Whenever there is a conversation about AI chatbots, there is a chorus of praise about how impressive the technology is, or the incredible things it can do.
People from all walks of life are noodling with chatbots to better their work and personal lives.
“Generative AI and chat technology is fantastic, but it’s only fantastic when you interact with it,” says Anthony DiStauro, solutions architect of the AMI Platform at BMC. “Otherwise, the technology sits there cold and idle until you go and work with it.”
It is becoming plain that the only way to get them working to their fullest potential is through prompts.
The AI Era in Mainframe has Begun
Most co-pilots packaged with IT tools do a passable job of answering questions and finding information when asked, but they fare somewhat underwhelmingly when it comes to acting on scheduled commands, or making smart decisions about how much to share with a particular user.
BMC is doing the work of infusing sixth sense into its co-pilot taking AI awareness to the next level. At the Tech Field Day Extra at SHARE Kansas City 2024, BMC introduced AMI Assistant, a conversational interface that is always aware of who’s at the other end.
AMI stands for Automated Mainframe Intelligence, an initiative aimed at making mainframe self-managing using AI, ML and predictive analytics.
A difficulty that all companies experience with mainframe is the prevalence of old practices and workflows, and a mixed demographic of workers.
“We have our 30+ year professionals. Bright and true, but they are on their way out as we segue to a new generation of mainframers,” says DiStauro.
Only companies that can synergize modern technologies like AI and cloud-native with mainframe stand the chance to stay competitive when the next wave of digital transformation hits.
But BMC disagrees that the relation between GenAI and tech is one of hammer and nail that it is often made out to be. Over-reliance on GenAI tools can in many cases be opposite of helpful.
At BMC, the thinking goes like this – GenAI is a value addition to IT products. If done right, it can significantly elevate the value of a product. But by no means is it the hammer to all IT’s problems.
The AMI Platform Bolsters Mainframe Processes
The AMI Platform provides services and capabilities that help a new generation of mainframers reach their potential without the care and feeding of the previous generation. Out of the box, it provides a single point of entry through the AMI console, APIs, CLI and SDK that new mainframe engineers are in the habit of using, useful data services, and a GenAI experience delivered through GPT services, tooling and the AMI Assistant.
Mainframe GenAI for BMC boils down to four key objectives – an in-product GenAI experience that eliminates the need for context-switching, flexibility of deployment that makes the product adaptive to any environment, ability to bring customers’ own data and LLMs, and the option to augment the platform using “tribal knowledge” assets.
“We’re using open source LLMs where we need,” informs DiStauro, “but we have a retrieval augmented generation pipeline that is where we go for external knowledge that is specific to your domain that the LLM wasn’t trained on.”
“We have a vector database infused with a bunch of BMC knowledge that is used alongside the LLM,” he continues. “But then we also allow you to bring in your own tribal knowledge.”
This could be rules, business logic, best practices, existing code, and so on. The platform processes and integrates this information into the built-in database. The in-house tooling used by the team at BMC will be available to the customers for building this database.
A Smart Assistant for New Mainframers
DiStauro describes the AMI Assistant as “an expert in a box that professionals can lean on”.
The AMI Assistant is an embedded interface inside the AMI Platform Management Console. Users can log into the platform using single sign-on (SSO). All products and services that are available to consume are displayed on the landing page. These services can be accessed directly without separately logging into each one of them as the SSO works across all services.
The AMI Assistant appears on the right-hand side of the screen. The chatbot responds to prompts on the fly, answering questions, providing explanations and summaries, and even taking actions on behalf of the users.
For example, if you ask AMI Assistant about the deployment frequency of a certain component, it will come back with the numbers. If you ask it to explain what the numbers mean, it will describe them for you.
But what makes it truly impressive is that it is an “active GPT”, meaning it can execute commands set-and-forget style. To help understand, imagine that you need to aggregate certain metrics over a time horizon and build a report around it. You need to email this report to a group of people in the company periodically. At a scheduled time every week or month, AMI Assistant can do this work for you without requiring repeated commands.
It can take this action based on calendar events, and additionally control all the data moving in and out of it using AMI Data Services. With a built-in security layer that polices all activities, it can control sharing of information based on role and role accesses.
DiStauro highlights that the AI engines working at the backend can also tweak the responses to match the skill level of the user.
“We give you the knobs and switches to rate our responses to keep fine-tuning the AI engine so that they get smarter as we go.”
DiStauro shared a few forward-looking things during the presentation that will be the part of the solution in the near future. Among them is a prompt builder. This application will help users fashion and manage prompts in a point and click fashion.
Additionally, a layer of explainability will be added for mainframe professionals new in the job to gain clarity and transparency on complex datasets.
Be sure to watch BMC’s presentations from the Tech Field Day Extra at SHARE Kansas City 2024 for a deep dive into the AMI platform. Also check out this Gestalt IT Rundown episode where Stephen Foskett and Tom Hollingsworth discuss the criticality of BMC’s AMI initiative to the revival of mainframe.
