My two most recent columns for Fast Company have discussed the impact that artificial intelligence/machine learning (AI/ML) and location intelligence will have on organizations. The use cases are fascinating, and the positive impacts on productivity, efficiency, and decision making are enormous.
We won’t have to wait long to see those use cases come to fruition. The emergence of generative AI is as transformative as the proliferation of personal computers in the 1980s and the mass adoption of the internet in the late 1990s.
But there is one major difference. In those cases, companies had years to experiment with these technologies, discover use cases, conduct pilot projects, orchestrate rollouts, and eventually make them cornerstones of their business strategy. The adoption of AI is playing out very differently. There is an immediate arms race, driven not only by the opportunities that companies see, but also by fear of being left behind. Rather than a true crawl/walk/run approach to adoption, the crawl phase looks a heck of a lot like a sprint.
That means that organizations and departments don’t have the luxury of staying on the sideline and letting early adopters blaze the trail. At every company I talk to, there is significant internal and external pressure to be an early adopter of generative AI combined with location intelligence. So, what should companies be doing to take their first steps?
Before I answer, though, I should give a disclaimer. No one has a crystal ball about the future of this fast-evolving technology. Even the most prominent AI experts in the world have been candid about how they do not know exactly how AI will impact companies and society.
With that said, it doesn’t mean that we are flying blind. My team and I talk every day with organizations that are launching initiatives to harness the power of AI/ML and location intelligence, and there are some important questions that come up again and again as organizations navigate these early days of the AI revolution.
In this column, I will address two key questions that your organization should be asking. I will address others in my next column.
DOES CHATGPT NEED A HELPING HAND?
Having a natural language interface is important for every AI implementation, but I would argue that it is even more important for one involving the combination of AI/ML and location intelligence. One of the overarching goals of these deployments is often to broaden access to powerful location intelligence insights by putting it at the fingertips of everyone in the organization—not just the handful of true geospatial experts. Using ChatGPT accomplishes that democratization goal by letting any user ask for information and insights from location-based data using natural language queries.
But anyone who has done any experimenting with ChatGPT quickly finds out that how you ask a query has a major impact on what you get back. That is true even for queries that don’t tap into technical data, but it is especially true for queries about location-related data like that found in geospatial databases, digital maps, and other related data sources. Location-based data is highly technical, has multiple layers to it, and is tied to contextual information—all of which makes precision important for any questions being asked via a natural language interface like ChatGPT.
For the early stages of an AI plus location intelligence rollout, it will often be a best practice to have an engineer with geospatial experience act as a “helper” who can refine the language of queries to get the best results using the most precise wording of questions that is attuned to how data is structured. Over time, this helper will likely become unnecessary as the AI interface, the AI model, and users learn more. Still, organizations should consider this “helper” model at the beginning.
WHERE SHOULD YOUR DATA LIVE?
One of the key early decisions that organizations need to make is whether to leverage a commercial service that will provide generative AI as a service or whether to build your own generative AI model. This decision is independent of the user interface decision and is completely transparent to the user. They continue to ask ChatGPT the question they want to ask, but what happens behind the scenes is very different depending on whether you build or buy generative AI.
There are a lot of factors that go into any build-vs-buy IT decision, but I will point out one that is likely to be decisive for implementations of AI and location intelligence: where your data lives.
Choosing a commercial service (the buy model) simplifies implementation in many ways, but it requires your data to be exported to the provider. When a user asks a question of ChatGPT, the query is routed to the provider who provides the generative AI-driven analysis of data living on their servers and then sends the answer back to your employee.
For many organizations, a cloud-only approach will be a non-starter because sizable portions of their data need to remain on-prem for regulatory reasons, or because the volume overall is so large that exporting to the cloud is not feasible. In response to those practical concerns, the best practice will likely be for service providers to develop hybrid models that combine on-prem and cloud in a way that is customized for each customer organization. For data that must stay on-prem rather than in the cloud, the generative AI model will be deployed on-prem as well. This will allow AI to live where the protected data lives and be integrated with geospatial applications that are already delivering important insights to those organizations. For data that can live in the cloud, the organization can export that data for analysis by centralized AI engines.
This kind of hybrid approach could help organizations navigate those thorny data protection issues and dramatically accelerate the timeline for AI initiatives.
In my next column, I will explore additional critical questions to help steer your first steps in working with generative AI and location intelligence.
Todd Slind is the VP of Technology at Locana, where he leads development of solutions that harness the power of location intelligence.