Artificial intelligence (AI) is increasingly being positioned as the key to faster software development, smarter customer experiences, and more efficient operations.
However, as organizations rush to build AI-powered applications, there is a growing recognition that success depends not only on the technology itself, but also on the controls surrounding the technology. The challenge is no longer just how to use AI, but how to use it safely and reliably in a way that aligns with business goals and customer expectations. Building apps with AI should make the process smoother, but humans need to be involved to add guardrails to development and ensure apps work safely and as intended.
At the Summit in London in March, Datadog, a supplier that provides observability services for cloud-scale applications and monitors servers, databases, tools and services through its SaaS-based data analytics platform, promised its audience to demonstrate its knowledge and prowess in the use of AI and show how it believes AI’s ability to power modern business operations is possible.
With AI capital spending expected to reach $725 billion in 2026, this surge in investment is driving business transformation as organizations increase spending and restructure operations around AI.
At the Datadog Summit, Yrieix Garnier, vice president of product management, said the company has launched “many types of AI agents” to help provide context, identify issues, and recommend fixes, as each additional change introduced increases “stress on the system.”
“It’s very siled, repetitive, and quite time-consuming,” Garnier says. “This is something Datadog is already solving. We help our customers close that end-to-end loop, allowing them to continuously monitor cycle stress in their systems. We want to give them the right information to detect problems and the information they need to remediate them.”
Governance requirements
The company made the announcement in London, specifically about the presence of its UK data center, along with the launch of its new site. Datadog says this will help customers meet data governance and security requirements as demand continues to evolve in the wake of questions surrounding Europe’s digital sovereignty.
In a recent London Stock Exchange Group survey, 82% of businesses surveyed said they operate in a multi-cloud or hybrid environment, and businesses are adapting to changing data governance requirements in the UK as cloud adoption continues to accelerate across regulated organizations.
Garnier said the company is “investing in adding more AI to our products to ensure we can provide a correlation of what’s going on in the environment. In doing so, we can cut through the noise and significantly reduce resolution times.”
This increased use of AI not only brings AI to infrastructure monitoring, Garnier argues, but also provides a more automated, more complete view of the entire property. “It’s really about being able to quickly detect and remediate what’s going on within your Kubernetes environment,” he says. “When you understand what’s going on within that environment, you can not only make good recommendations, but also apply fixes on top of that to change the environment.”
Introducing the concierge
At the summit, Mark O’Neill, senior manager of AI software engineering at Datadog customer Virgin Atlantic, will discuss Virgin Atlantic’s AI achievements. The company’s investment and development centered around chatbots for websites, with the goal of making chatbots more than just answering questions or guiding visitors through a series of prompts to provide real assistance.
O’Neill described the concept of implementing a customer-facing AI chatbot as “challenging.” Especially since it’s only been 90 days since launch, “especially when brand reputation is so important.”
Virgin Atlantic’s brand is built on service, personality and trust, so the concierge chatbot needed to fit these parameters and support help and Q&A, flying club queries, flight searches and holiday discovery, he said, adding: “For travel planning, rather than just prompting LLMs, we looked at how our frontline teams support customers and built those patterns directly into concierge.”
Negative experiences with AI-powered chatbots and their failure to solve problems rather than send users in circles are creating a need for companies like Virgin Atlantic to build better chatbot experiences.
Mr O’Neill said that while traditional chatbots guide users along a set path, generative AI (GenAI) can effectively tackle any starting point of the journey, and this is a development that Virgin Atlantic is “most proud of”. In particular, the Holiday Discovery feature allows users to find flights to specific destinations on specific dates through the concierge. “The benefit of this technology is that it gives you the flexibility to have more diverse conversations,” he adds.
Developed using OpenAI, O’Neill admits there were three concerns about the concierge offering. One is providing incorrect information to customers, leaking personally identifiable information (PII), and making sure to portray a brand tone that is not “the real Virgin Atlantic.”
O’Neill said the decision was made for Concierge to not include any personal data within its systems to protect PII. “Therefore, we do not allow our models to access or process personal customer data,” he added. “Concierge has no account context, no reservation retrieval, no session memory associated with an ID, and only supports read-only operations.”
Perhaps keeping in mind the 2018 British Airways data breach, concierges can obtain information but cannot make transactions, change reservations, or update account details.
Do you want to build your own LLM?
Should companies build their own LLM to support the use of GenAI? O’Neill said that rather than building its own LLM, Virgin Atlantic used OpenAI’s models and application programming interfaces (APIs) as the foundation for Concierge and added a search augmentation generation (RAG) database that can answer custom prompts and common questions.
He said the company recognized that it didn’t necessarily have the knowledge or skills in-house to build the system, so it worked with TomorrowAI, OpenAI’s consulting arm. He says this gave him the knowledge and expertise he needed to move forward with the project.
“We had the experts in the room explain to us, this is how we can do it safely, this is how we can do it safely, and that was certainly something I pushed for,” O’Neill said. “What we recognized as a business is that we fundamentally believe in this technology and that it is here for the long term.”
So where does Datadog fit into Concierge? O’Neill says it functions as an end-to-end observability platform. From the front end to every interaction with OpenAI, “we have a complete trace of everything that’s happening. We use that to monitor, alert, and perform evaluations.”
O’Neill says Datadog is also used to check the accuracy of answers during testing and development, and is involved throughout the lifecycle.
AI and human elements meet
The message from both Datadog and Virgin Atlantic is clear. AI can accelerate development, automate operations, and improve customer experience, but only when supported by strong visibility, careful governance, and clear boundaries of what is allowed. Human oversight remains essential, including monitoring infrastructure, validating responses, and ensuring AI systems reflect your brand’s values and tone.
As companies continue to increase their investments in AI, the winners will be those that balance speed and control. Organizations that combine observability, security, and human judgment are best suited to build applications that are not only more capable, but also more reliable.
