Generative AI Is Part Of Application Modernization Now

August 21, 2023 Timothy Prickett Morgan

Generative AI is funny in that it is both an application in its own right when equipped with a chatbot interface and some intelligent prompting with guardrails to keep it from hallucinating and it can also be used as a tool to either generate chunks of applications based on large language models tweaked with a library of code.

This is the first time that we can think of where the application is the tool or the tool is the application, and it presents a kind of chicken and egg conundrum for IT departments all over the world. Mainly: Where do you start to deploy generative AI within your organization.

To help the CEOs, CIOs, and CFOs on Earth think about it, IBM’s Institute for Business Value, a kind of think tank within the IT vendor, have put together a collection of documents called The CEOs Guide To Generative AI, with three chapters already done relating to application modernization, customer service, and talent and skills.

According to surveys done by IBM, 64 percent of boardroom executives say they need to modernize their applications to make use of generative AI capabilities. But where do you start? With three things, just like most thing that have to do with human beings. And here they are in a nutshell from Big Blue:

Grab low-hanging fruit by applying generative AI to apps that have already been modernized.
Go after opportunities that were previously “off limits” – such as applications and processes in core systems.
Stop measuring business and IT goals separately – explicitly prioritize IT projects with the strongest links to business value.

This is bold advice. Telling people to mess with core applications, which might have millions to tens of millions of lines of code and which have been through rigorous testing in a business environment that abhors the risk of introducing any new code except where absolutely necessary, seems to run counter to what most IBM System z and Power Systems customers believe. But we do think that eventually, application development tools will have code conversion features based on large language models that can, for instance, convert from ILE RPG to free form RPG or even to other languages like Java, PHP, or Node.js. IBM already has code generators based on generative AI that can create Ansible runbooks based on plain English prompts, so why not?

You always have to remember what IBM’s interests are when it comes to generative AI. It is not going to make money selling GPU-laden hardware, which would involve big bucks but possibly only modest profits given the high cost of Nvidia GPUs these days. Nvidia is profiting handsomely, but it is not clear how its partners are doing when it comes to profits. IBM, on the other hand, wants to sell consulting engagements and help customers figure out how to introduce generative AI to their IT infrastructure. Big Blue, which is an avid user of this technology itself, is in a good position to give such advice.

Interestingly, after reading what the Institute for Business Value has said about how to add generative AI to organizations, we think that this might be a moment like the Y2K date crisis. At that time, in 1998 and 1999, there was a lot of legacy, homegrown code in the System z and AS/400 bases and third party application developers made hay getting people to throw out those applications because it was more expensive, disruptive, and risky to update the dates in those applications than it was to move to SAP, PeopleSoft, Oracle, or JD Edwards applications. And so many shops did just that, getting a fresh start and modern applications that could be updated externally in the bargain. The difficulty of training generative AI models and integrating inference APIs into applications might just have IT shops in the IBM i midrange market throwing their hands up and just moving to applications that have the AI functions already built in.

Then again, some will do it themselves out of pride as much as necessity.

We shall see. It is still early days for generative AI. What we do know is that companies will have to do something.

This Issue Sponsored By

Table of Contents

Content archive

Recent Posts

Subscribe

Pages

Search