Artificial intelligence (AI) is rapidly evolving, and organizations of all sizes are looking to leverage the groundbreaking technology — as quickly as possible.

But getting apps into production can be problematic (and not just in AI): Developers face obstacles and bottlenecks that slow them down. This could be due to lengthy approval processes, inefficient workflows between departments or a lack of resources.

This is where platform engineering is coming into play. The concept has only emerged in the last few years and is what some call a progression of DevOps, which integrates and automates development and IT operations to improve and shorten the development lifecycle. But, as opposed to focusing purely on development and operations, platform engineering enables self-service capabilities and automated infrastructure operations.

“It’s really the maturity of DevOps in the AI age,” said Erin Boyd, distinguished engineer and director of emerging technologies at Red Hat. “Platform engineering provides an opportunity for a cultural shift in our industry to be able to more keenly understand the needs of all the people involved to make the work successful.”

Platform engineering explained According to Gartner, the platform engineering process starts with a dedicated product team that creates and maintains a platform that supports the needs of developers and others. This platform provides common, reusable tools and capabilities and interfaces to complex infrastructure.

Platforms depend entirely on end-user needs in specific situations, and the overall goal is to modernize the enterprise software delivery process and improve the developer experience and productivity by enabling them to run and manage their applications independently. The practice also helps ensure reliability, security and talent retention, experts say.

“The development of a new generation of tools has made platform engineering one of the hottest topics of conversation within the DevOps community,” said Paul Delory, Gartner VP analyst. “These tools aim to make building and maintaining platforms easier.”

In fact, Gartner named platform engineering one of its top 10 strategic technology trends for 2024, and the firm estimates that by 2026, 80% of large software engineering organizations will have established platform engineering teams that provide reusable services, components and tools. Delroy noted that the practice can help solve issues around cooperation between developers and operators.

Red Hat, for instance, recently worked with a large bank in Australia and New Zealand to design a developer self-service-based OpenShift platform. In eight months, the financial institution sped up application onboarding through self-services, reduced wait times, internal tickets and errors and achieved successful migrations  in specific time frames without experiencing business disruptions.

Platform engineering is a “really good blend” of DevOps and site reliability engineering, which focuses less on the end-to-end lifecycle and more on the delivery and stability of production environments, said Boyd. Prior, those two areas were very distinct; platform engineering can create better integration between them.

Ultimately, platform engineering helps ensure that developers can get to work as quickly as possible, while also ensuring that the appropriate guardrails are in place to adhere to company policies, she said.

“It's a new way of looking at how we can mature out the models and processes we've been using for years in a way that is more efficient, and more aware and provides more impact back to the organization,” said Boyd.

Platform engineering helping to operationalize AI Generative AI is moving fast, as is the hardware that supports it.

“We see a lot of potential for improved development cycles for better automation, better decision making through data that wasn't possible before,” said Boyd.

However, there are concerns around data exposure (specifically, proprietary or sensitive personal data), which can occur when models are pushed out too quickly, or used incorrectly or even nefariously.

Platform engineering can help securely support the operationalization of AI, Boyd noted. The goal is to support speed and innovation in a way that is safe and protects workloads and data (and at reasonable costs). Organizations must also have the ability to deploy AI in a hybrid fashion — in the cloud, on-premises, or a mixture of both — and constantly train models.

“With all applications, once you deploy, the work is not done,” said Boyd.

With AI that's no different: Once models are in production and applications are using them, how are they being maintained? How are bias and drift being measured?

It’s important to work toward “secure by default” platforms that allow organizations to monitor data over time to ensure models suddenly haven’t introduced unnecessary bias, or that they haven’t been poisoned or fed false data to work around their safety mechanisms, said Boyd.

Blending SRE and DevOps in one platform can centralize all these processes that need to be managed across application lifecycles, she said. It helps ensure guardrails are in place at inception.

“Platform engineers will really be key in protecting the organization and providing that onramp to use generative AI safely,” said Boyd.

Building a product management culture Platform engineering looks different for every company, experts point out, but Gartner offers some valuable, high-level advice.

Efforts typically begin with internal developer portals (IDPs) — which provide a curated set of tools, capabilities and processes — as these are most mature. Subject matter experts select and package them for easy use by development teams.

It is essential, Delory said, to build platforms that have reusable, composable and configurable components, knowledge and services. Also, he advised, “treat the platform as a product,” by working with users to identify and prioritize the technical capabilities, tools and processes that are most valuable to them — and then build a platform around those.

Also critically, organizations should build a product management culture in which routing collaboration occurs between platform engineers and end users to “share bidirectional feedback in a safe and productive environment,” Delory said.

Ultimately, said Boyd, platform engineering is critical for enterprises — particularly larger ones — to be successful in the age of AI.

“It's a great way to adopt a new full stack development point of view and understand all the concerns,” she said. “The more you understand about the application down to the hardware all the way up to how the user uses it, it becomes a much more effective way to meet the needs of customers.”