Roshan Soni is Vice President of Engineering at Codal, a global digital consulting firm, where he leads the engineering, solution architecture and technology teams. With over 20 years of experience, he advises clients from the C-suite to development teams on technical strategy, solution architecture and organizational change across many different industries. At Codal, Roshan leads the company-wide practice for AI-augmented delivery. He helps enterprise teams move AI from proof of concept to production by building the content, data and workflow foundations it needs to scale.

Past the Pilot: Why Your Content Model Decides Whether AI Scales

AI pilots in content ops are everywhere, and very few of them reach scaled production. Teams try a drafting assistant, a translation experiment, or a chatbot over the help center. Then progress stalls. Review queues grow, output needs heavy rewriting, and the savings never reach the budget line. What holds them back is usually the content underneath the model.

Agents now create, review, translate and consume your content, and they can only act on what your content model makes explicit. A page built for how it looks gives an agent a blob of text. A model built for what content means gives it facts, relationships and rules it can reason over and act on.

This session shows how page builders, component-driven schemas and fully semantic structures change what agents can reliably do, and what each one costs the people who work alongside them. It then walks through four agentic workflows that fit day-to-day content operations: localization, content audits, reuse across channels, and agents as consumers of governed content. For each one you’ll see the business outcome. You’ll leave with a readiness scorecard for your own content and a practical path from AI experiment to execution.

Here’s what you’ll learn in this session:

• Why AI pilots stall in content ops. The three failure modes behind most stalled pilots: agents working from unstructured pages, business rules that live only in people’s heads, and AI output with no workflow to land in.
• Semantic structure is what makes content agent-ready. How modelling content by meaning lets agents find, update, translate and reuse it safely. It also makes content easier to govern, easier to personalize and easier for editors to manage.
• Content, context and workflow. A three-layer view of agent-ready content operations. Structured content is the foundation, a machine-readable context layer tells agents the rules, and designed workflows put people at the decision points that matter.
• Principles and a scorecard to act on. A short set of modelling and workflow principles, plus a self-assessment you can use to find your first workflow and measure the cost per piece of content before and after.