The Reuters news staff had no role in the production of this content. It was created by Reuters Plus, the brand marketing studio of Reuters.
Produced by Reuters Plus for
Case Studies
See real-world examples of how businesses are putting insights into practice to innovate, modernize and unlock value.
Microsoft
EY Consulting and Microsoft’s Launch AI simplifies financial product launches with GenAI.
Microsoft
Featured articles
The new era of intelligent finance powered by AI and ML
LEARN MORE
How AI agents will take GenAI from answers to actions
LEARN MORE
Disclaimer: The Reuters news staff had no role in the production of this content.
It was created by Reuters Plus, the brand marketing studio of Reuters. To work with Reuters Plus, contact us here.
Xoople
How can AI provide valuable insights to enhance farming? Bayer Crop Science improved accuracy and saved time in answering questions on its agronomy products with a collaborative GenAI tool. Learn more
Bayer Crop Science
EY AI Insights
LEARN MORE
EY.ai Value Blueprints
LEARN MORE
The dividend age: How AI is turning promise into payoff
LEARN MORE
Bayer Crop Science
How agentic AI should reinvent work beyond automation
LEARN MORE
How to empower businesses with trusted conversational AI
LEARN MORE
Local data centers: the key to speed and compliance in business
LEARN MORE
EY and Xoople teamed up to turn Earth data into actionable AI—helping businesses anticipate disruptions and align operations with live data.
How AI Transforms Earth Data for Good
The dividend age: How AI is turning promise into payoff
LEARN MORE
Watch more: On the Road
CES 2026
SXSW 2026
The AI Summit New York 2025
Show Transcript
How AI Transforms Earth Data for Good
Our customers want answers on what's happening on the earth. We realized that in order to provide those earth intelligence solutions to our customer, we needed the right data and we needed right tools. So we are working to produce and curate the best Earth data set possible. We don't always look at data as something that can be utilized in new and differentiating ways. And this project has really impacted the way that I think about the future of AI and the data that's going to enable it.
Xoople is an Earth intelligence company that we created with the objective of providing earth data and EarthAI as a service to companies worldwide. Earth data is actually everything that describe what's happening on the ground. It can be weather information, it can be traffic. EarthAI is a new category where Earth data is united with a new kind of multi-modal AI with the idea of being able to extract, at scale, information from the complexity of the Earth data so that companies can use that information for better, more insightful, and sometimes even faster, decisions. EarthAI is going to be revolutionary, we think, in the way companies operate in their physical world. And this is everything from supply chain to agriculture, to insurance, financial services, construction. That's why we are working with EY at the global level, because really the last mile of the delivery of our solution is so key. It's not just about developing the solution, it's about integrating our solution in the enterprise workflow.
Xoople and EY strive to solve three key challenges. The first was to technically enable very complex geospatial data into AI ready data sets. The second was to research and validate the market fit for these data sets. And lastly, to create the go-to-market strategies and programs to get this data into the hands of the clients that it can enable. To address the key challenges of this program, we teamed with Xoople in a number of different ways. The first was we brought together our ecosystem of hyperscaler partnerships with key organizations like Microsoft and Databricks to bring the technology that would help scale these fast within the enterprise. We also helped to build ready-to-go, go-to market playbooks that can be utilized to explain to clients the value that can be generated by these datasets, as well as the playbook of how to integrate them in at an enterprise level.
EarthAI is not just about collecting the data. It's really about getting, on one side, the right data, but on the other side, the right AI, the multi-modal AI to process that data at scale and extract that information. And again, that's not a standard LLM. This is a complex AI that has only been developed in the last few years and actually is just at the prototype level. There were key moments during this program where we had to pivot our approach. And one of those was the onset of generative AI. When we started this program together, generative AI wasn't a mainstream technology. But as it did emerge during the program, we realized that the acceleration of the availability of this data was going to become increasingly more important.
The real powerful EarthAI is what's coming. Privacy and security are key for being able to deliver not only an effective solution, but a trustworthy solution. And so what we are working with is with technological partners on a unified global data approach that has an added extra layer of cybersecurity on top of that so that we can be sure that we have secure and efficient access to our data and our customer can have the trust that their data is also secure when creating an EarthAI solution.
This project has really impacted me as I think about the future of AI. First of all, I think the applicability of this information across various different sectors is really an example that a number of different industries should take notice of. The second is the democratization of AI. Over the last three years, generative AI has put AI into the hands of everybody, but we have to go further to make the data democratized to enable that AI, and this program was a great example of how to do that.
What is really exciting for us is the fact that we can finally do something that can bring benefits at scale, not only to customers, but to people. So for the first time, it's not a tech push, it's really a solution that can have an impact on any industry.
Show Transcript
Show Transcript
We were exploring the possibilities of generative AI technology across the finance ecosystem in pursuit of very specific business outcomes like time efficiency, productivity for users, and automation.
When you think about generative AI, you think about really transforming the way we work and transforming organizations. That's what we do at EY. Microsoft had tens of thousands of finance files scattered across many different locations. And if you could somehow distill all of that knowledge and then tailor it to each new product launch, you could really drive value and decrease time to market. I think early on, both EY and Microsoft realized that this was really the perfect opportunity to deploy generative AI and specifically retrieval augmented generation or RAG.
With our partnership with EY, what we were trying to solve for was scale.
Working with Microsoft, we created Launch AI, which was really a generative AI system meant to simplify the complexities of taking products to market and particularly gathering requirements on the backend financial side of things. We're having great user adoption and really driving towards our goal of reducing time to requirements gathering from six to eight weeks to three to four weeks. This is really the first proof point of using generative AI to drive efficiencies within the finance organization and across Microsoft, and in general, really a proof point for all enterprise companies that have these sorts of processes.
