Much of the AI conversation in news media often starts from a defensive posture: Traffic is fragmenting, search is shifting, subscriptions are plateauing, ad revenue is volatile, costs are rising.

But stepping back over the past three modules of our master class, a different framing emerged. This is not primarily a cost story. It is a capacity and growth story.

Agentic AI — autonomous, multi-step, action-oriented systems — represents a structural shift in how work gets done and how value is created. These systems don’t just assist. They plan, orchestrate tools, adapt, collaborate, and execute. Increasingly, they will sit between users and the Internet itself. There are many knock-on effects.

Agents are starting to change the interface (see an example below), which is likely to change the economics. We are moving toward an agentic layer where agents become the default expression of how users interact with information, services, and commerce. That makes this more than a newsroom conversation. It is a business model conversation.

At the same time, journalism itself is evolving. It is no longer sufficient to think in terms of finished articles or isolated formats. Journalism increasingly needs to function as structured, machine-readable infrastructure. Attribution, timing, context, and editorial logic must survive inside AI systems.

Several examples showed what this looks like in practice. Stories broken into atomic units (sentence-level components with clear metadata and separation of fact from analysis) allow a single source of truth to power multiple outputs without distortion. When journalism is structured this way, it becomes reusable, adaptable, and resilient inside agentic environments.

This is not theoretical. Structured journalism is already being packaged as a product, just look at AP Intelligence. Context and metadata are becoming commercial capabilities. Trust is no longer just a brand attribute; it is a system requirement.

The interface shift reinforces this. As we move from user interfaces to agent interfaces, value shifts upstream. It moves away from clicks and pageviews toward structured access, control, contribution, and machine-readable rules. In that world, competitive advantage comes from infrastructure: who controls access, who owns context, who can verify provenance, who can measure and monetise usage.

That infrastructure thinking is already driving execution. Some organisations are embedding agentic systems directly into engineering workflows, increasing feature velocity and reducing friction. Autonomous agents mirror team structures — planning, building, testing — with humans governing the loop. The outcome is not job replacement (for growth minded companies); it’s expanded capacity.

Investigative journalism is being strengthened through layered agent architectures that handle data collection, cleaning, statistical modelling, and anomaly detection before a human editor interprets the findings. Institutional knowledge — archives, transcripts, notes — is becoming computational. Instead of static history, it becomes queryable intelligence. Raw content turns into structured advantage.

AI cannot remain a side project focused on efficiency. The real opportunity lies in growth: API-first infrastructure, service-based revenue models, and inference economics built on Rights → Access → Payment. Monetisation frameworks are being shaped now, particularly around real-time inference and structured access, and there is still everything to play for.

Distribution is shifting, too. AI-mediated traffic is more qualified and more intentional. That makes control of access, authentication, and licensing central rather than peripheral. Standards around attribution and agent authentication will determine whether monetisation is enforceable or aspirational.

On the audience side, product innovation is focusing on friction. Personalised, adaptive formats that respond to time constraints and allow in-flow interaction are increasing engagement. But they also raise the bar for accuracy. Testing AI products is not the same as testing traditional features. Error categorisation, benchmarking, and transparency become core product disciplines.

Across all of this, the pattern is consistent. AI is increasing capacity.

→ Capacity to build faster.

→ Capacity to analyse deeper.

→ Capacity to personalise intelligently.

→ Capacity to monetise in new ways.

And capacity creates growth.

The organisations that treat this as an infrastructure moment, not a feature moment, will move differently. They will invest in shared foundations rather than one-off tools. They will collaborate on standards. They will structure journalism for machine readability. They will design monetization around inference and service, not just licensing headlines.

None of this removes uncertainty. Models will continue to evolve quickly. Economics will continue to shift. But the direction is clearer than it was even a year ago.

We are entering a period where institutional intelligence, structured journalism, and agentic systems converge.

This is not about surviving disruption. It is about redesigning for advantage.

We have the tools.

We have the examples.

We have a window where the rules are still being written.

The question now is not whether AI will reshape our industry, it’s how. The opportunity is real, and it’s ours to build.

If you’d like to subscribe to my bi-weekly newsletter, INMA members can do so here. 

Banner photo: Adobe Stock By Deemerwha studio.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .