The first step is to identify a problem within an enterprise that can be addressed with AI.
Then comes the discussion and vetting phase: Brainstorm solutions—assessing each for feasibility, effectiveness, and potential impact.
Building the actual prototype should be highly developer-led. Take a bottom-up approach, incorporating developers early to investigate new capabilities.
Join us on June 15th for our Generative AI Salon — a monthly gathering of top AI builders.
In the ever-quickening pace of the AI arms race, every nanosecond counts. At this point everyone and their eccentric aunt on Facebook is working on their own neural network pattern recognition system.
The avalanche of new products keeps gathering speed. Already we’ve seen AI outperform doctors, bamboozle banks, and even nail The Weeknd's falsetto. The World Economic Forum research found that 75% of companies plan to implement AI in the next five years. As the demand for AI solutions surges, experts from MIT propose that the future of generative AI will lean towards niche-specific applications, focusing on tailored solutions for industries and unique use cases, rather than adopting generalized, one-size-fits-all models.
The question remains, however: Will any of these shiny new things drive a profit? And the answer seems to be, Yes—IF they can solve big problems for enterprise companies.
That’s why A.Team hosted its first-ever Generative AI Hackathon: to test out this very idea. It was a two-day sprint to research, discuss, and prototype AI solutions for enterprise problems. Each team of expert AI builders was paired with a Fortune 500 executive with an insider’s perspective on the hairiest problems facing their industry.
Sponsored by OpenAI, Baseten, and CooleyLLP, the hackathon prize included $3,000 in cash, plus $2,500 in credits for OpenAI, and $1,000 in credits for Baseten.
The result was six compelling prototypes and three finalists. On Monday more than 350 founders, investors, enterprise executives, builders, and journalists gathered at the A.Team Clubhouse in NYC (and online) for our first Generative AI Summit, drinking GPTinis (made with Blank Street espresso and vodka) and watching demos before the judges picked a winner. You can watch how it all went down right here.
If you want to create your own generative AI enterprise prototype (and beat your eccentric aunt to the punch), we've compiled a basic roadmap from what we learned from the hackathon teams and their Fortune 500 advisors.
Discovery: Identify the Pain Point
The first step every team took was to identify a problem within an enterprise company that could be addressed with AI. The idea was simple: Solve non-artificial problems with artificial intelligence.
Unlike other computing mega-trends such as mobile and crypto, which gave an edge to startups starting from scratch, generative AI may benefit incumbents more than new disruptors, as they possess the most significant asset for LLMs: data.
Generative AI may benefit incumbents more than new disruptors.
The teams met with their enterprise partners to try to understand the pain points and spec out potential solutions. Most enterprise companies won’t just lend you their proprietary data for the sake of innovation, so the teams worked off of publicly searchable data sets.
JAY, an intelligent claims assistant, is designed to revolutionize the way claims are processed in the auto insurance industry. Anu George, a Digital Transformation Executive, advised the team.
"JAY will be able to help the insurance industry reduce costs by almost 90%," and enhance customer experience with its human-like capabilities, George said.
JAY can handle both text-based and speech-based conversations and offers services in multiple languages with better response times. "We believe that the future is in speech to speech," said Alexander Whedon, an AI expert on the A.Team network.
It also utilizes a novel approach called goal-oriented staging to ensure the system follows a rigid structure and collects the right information in the right order.
Most importantly from the insurance industry perspective, JAY identifies opportunities to upsell. “With this alone, insurance companies can save a lot of money,” Whedon said.
Product Spec: Vet Potential Solutions
Once a problem was identified, the teams entered the discussion and vetting phase. They brainstormed solutions, assessing each for feasibility, effectiveness, and potential impact.
This stage was crucial. Teams had to balance creativity with practicality. They had to ensure their solutions would not only solve the problem but also be coherent and easily demonstrated within the hackathon's two-day timeframe.
mAI CFO serves as a virtual CFO for small businesses. As Paul Sangle, a Product Manager and A.Team builder, explained, "Our platform offers small business owners the ability to get answers to complex financial questions with customized insights."
Imagine a world where an AI CFO can fill out tax forms on your behalf.
It can generate tailored financial documents like P&L statements based on a business's unique financial situation. The mAI CFO dashboard offers a user-friendly interface with recommendation cards based on user history and best practices. The app team envisions future features such as filling out tax forms, applying for business loans, integrating with third-party services, and guiding users through raising capital.
The enterprise advisor, Sebastian Gunningham, Chairman of the Board of Santander and Board member of Saks Fifth Avenue, coached the team to build something that empowered small business owners, which he identified as $12B addressable market.
"Imagine a world where an AI CFO can fill out tax forms on your behalf and suggest ideas for tax efficiency next year," Sangle said.
Implementation: Build a Prototype
The last finalist, Floorplan.ai, built a "Dream Floor Plan" tool that leverages machine learning to create floor plans based on text and simple dimensions, bypassing time-consuming and expensive architectural consultations.
Amir Kazmi, Chief Information and Digital Officer at WestRock and A.Team CxO, helped the Floorplan.ai team identify pain points and vet solutions. But ultimately this stage—the actual building of the prototype—was highly developer-led.
The actual building of the prototype was highly developer-led.
The hackathon took a bottom-up approach, incorporating developers early to investigate, learn new capabilities, and bring back the findings to the product manager. This allowed for a fluid, organic process, with developers and project managers collaborating to bring their AI solutions to life.
Floorplan.ai gave us a peak at the future of automation for many processes tied to blueprinting and planning.
"Our approach has applications in many other environments and industries, and we hope to see industries starting to explore the endless possibilities of generative AI," said Ric Arcifa, an AI expert on the A.Team builder network.
The Winner: JAY, the AI Insurance Agent
This weekend we learned that with a clear process and an innovative mindset—plus major input from enterprise executives—it's entirely possible to build an AI prototype in two days.
By focusing on customer satisfaction and revenue generation, JAY ultimately best embodied the ideal of using AI to solve workflow problems so that people could spend more time thinking and being creative.
The hackathon demonstrated the power of generative AI and how it can be harnessed to solve real-world problems. The future of AI is not just about general, one-size-fits-all models, but about niche-specific applications tailored to industries and unique use cases.
Whether you're an AI expert, a developer, a project manager, or an eccentric aunt on Facebook, the path to building your own AI enterprise application is clear. Identify the problem, brainstorm and vet potential solutions, and then get to work building your prototype.