Sitecore shifts to a composable architecture and focuses on AI innovation

If you want to know how Sitecore is embracing generative AI, there's no better person to talk to than Dave O'Flanagan, Sitecore's Chief Product Officer. O'Flanagan took some time to tell me about the changes to Sitecore's architecture and how they are implementing AI within their platform to support customer needs.

O'Flanagan joined Sitecore almost three years ago when the firm acquired Boxever, an enterprise CDP and personalization‌ platform. He was the CEO and co-founder of Boxever, and with the acquisition (one of three others around that time), he became the Chief Product Officer (CPO), reporting to Steve Tzikakis, CEO. 

From on-prem to cloud and composable

Sitecore knew it didn't want to rewrite all its products as cloud-native, so it acquired several core strategic products and rewrote some Sitecore products. The result was Sitecore DXP, a composable, fully native cloud solution that enables customers to pick and choose the capabilities they want when they need them.

O'Flanagan was part of the leadership team driving this new composable strategy under the direction of Tzikakis, who joined the company as CEO in September 2020. He said Sitecore is uniquely positioned against its peers by offering a completely end-to-end cloud-native solution that is headless and offers flexible implementation and deployment. 

It's a significant change for Sitecore, which is known for being a monolithic platform that was challenging to work with and often required hard-to-find, experienced resources. According to O'Flanagan, the last two years have been challenging but rewarding from a technical perspective. The company has done a lot of work re-educating customers, partners, and the market. 

The other big change O'Flanagan talked about was that customers no longer need a lot of Sitecore .Net developers to build their experiences. Now that Sitecore is headless, customers can access standard web APIs and build their experiences in modern languages like Javascript and React - something that opens customers to an entire world of development resources. 

Sitecore isn’t alone in this challenge to shift to the cloud and a composable strategy. The entire DXP (digital experience platform) market has been traditionally on-premise and slow to evolve, but things are finally changing.  

AI has always been part of the Sitecore platform

Naturally, our conversation shifted to AI, and we talked about how it had been a part of Sitecore long before the hype of generative AI. In the past, Sitecore leveraged AI to analyze customer behavior and provide insights and recommendations. Boxever also used AI and machine learning for automatic testing and personalization. So, AI isn't new to Sitecore. 

O'Flanagan said that Sitecore believes that AI should support a feature rather than be the feature itself. He said many startups are conflating 'I've got AI' with 'I've got a solution to a business problem',  whereas he argues that Sitecore is trying to solve business problems more efficiently by integrating AI deep into its products.  

Generative AI has been a whirlwind for many, and Sitecore is no exception, but it provides many innovative and new ways to solve customer problems, he said: 

When we think about generative AI and how we are implementing that in our solution, we're really thinking about how do we introduce better efficiencies in the workflows of our marketers or in the interactions of our customers, so that things become easier or more relevant or more simplified?

Improving content creation is one way. Sitecore manages the entire content lifecycle, from ideation and planning to creation, production, and delivery. And they've spent time understanding how AI impacts that work, he explained:  

If you think about the first stage in the workflow: I want to create a campaign for a brand to sell more flights online. So whatever that is, we want to be able to create ideas around that leveraging ChatGPT, and almost instantly, you can imagine using ChatGPT to give you ideas. So I think the key thing as we think about introducing AI is leveraging these large language foundational models from Open AI and Microsoft, and then using prompt-based interfaces to influence or improve the interface.

Another area where AI can help is content production, including creating a first draft or creating images based on stored assets. It's more than creating assets, though. It's also creating derivations of assets, summarized content, SEO tags, and so on. And while generative AI can do a lot, O'Flanagan said, you still need to have guardrails and controls in place to ensure what's created is appropriate and correct and does not expose customers to copyright challenges.

A third area is delivery. O'Flanagan said you can use generative AI for personalization and search. He explained the work they are doing with search, where they are combining the power of LLMs with first-party data to make more powerful recommendations even when they know little about the customer. 

The key takeaway is that Sitecore isn't trying to go big and do everything at once with generative AI. A partnership with Microsoft on OpenAI capabilities is critical to Sitecore moving in the right direction, as well as other partnerships. They are implementing basic functionality and, at the same time, testing new features in their labs with customers to see where the actual value is. O'Flanagan said: 

The other thing I've learned in my years using doing personalization, conversion rate optimization at Boxever - you never know exactly what's going to work, right? You could think that the product is awesome, but maybe the less sophisticated version has more utility. So I think one of the things that I'm keenly aware of as a product leader is that it's not yet clear what the right capabilities are. I think there's lots of opportunity. I think there's a lot of hype around this. We don't quite know yet what the right features are. So we're entering a phase in Sitecore where we're going to be super iterative and responsive.

The user experience is also top of mind for O'Flanagan. He said Sitecore doesn’t want marketers to have to be prompt engineers. The interface needs to be simpler, taking the prompts in a way that enables marketers to feel confident and competently interact to get the value they need: 

So there's a lot to think about, I would say, and it would be remiss of me to say we've got it all solved. But I think what I want to do is create a framework of fast innovation and iteration in partnership with Microsoft so that we can get these features into the hands of our customers as quickly as possible and then learn from how they're using them so that we're creating the features that generate the most utility.

Because I've seen in previous lives working in AI companies and things that you can go super deep on this stuff, and you can be really focused on eking out the next two, three, five percent out of the model efficiency, but that doesn't necessarily translate into business benefit, right and just trying to get the right business benefit versus technology effort. I think that's the tricky bit at the moment. But the fact that we can leverage these elements so easily by APIs is a real door opener to innovation, which is awesome.

Sitecore doesn’t want to provide all the AI ideas. It wants to provide the capabilities, platform, and infrastructure so that ‌builders can build things quicker and innovate themselves. That means they will bake AI into products where it makes the most sense. 

My take

O'Flanagan said that customers want to know that Sitecore is working on generative AI capabilities in the DXP, but they still have concerns about IP exposure and issues like hallucinations. 

Implementing basic capabilities into the DXP that help the marketer's workflow are quick wins in terms of introducing AI into the product and helping customers find value without worrying that something will go wrong. But it's just the start, and Sitecore is doing some interesting work with partners like WPP, Microsoft, and Nvidia that will take its products to new levels of generative AI use. 

It's good to see Sitecore is taking its time to do things right. Every technology vendor needs a generative AI story today to be considered a feasible option for buyers, but slapping AI in just to say they have it isn't the right path forward. Get the basics in place and start testing and trying things. Then, bring the right innovations to market.