportunity may be bigger with existing investors
The most interesting application I came across is Invest KOREA’s predictive targeting system. UNCTAD describes an AI system that identifies companies likely to make additional investments, which I read as a focus on investors already in the country.
UNCTAD reports that between 2023 and 2025, 38.5% of these companies went on to invest, against a 17.7% baseline, and that the system saved over 1,000 staff hours a year. Invest KOREA won WAIPA’s 2025 award in the AI category for it.
That points to a possible change in how IPAs think about aftercare. Instead of asking only how to attract the next investor, they can ask which of the companies already here is showing signs of its next investment. It is also a case where the IPA has something no new targeting exercise has: a relationship and a history with the company.
A research-centre summary of the UNCTAD study draws a broader lesson from Estonia and Korea: the greatest benefits arise when AI is integrated into a coherent data ecosystem, rather than used ad hoc. The tool matters less than the data and relationships behind it.
What we don’t know yet
There is however limits of the evidence. The same summary of UNCTAD’s study notes that few results exist to assess return on investment, and that no agency has set up a protocol to measure AI’s causal effect on attractiveness or FDI flows.
That matters for how to read Korea’s number. A flagged company that invests more often than average is promising, but it doesn’t show that the AI caused the investment, or that a human analyst wouldn’t have picked the same names. Higher conversion from a chatbot, like the 20% UNCTAD reports for one deployment in the Democratic Republic of the Congo, is real, but it is also a narrower claim than “AI attracts more FDI”.
There are also risks. UNCTAD flags weak data, limited technical expertise, data protection and cybersecurity risks, and the need for human oversight on high-stakes decisions. Poor data, it notes, can lead to poor results and damage investor trust.
Where AI reaches its limit
Suppose an AI system flags an Indian company because it has increased European exports, hired several people in Germany, acquired a Dutch company and begun talking publicly about international expansion. It would be reasonable to classify it as high potential.
Now suppose an experienced professional speaks to the company and learns what the data couldn’t say. The acquisition was about technology. The German hires support an existing customer. And there are no plans for another European investment for three years. (Again, this is an illustration, not a case.)
Was the AI wrong? Not necessarily. It identified a pattern. What it couldn’t reliably identify was the reason behind the pattern.
Companies don’t make investment decisions because a set of data points reaches a score. Decisions are shaped by customers, competitors, management priorities, capital, risk, timing, and sometimes a conversation that changes how a market is perceived. Those don’t always leave a clean digital trail.
So what changes for the IPA professional?
If the groundwork shrinks, the question is what the time is spent on instead. Perhaps the role moves gradually from:
Research → shortlist → pitch
towards:
Signal → hypothesis → validation → relationship → facilitation
The professional spends less time finding basic information and more time asking better questions. Why now? What has changed? What would trigger the investment? What could stop it? Who is actually deciding? And perhaps most importantly: is the signal we’re seeing really an investment signal?
If AI can give every IPA a better list of companies to call, the advantage won’t come from the list. It will come from knowing which signals deserve attention, what they mean, and having the credibility to explore them with the company.
The more honest conclusion may be a modest one. AI is already helping some IPAs look in better places and prepare better conversations. Whether it helps them win more investment is something the field hasn’t yet measured.
Which raises a question I would like more IPAs to ask: if you adopted an AI tool tomorrow, how would you know it was working?
