AI for Backup: The Future or Just Hype?

Stephen Manley, CTO

Modern computers offer almost unlimited capabilities… including the power to make us feel incompetent. Most software boasts hundreds of features we don’t know how to use, especially since we’re all trying to run a dozen different products just to get our jobs done. We’ve all stared at a console silently screaming, “Please, just do what I need!” 

With Generative AI (Gen AI), we stand at a fork in the road. Will Gen AI be the next technology to make us feel inadequate, or can it finally make our products work for us? Druva believes that Gen AI is the next giant leap in our journey to building autonomous data protection. Druva wants to do the job of protecting data for you, and AI will help us — today, tomorrow, and in the future. 

In this two-part series, we will first discuss the basics of Gen AI and how it can be applied to backup. In the second part, we will explore how Druva is using Gen AI to help us help you. 

Generative AI Solutions — Do’s and Don’ts

Gen AI cannot solve all technology problems, but if you avoid the dangers and superficial uses, you can do amazing things. 

The dangers to avoid: 

  1. Cost — When “big data” was the hot technology, companies built massive Hadoop clusters, waited for MapReduce to calculate the secret of life, and only got… a huge bill? AI needs a focused use case, or your bill will be staggering.

  2. Security and Privacy — When the cloud emerged, every business unit engaged in “Shadow IT” which put data at risk. “Shadow AI” is an even bigger threat to your data privacy. 

  3. Hallucinations — Generative AI is irrationally confident. It gives you an answer, even when it is making it up. You need guardrails to keep your AI solution in line. 

The superficial uses to ignore:

  1. Chatbot — While useful, they are standard functionality that doesn’t integrate with the product. It’s better than a search interface on documentation, but not by much. 

  2. “We expose the data to you” — If somebody is making AI your problem, then they’re making you work for their product. 

What to look for in a real AI solution:

  1. Clear Use Cases — If there isn’t a clear use case, they’re selling “Do it yourself” and putting you to work for them. 

  2. Integrated Security and Privacy — The AI solution needs to be part of the product’s existing security and privacy model. If it sits outside the standard, you’re taking on risk.

  3. Trusted Data — To minimize hallucinations, validate if your solution provider has access to a large amount of clean, accurate data on which to train the model. 

Real-World Backup Example of Generative AI in Action

To explore the role Gen AI can play in backup, let’s break down a common problem and see where Gen AI applies. 

Every customer has struggled to diagnose backup failures. The team has to assess whether there is a systemic issue or just a transient failure. If there is an issue, they have to identify the cause, which can include backup software bugs, backup storage capacity, network performance, data source outage, administrator error, cyber attack, and more. After diagnosing the issue (or issues), the team needs to create a resolution plan. The process is complex and is made even more difficult because most user interfaces are built for the “happy path.”

Gen AI can help diagnose and remediate your backup failures:

  1. Conversational Interface — Most IT admins are not comfortable navigating their backup UI because they use it so infrequently. A conversational interface allows the admin to ask a question, which the AI can then understand and translate into action. 

  2. Reporting/Data Retrieval — The first step in any troubleshooting exercise is to gather data. The Gen AI can call the appropriate reporting APIs to bring back just the data you need. 

  3. Intelligent Response — Every system, no matter how simple, has unique terminology and issues. The Gen AI can locate and summarize the appropriate documentation and KnowledgeBase articles to help the IT admin assess what is happening. 

  4. Recommendations — Everybody wants to know their options, and what are best practices. The Gen AI system can summarize the pros and cons of all the options, so the user can make the best decision. 

  5. Navigation — The Gen AI can drive the IT admin to the appropriate locations in the UI to put the plan into action. 

Gen AI can simplify every step of a complex problem, speed time to resolution, and improve the quality of the response. But only if you have a plan. 


Generative AI is at the center of an enormous hype storm. Used improperly, it can be an expensive security risk that gives you the wrong answers. Used superficially, it can become another tool that expects you to do all the work. Used properly, however, Gen AI can cut through the fog of the solutions in front of you. 

Therefore, when your vendor comes to you with a Gen AI solution, ask them:

  1. Is it part of the product or totally separate? 

  2. What data are you using? 

  3. What use case are you solving? 

If they have a good answer to all three questions, they’ve come to help you. 

In the next post, we’ll give Druva’s answers to the questions, and you can judge for yourself whether we can help you solve your data protection challenges.