Big Data LDN 2026, which took place on 23rd and 24th September at Kensington Olympia, opened with the theme of FROM OUTPUT TO OUTCOME - The Good, the Bad and the .
The talk of the show floor this year was the context layer and agentic AI in production. In previous years, BDL conversations centered around Generative AI, then agentic AI, and now it’s all about the ontologies and putting those AI investments into production.
Essentially, the data wizards answer what AI needs from your data before it can be trusted with real work. It came up in the keynotes and on the exhibition floor, and again in nearly every interview we recorded.
The UK's biggest data, analytics and AI event ran with 411 speakers listed, 16 theatres and more than 200 exhibitors. This year's debut was the AI Layer, a dedicated area that put AI vendors side by side so attendees could compare products and put questions straight to the people building them.
The keynote programme went well beyond the technology itself. Andy Fletcher, VP of Technology and Digital Products at Liverpool FC, covered how the club uses AI to understand and engage supporters worldwide. Ritika Jain, Sr. Director of IT Operations at AstraZeneca, looked at physical AI and autonomous manufacturing, where systems act on machinery rather than just analyse information. Karthik Ravindran of Microsoft closed the loop with AI Hope: A Future Worth Building with AI.
The week had opened a day earlier with Data Driven LDN, the smaller, senior-level conference, where Professor Marcus du Sautoy, Simonyi Professor for the Public Understanding of Science and Professor of Mathematics, University of Oxford, gave the keynote on AI: Promise or Peril?
This was our third year as an official media partner of Big Data LDN. Our team spent both days recording podcasts and short vox pop interviews with data leaders, solution providers and analysts.
This included a live podcast event with Striim at the Y-Axis Theatre on Revisiting Enterprise Data Pipelines For Agentic AI: An Edgy, Shift-Left Live Podcast. This live BDL special episode was hosted by Christina Stathopoulos, Founder, Data & AI Evangelist, Keynote Speaker, Educator, Dare to Data, with guest Alok Pareek, Founder and EVP of Product Development at Striim.
All interviews can now be watched on our Video Resources page, and here is what stood out.
Context Dominated at BDL 2026
“If Context is Your Moat, It Should Be Open Source,” said Suresh Srinivas, the CEO and Co-Founder of Collate and OpenMetadata, in an interview with EM360Tech’s Shubhangi Dua, Podcast Host, Producer & B2B Tech Journalist.
“I see three themes, of course: one is context, then agentic AI and finally, trusted data,” he adds. “The general consensus is that context is very important for enabling AI, but there are lots of words being thrown around in the context salad, such as metadata, semantics and ontologies,” he responds.
Collate specialises in the context layer. Srinivas tells Dua that Collate has a comprehensive view of what a context is, which they define by highlighting 3 Ms: Metadata, Meaning and Memory.
Meanwhile, Phillip Miller, AI Strategist and Director of Product Marketing at Progress Software, tells Dua that the most exciting conversations on the BDL floor have been about how enterprises are trying to solve world challenges using AI.
Progress Software looks at solving these world challenges with three ‘Cs’: Context, control and confidence.
“It’s become more critical to bring that data together so the data can be connected to policies, governance, ontologies, or taxonomies and everything that goes into turning that data into information,” Miller said.
When Vector RAG Stops Working
Why retrieval strategy now hinges on matching Vector, Graph and hybrid RAG to the questions AI must answer across complex enterprise data.
Paul Kryla, Global Director, Field Engineering, Matillion, also voices how foundational AI deployment relies heavily on high-quality data and rich context.
“It all wraps around context. Data tells us what happens and context tells us why it matters,” Kryla tells Dua, alluding to what CEOs in the enterprise need to pay attention to.
Data Architecture AI Risks & Unstructured Data
Reece Williams Griffiths, Field CTO at Collibra and former Co-Founder and CEO of Deasy Labs, tells Dua that many in IT and data architecture foresee the risks of attaching copilots or agents to very highly unstructured repositories in SharePoint and OneDrive.
Furthermore, Griffiths points out to Dua that these teams currently lack both the necessary budget and authority to make substantial investments in a knowledge management initiative of this scale. A major reason for this shortfall is that "there isn’t a direct tie to the use case."
“The reality is that enterprises will only invest in agentic AI systems once they’ve deployed AI agents and seen the consequences of hallucinations,” he adds.
Meanwhile, Ole Olesen-Bagneux, Vice President & Chief Evangelist at Actian, brings it back to his panel discussion at the event on “Where is the Context?” He tells Dua, “It’s interesting how we make it real by providing technologies and setups to capture context, and technologies that use pieces of context to instruct AI agents.
The Actian VP adds that he sees huge potential in unstructured data. “Unstructured data has not been harnessed enough.”
“The vast majority of data in enterprises is unstructured. We can use it for AI use cases. It has the biggest potential. We have the poorest-defined engineering disciplines around unstructured data.”
Inside the Agentic SOC Stack
See how unified telemetry, correlation engines and agentic AI workflows rebuild SOC architecture for autonomous detection and response.
Tech Stack is the Way to Go, Not AI Stack
Dua also interviews Sanjeev Mohan, Principal at SanjMo and former Research Vice President for Data & Analytics at Gartner and Author of Designing the AI-Driven Data Foundations: Architecture, Principles, and Practice. He tells Dua that the power of AI now exceeds human imagination, but “we are using a small portion of its capabilities.”
“CEO’s should look at AI with an open mind. AI is not a replacement for the good old deterministic SQL queries. It has the power to really transform every aspect of a business,” he adds.
Mohan points out a common misconception in enterprise management regarding the notion of an "AI stack." Rather, he emphasizes to Dua that the reality is simply a "tech stack."
“There’s only one stack with data as the foundation, and AI is on top of it,” he says. “It's a wrong move when someone equates AI governance to data governance.”
On a more positive note, Cooper Kenny, Founding Sales and Alliances Manager at LakeFusion, tells Dua that instead of having a tool for ETL or a tool for ML or a tool for AI, enterprises and their CEOs are consolidating their spend for a more modern tech stack which would allow them to see, read, and have their data communicate back and forth more effectively.”
Kenny also talks about the growth in the industry over the last two years and what’s expected for the next few years, and the impact generative AI will have on it. “Now the question is what’s the adequate technological fit to get there.
Avoiding The Next AI Winter
Learn why unchecked hype, weak controls, and poor data discipline still threaten AI value, and how leaders can close the gap.
SaaS products have exploded over the last decade. He says to think about the growth of centralised data platforms, the hyperscalers such as Databricks and Snowflake. “What these vendors have done really well is build specific customer-centric data products within their data platforms.”
“Instead of having a tool for ETL or a tool for ML or a tool for AI, enterprises and their CEOs are consolidating their spend for a more modern tech stack which would allow them to see, read and have their data communicate back and forth more effectively,” he tells Dua.
Tokenmaxxing is Dead
One of the key insights at BDL that most IT leaders agree is that tokenmaxxing is dead. Jane Smith, Field Chief Data & AI Officer for EMEA at ThoughtSpot, says that the industry has moved from the theoretical promise of "Agentic" AI to practical implementation, focusing on context, semantic layers, and managing token costs.
The mindset is moving to enterprise economics. “The enterprise economics of token costs, also referred to as tokenomics, are the economics of engineering costs to support the token costs,” Smith says.
“The bill has come due, and people are reflecting on the past when they claimed how AI would save them money.”
“Tokenmaxxing is dead,” Smith asserts. There are many ways to optimise for tokens. Enterprises should build proper architecture and infrastructure first. Also, they shouldn’t bypass good engineering principles.
Comments ( 0 )