Artificial intelligence (AI) is already having an impact on corporate decision-making. Tech companies, for example, have sacked more than 50,000 people so far this year, according to the Layoffs.fyi tracker. Analysts argue this is in part due to them reshuffling their resources towards investing in AI.
It has also become a prominent force in foreign direct investment (FDI). In 2023 alone, AI-related FDI projects in the software industry worth more than $9bn were announced by companies worldwide, according to fDi Markets. Since 2021, there have been more than 1400 of these investments, from the likes of Google, Microsoft, Databricks, Deloitte and Samsung.
Even though the technology has been around for decades, generative AI, such as OpenAI’s ChatGPT and Google’s Gemini, allows individuals and businesses to freely experiment with it. Professional services firm Accenture estimates that 40% of all working hours across industries could be affected by large language models. The IMF predicts about 60% of jobs in advanced economies could be affected by the technology.
But can generative AI also help companies on their investment location strategies? For site selection consultants, who perform this very task, general pre-trained transformer AI models hold great potential. Finding suitable sites for companies is complex and nuanced. But it still involves repetitive tasks that could potentially be automated by AI, such as collecting and analysing publicly available data.
Consultants also have to communicate with clients to provide them support before, during and after their investment decisions, a task potentially well suited to AI chatbots or responses that could be automated.
Several site selectors tell fDi that they are experimenting with the technology for generating speeches and initial research, but stress it cannot replace the human touch needed for tasks like incentives negotiations. Nevertheless, some AI companies show the potential for the technology to change the way businesses of the future find suitable locations to invest.
AI site selection study
A white paper published in January 2024 by the Site Selectors Guild (SSG), a trade body, provides useful insight into AI’s potential in site selection. The study undertaken by David Chiu, a third-party computer science professor at the University of Puget Sound in Washington State in the US, explored how AI fared against human consultants in two US site selection scenarios.
One was for a software company relocating its headquarters out of San Francisco. Another was for a manufacturing business finding a location for a new facility to increase its capacity. Three widely used AI models were asked to generate and rank 20 potential locations for both the office and industrial project. Two types of prompts for the models were used: one with a single large paragraph of text, and another with several shorter sentences.
“The results were super unreliable,” says Mr Chiu. “You can give [these AI models] the same prompts over and over again, and nine times out of 10, it’ll give you a different response.”
The results differed by type of project, model and prompt. The study found almost no overlap between the AI-generated and consultant-made shortlists of recommended locations for industrial projects. AI-generated locations for the office project were slightly more promising, but still fell short. Two anecdotes epitomise this. One time an AI model suggested that Mr Chiu refer to a site selection consultant to get this information and generated no results. Another time a model suggested San Francisco as a suitable location for the company to relocate to, the very city it was trying to leave.
Larry Gigerich, executive managing director of Ginovus and chair of SSG, says that currently available AI tools are only used “on a limited basis”, but do hold the potential to make site selectors more efficient.
“We use this technology to help us analyse site selection data to identify key high-level takeaways early in the process for our clients’ projects,” he says, adding that the current tools “have a long way to go” before they can address more technical issues. He also argues that site selection is “part art and part science” in tasks like structuring financial transactions and evaluating talent.
Emerging platforms
Some companies are, however, developing AI-powered tools to help businesses in their growth journeys and, as a consequence, bypass any other intermediary such as site selectors. Peachscore, an accelerator platform for start-ups, provides recommendations for business owners and founders to help improve their businesses, attract funding and suggest suitable locations for them to scale and open new offices.
Through its own generative AI model, Peachscore claims to recognise patterns of successful companies and benchmarks this against start-ups on its platform. It then helps these start-ups to identify investors, customers and competitors to shape their growth plans.
Alex Mojtahedi, CEO of Peachscore and a former venture capitalist, tells fDi that start-ups are using the platform to assess how to enter the US market by suggesting the type of finance structure and locations best suited to their industries and target market.
“Our system is able to analyse legal information like your entity type, your location, and even can recommend [the best course of action],” he says. Through analysis of the sectors in which a company operates, Peachscore generates the probability of success of opening an office in a particular location.
“If there is a new company that wants to scale and expand to a specific city, you can see the [percentage probability of this venture] becoming successful,” explains Mr Mojtahedi. Peachscore tracks about 150 major tech hubs and 200 other mature hubs globally and is working with government agencies, including in Germany, seeking to help internationalise their start-ups.
Peachscore is not alone. Kobold Metals, a mineral exploration company backed by US billionaires including Bill Gates and Jeff Bezos, is using AI to determine which countries to target. In December 2022, Kobold agreed to invest $150m in copper exploration in Zambia. The company claims AI “transforms the unit economics of exploration”, arguing it can more accurately predict the copper grades in the ground.
A global survey by Kearney of executives, at companies with annual revenues of at least $500m, found that 64% of them expect to increase their use of AI when making investment decisions over the next three years.
Data quality
John Evans, managing director of advisory firm Tractus Asia, stresses that AI’s usability rests on data availability and quality. “Site selection across Asia is still done with a lot of primary research and interviews,” he says, adding that there is a lack of databases with accurate information in the region.
This sentiment is echoed by other consultants, who stress that regional and city level data is far less available and reliable. A lot of site selection involves doing due diligence on the data provided.
“I think where AI will be the most valuable quickly is for location study of several countries,” says Mr Evans, who adds that AI is not going to be able to negotiate an incentives deal for investing companies.
Deployment of AI in the site selection industry remains in its infancy. The majority (57%) of consultants surveyed by SSG in November 2023 said that they do not use AI for their office or administrative processes. The same was true for assisting in a client’s location search.
The technology is still reckoned to have the potential to free up the time of over-stretched site selectors. However, in practice, efficiency gains may be limited or even totally lost by the need to validate the results of any AI-generated study. Just one third of site selection consultants trust the results generated by AI, but this does not mean the technology should be discounted entirely.
While AI deployment for location decisions is at an early stage, as platforms like Peachscore and AI tools improve, the technology could play a more prominent role in the future.
fDi
