Migrations and Modernizations in the Era of AI

It used to be that people wanted to migrate to the cloud so that they could save money, leverage managed services, and for a million other reasons too. Things have changed a little bit out there. We are now seeing companies that previously had not considered migration to the cloud submit and start planning to move workloads into a cloud ecosystem. So what changed?

Here is my take on the “changing of the winds.” Almost every company that we speak to is thinking really hard about how they can leverage AI: large language models, retrieval augmented generation, and optical character recognition (oops, this one is just regular old machine learning, for now). One thing is clear- it’s cheaper and easier to leverage AI if your data, compute, and manpower are already in the cloud.

The AI Value Proposition

So I guess the incentives to move to the cloud just weren’t good enough before AI came along. Now, the companies that have already made the move and modernized their workloads are leveraging AI for a plethora of use cases. The companies that are still running legacy tech stacks in the data center are literally scrambling to “get their data house in order” and seeking to understand what parts of their application need modernization to best leverage large language models. 

I hate to say I told you so. But really, we did. Migrating to the cloud has always been about enabling your organization to innovate faster and leverage cutting-edge services. The AI Value Proposition is just much clearer and more convincing than the old “you need serverless” or “global availability” or “just move to the cloud and then use savings plans; it will be cheap, I promise.” Whether the promises were empty or not, the truth is, that the cloud is where AI is happening, and at a breakneck pace, I might add. If you think AI is just a fad and large language models won’t be relevant in the coming years, then you are probably also still building applications in ASP.NET and will never change.

Of Course, you can Leverage AI in the Data Center

To be honest with you, reader, you can leverage AI from the data center. But the real correlation is that companies that have moved to the cloud are also more likely to have cleaned up and modernized their data. Meaning, they may already have columnar data stores, data lakes, data warehouses, huge data sets with all the metadata that you could ever desire, and really great scaling and availability. They are already in a better position to leverage services like AWS Bedrock, Sage Maker, Rekognition, Textract, etc.

These same companies also typically have mature DevOps, Security, and Governance platforms in place, meaning they are better equipped to experiment and innovate with AI and Machine Learning when compared to companies that have been kicking the can down the road for years or even decades. It’s not just about the capabilities of the tech stack that you use; it’s also about your ability as an organization to seek out AI use cases that make sense. Think-product mindset. A company that hasn’t realized that Delphi 6 is maybe a bad choice to continue building a product is also the same company that will simply build an AI-enabled tool just to do it. In other words, they are bad at making informed decisions. 

A Rant: Just Modernize

We shouldn’t need a bright and shiny, transformative new tool to convince us that we need to write good code, maintain and update legacy systems, and do good engineering. Although the carrot on the end of the stick changed from whatever we were looking to achieve before to let’s build this awesome AI use case with our data. The path from point A to point B is still largely the same. Migrate off of the data center into a cloud, and modernize your applications, especially your data platform. Please, for the love of God and all that is holy, don’t try to tack AI onto your 20-year-old client-server application with its out-of-compliance databases and unsupported end-of-life EVERYTHING!

There are no freebies in this ecosystem. If you want to do what the big tech companies are doing, then get your data house in order. Update your systems, and build a decent IT organization that is capable of experimenting and innovating. You can’t put the cart before the horse with this one, and you cannot take shortcuts here. You can, however, go ahead and leverage some AI tools to help you migrate, modernize, and aid in your informed decision-making process. It is strange that we are getting to a point where we can use AI to help us leverage AI better.

Inference for Everyone

I have begun to notice a pretty large disconnect between what companies think they need to do to have a decent AI strategy and what the reality is. This is like the “perceived future state” vs. the “actual future state” conversation. Unless you measure your data size in the 100s of Petabytes, you probably don’t need to train your own model or even fine-tune a base model. What you need is to enable your organization to sort, categorize, and index your data, create access patterns that make sense and are secure with the right technologies (like a vector database), and then use existing models in tandem with your data (via data retrieval) to enable most use cases.

To be clear, your data set is not needed to train models. Your data set is needed to perform inference across many models and provide AI solutions that are specific to your organization and your goals. The models are already good enough, and unless you have a really good reason to train or fine-tune one, don’t! Now that we know what the target is — having good, clean data sets, data governance strategies, and a way for LLMs to retrieve your data and stick it in a prompt —let’s go work towards that goal.

Conclusion

Whether you are moving into the cloud for all of the reasons that companies previously made the switch or you are moving because you just want AI, Ippon is here to help you from start to finish. We are an AWS partner with Migration Acceleration capabilities (MAP), and we are also a Snowflake partner. If you don’t know what Snowflake or Databrick is, go read up on it. Here is a hint: it will enable you to use AI.

If this blog came across as sounding like a rant to you and you found yourself shaking your head yes, then you are already in the know and on your way, my friend. If you are offended by this blog, then please reach out to sales@ipponusa.com; you need our help!