Moving GenAI from proof of concept to production on AWS
A prevailing critique of generative AI tools is that they can't be put into practice. However, some organizations now use vendor tools to produce capabilities for their clients.
NEW YORK -- Nearly two years after the introduction of ChatGPT, generative AI proof of concept is now moving toward production for enterprises' internal and external processes.
In the early days of generative AI, some of the data and security problems associated with these large language model (LLM) tools made many organizations skeptical of inflated expectations about GenAI.
However, as the industry moves from infancy to toddlerhood, more organizations have adopted internal processes that allow the use of GenAI to boost productivity, without compromising customer data.
Ensono's internal and external processes
One organization doing this is Ensono, an AWS partner and provider of managed services for enterprises.
The company is known for its Modern, Cloud-Connected Mainframe, or MCCM, an updated hybrid version of the mainframe computer.
"Mainframes kind of run the world," said Brian Klingbeil, Ensono's chief strategy officer, in an interview with TechTarget Editorial at AWS Summit New York Wednesday. "All the banking systems, airlines, ATM machines, the IRS and Social Security, a lot of … state and local governments, they still depend on these things as the backbone of their IT."
While managing mainframes and providing cloud migration processes for customers on AWS, Ensono has developed some internal processes using GenAI on the back end.
One of those processes uses DocuSign's AI capabilities to improve workflows and redline contracts by highlighting changes between different drafts of a document.
DocuSign's new Intelligent Agreement Management platform uses AI technology to extract the most essential terms within a contract. It also proposes new language.
Another application moving toward production is the Ensono Predictive Engine.
The idea behind the system is to compile internal information about IT system problems and automatically create tickets for items forecasted to need repair or maintenance.
"The whole service provider mentality is shifting from monitor alert -- to fix, predict, prevent and optimize," Klingbeil said.
We'll be leveraging other people's embeddings of AI into their tools and taking advantage of those.
Brian KlingbeilChief Strategy Officer, Ensono
Ensono is looking at ServiceNow's AIOps-based Loom technology to build its predictive engine tool.
As a service provider, Ensono is not buying up the expensive and currently scarce Nvidia H100 GPU chips that power GenAI systems, nor producing its own LLMs, Klingbeil said.
Instead, "We'll be leveraging other people's embeddings of AI into their tools and taking advantage of those," he continued.
Ensono's predictive engine is not only for internal use but also helps provide and manage processes to customers.
For example, Ensono recently built an AWS cloud-based process for an airline in which every time a plane lands, airline engineers can look through the GenAI-generated data sets to determine what parts of the plane need repair.
"This is a great example of how if we did something for ourselves, we can apply some of the same tools for our customers when they have a use case that matches that," Klingbeil said.
IT leaders gather at AWS NY Summit to learn more about the cloud provider's GenAI products.
Brainbox AI ARIA
Similar to Ensono, Brainbox AI, a vendor that works with commercial real estate owners and managers to make HVAC systems more energy efficient using AI, has started to take some of its GenAI ideas from proof of concept to production.
As one of the original beta users of the Amazon Bedrock GenAI platform and now a full-fledged AWS customer, Brainbox has been using the service to build an autonomous AI agent.
The vendor introduced the AI agent, Artificial Responsive Intelligent Assistant, or ARIA, in March and plans to release it in beta next month.
ARIA helps facilities operators and building managers query and command building infrastructure and devices using voice or text.
When building ARIA on Bedrock, BrainBox could update the different LLMs undergirding ARIA with the latest LLM upgrades available on the market, said Jean Simon Venne, the vendor's chief technology officer and co-founder.
"The entire architecture of ARIA doesn't care about what model [it uses]," Simon Venne said.
"We're always floating with the latest on the latest. It's making our life much easier in a sense that we're focusing on other problems than the problem of which model should we use and do we have access to that model," he said.
Other than working with customers like Brainbox AI to turn GenAI proof-of-concept plans into production, AWS is also collaborating with partners such as Deloitte to help organizations scale their AI workloads.
"This new strategic collaboration agreement is moving our clients from tinkering or proof of concepts into value creation through GenAI and moving into production," John Byron Thomas McGinnis, principal at Deloitte, said in an interview with TechTarget.
"We're not only looking to solve a business issue or create business value for a specific client," he continued. "It's meant to have a market ripple effect across the industry."
Esther Ajao is a TechTarget Editorial news writer and podcast host covering artificial intelligence software and systems.