As we started the project, we were dealing with very discreet processes in inputs and outputs. As we engaged on additional engagements and projects and expanded the portfolio of commitments, we quickly identified common patterns, the least common denominator. So we started investing on what happens if we can democratize the use of a lot of these models or capabilities without needing to custom tailor the solutions for a specific process or business and we were able to do that and get the same efficiency or get the the same value in most cases as they would if we custom tailored a solution for them.
A thing we realized as we were building this particular use case is really the need for a platform-based approach for generative AI at Microsoft and at Microsoft Finance. With Microsoft, we then created Replay AI, which is a platform for generitive AI use cases and applications, really meant to create a shared level of governance, but also let people building generative AI applications leverage common patterns or common code that are used across generative AI use cases.
The solutions came together through collaboration across many different teams, inclusive of EY, inclusive of our own engineering team and the brainstorming and the design and our counterparts in IT, in Microsoft, to make sure that we were focusing on the right problems and on the opportunities and that what we were building scales. The feedback that we continuously get from our users is that they love the technology. We've heard the words, it's magic, used continuously. A lot of manual work, a lot of manual processes are now fully automated. Tedious activities, especially when dealing with hundreds of documents or thousands of pages, can now be queried and prompted in seconds and a lot of times automatically.
We learned that when building generative AI systems, it's very important to build in a modular fashion. And if you don't take time in that design phase, you're gonna find yourself re-engineering the entire system later. This is only the beginning. Technologies like this are a first step in a much larger transformation, not only of finance, but of the way we work in general. And I'm excited to see the future and what comes.
The partnership with EY helped us both scale our team and scale the skill sets that we had across the team, especially with a very emerging and nascent technology like generative AI. For businesses to get started, I would say take the leap. I would think of it in two fronts. One, Microsoft offers a great platform with Copilot. We're continuously evolving that platform with knowledge and skills, and it will become more scalable and more productive with every cycle. We also have bespoke processes that are highly complex and that need maybe more custom-tailored solutions. I would also encourage individuals and organizations like ours to take the leap and explore what extending from that copilot platform can achieve for them and their organizations in terms of automation and productivity.
At Bayer Crop Sciences we have a lot of frontline employees that support farmers and dealers. Sometimes they get asked questions they don't know the answer to. And getting an answer can take time. And in farming, that insight can be urgent. I'm Dan Kurdys, I'm the Global Business Lead for Gen AI at Bayer Crop Science. We saw that Chat GPT's 3.5 performed well on the Certified Crop Advisor exam. That is a standardized test on agronomy principles. And that set the light bulb off. If it was performing well on general agronomy principles, we thought, well, gosh, if we added Bayer's agronomic intelligence and vetted it with our experts, how good could we make it on applied agronomies in our products?
My name is Ed Bobrin. I lead our Microsoft AI and Data Practice at EY Consulting. Bayer came to EY questioning whether we could build something using Bayer Agronomy data that would outperform a standard chat GPT model. That was our challenge, figuring out how we could use their data and produce a credible result for Bayer. The top priority when we were implementing the gen AI solution was to make sure that the answers it was going to deliver were credible. So using a validation framework and a scoring mechanism we developed together with Bayer, we were able to establish a scoring chart, which was kind of the matrix between crops and product types to show exactly how we're performing from a credibility perspective with our answers to meet that high bar that Bayer set around 93% accuracy. That automation was critical because if it's not credible, it's a non-starter.
The three quantifiable outcomes that we've seen at Bayer on this solution is that we're seeing 42% improvement in answering with accuracy on our products in agronomy over Chat GPT-3.5. We've also seen that on our frontline agronomy team, 90% willing to use and recommend it to a colleague. And finally, as we're getting more feedback, we're seen time savings of our frontline staff of one to four hours a week.
One of the aha moments we had after we got through the first phase was that we had to ensure the data was always the most current data. All this data changes often. So what we're answering with today, that version of the document that's got, you know, version two, version 2a and 2b happened since we actually stood up the application. So now we have to think about who can actually put the data into the system. Is it the latest version? So coming up with automation that would help guide that process to make sure that the model is always grounded in the latest document to make sure that that level of accuracy is maintained. One of the interesting things we've learned along the way is that we understand specifically where you're located. It's a different answer in Pennsylvania than it is in Kansas, right? So geography is an important component to answering the question correctly.
EY Consulting was committed to the mission and the purpose and then willing to come to the table with expertise around the technology of their own and challenge us and us challenge them to really push what was possible. And I think that was really part of the magic in getting to this solution.
Some of the lessons we've learned that we could share with other businesses is getting your data ready for AI means different things. It's unstructured data. Getting that data to a usable format for a large language model takes a little more effort in terms of using technology to break it down and make it addressable for a gen AI application. That applies across all different industries.
Some key lessons that would help with implementations on gen AI are first, allow your responsible rebels to tinker with gen AI solutions and build prototypes. Then, encourage them to focus on use cases that are aligned with your strategy and mission, and then have them really focus on getting quantifiable results. Those are the two that I think are really important in terms of moving forward and innovating with gen AI solutions.
This was inspiring for me personally. I do see an amazing opportunity for this application to go really broad in terms of its underlying capability. So that's pretty exciting. We're getting past the curve of, gen AI is gonna take my job and replace jobs. We're finding that spot where we're saying, no, we can take the data that's locked away in your organization and make it easier for you to do your job.
Bayer is focused on health for all and hunger for none, and using regenerative agriculture as a model of agriculture for that future. And here in gen AI technology, we have proven that we can give better access to applied agronomy expertise and knowledge to support farmers in improving their livelihoods, producing more sustainably, and adapting to climate change into the future.