Guest: Alois Reitbauer (LinkedIn)
Company: Dynatrace (Twitter)
Show: Newsroom

Dynatrace is expanding its Davis AI, which has been around for a very long time, with a new generative AI component called Davis CoPilot. It is considered the very first hypermodal AI platform in the industry for observability and security.

In this episode of TFiR: Newsroom, Dynatrace Chief Technology Strategist Alois Reitbauer shares the details of the latest iteration of Dynatrace’s powerful platform.

The new Hypermodal AI platform:

• Dynatrace’s hypermodal AI platform is a combination of 1) predictive AI, which does forecasting and anomaly detection, 2) causal AI, which can do cause-and-effect analysis and uses a multitude of mechanisms to do this behind the scenes, using an ascending environment with Smartscape analysis, understanding the topology and of the environment, etc. and 3) Davis CoPilot generative AI, which is the large language model (LLM) component.
• Dynatrace takes raw input data (traces, metrics, logs, user sessions, data events, etc.) and semantically enriches them. It takes these modalities of data and converts them into a “hypermodality”.
• Example: Input is a time-series stream of data. Dynatrace will be able to determine that this time-series stream is actually response time. It belongs to a certain service that is linked to a certain service level objective. It’s behaving this way and is predicted to behave this way going forward. This metric was involved in a couple of problems in the past.
• Then, Dynatrace will take the predictive result and will feed it into the causal AI to do the root cause analysis. It will then provide the list of components in the system that went wrong and the reason (e.g., this changed in this deployment and impacted these users).
• Then, it will take all of this information to create a remediation proposal or do other types of automation based on it.
• Dynatrace could also go the other way around. If the customer asks something, it can augment the query and automatically build knowledge notebooks, dashboards, workflows, or even answer questions about what they should be doing in their environment.

On Generative AI:

• It’s going to make us rethink how we interact with systems and how good we want systems to be.
• LLMs play a crucial role, but they do not comprise a full AI platform. Asking an LLM to answer a question or do a simple calculation is fine, but that’s not what it’s built for. What it could do is figure out that you wanted to do a calculation, use a third-party plugin to actually perform the calculation, or outsource it to another model to another service. That’s where we will see a lot of development.
• People have to get over the technology enthusiasm and learn how to build real systems on top of it.
• The quality of the prompt will impact the output, so instead of going for “one model, one technology,” it is better to go for an integrated system platform approach.

On Observability:

• It has changed from what used to be a niche, highly specialized expert technology in large corporations to the more accessible technology that is widely adopted by not just expert teams, but individuals.
• When Dynatrace entered this field, they wanted to make it more accessible to even more people so they’re using generative AI and combining it with causal and predictive AI to lower the entry barrier.
• There will be a commoditization of the usage of observability tools. It will require less expert knowledge and the tools will explain even more what they’re showing and why they are showing it.

How Dynatrace helps customers:

• It combines different types of AI together. Rather than looking at the individual pieces, applying AI on some data, and then havinghave the customer decide what to do with the result, they leverage the hypermodal AI to make the choices and ensure that the customers get the right data without needing to become a data scientist or machine learning expert to use it.
• The Dynatrace approach is to enable the customer to get in the car and drive, and not have to open up the hood to understand how the motors work.
• It can trigger automations based on the results of the root cause analysis. For an automated AI system, this can be done in milliseconds, which is a huge boost to mean time to repair (MTTR), especially for customers in travel, ecommerce, and financial industries where one minute of downtime can mean 6 to 8-figure loss.
• It provides easy-to-comprehend analysis steps and remediation hints that customers can build and chain together.
• It allows customers to tailor it to their use case so that everybody in the enterprise gets exactly the view they want without having to learn another tool.

This summary was written by Camille Gregory.