News publishers should think of agentic AI not as another tool but as an entirely new workforce capable of scaling subscription operations across the funnel — from onboarding to retention — according to speakers at the INMA Media Subscriptions Summit in Toronto.

In a joint session, Robert Whitehead, lead of the INMA Digital Platform Initiative, and Débora Pradella, executive manager of digital product at Grupo RBS in Brazil, outlined how agentic AI systems can expand productivity for news organisations while enabling faster experimentation across product, marketing, and user experience teams.

Whitehead framed the shift as a fundamental change in how news companies operate.

“What would happen if you had an extra 1,000 people?” he asked the audience. “I’m going to call them members … . Think of them as team members, and they are here to grow your productivity.”

Rather than single AI tools performing isolated tasks, agentic systems coordinate multiple specialised agents that can monitor user behaviour, trigger actions, and optimise outcomes across the subscription lifecycle.

Moving from tools to agents

Whitehead explained that agentic AI sits on top of existing AI systems such as machine learning and large language models, acting as a layer of orchestration, Whitehead said.

“Agentic is a layer of orchestration; it’s not a type of AI,” he said.

These systems allow publishers to deploy multiple agents that work together toward shared goals. In a subscription context, one agent might track subscriber behaviour, another might trigger engagement campaigns, and another might coordinate onboarding experiences.

Whitehead compared this structure to an orchestra in which individual agents perform specialised roles but operate as part of a coordinated system.

“One agent does one task in subscription… but it makes this big story,” he said.

He challenged publishers to imagine assigning an agent to every subscriber in their audience.

“If you’ve got 80,000 digital subscribers, what would happen if you had one agent monitoring every single subscriber’s behaviour from the moment they click pay?”

Those agents could identify engagement gaps and trigger automated responses such as newsletters, product prompts, or re-engagement campaigns.

Rapid experimentation for subscription teams

One of the practical advantages of agentic systems, Whitehead said, is that they enable non-technical staff to experiment quickly without waiting for engineering resources.

“Anyone can now do rapid prototyping,” he said. “You don’t need to wait for that six-month delay… just go and try something.”

During the session he demonstrated how a simple AI-generated onboarding assistant could be built in minutes to guide new subscribers through steps such as downloading apps, signing up for newsletters, and setting preferences.

The goal, he said, is to accelerate the processes that drive habit formation and subscriber lifetime value in the critical first months after conversion.

Whitehead argued the technology is advancing so quickly that publishers must begin experimenting now.

“The progress has been so rapid in the last three months in this space,” he said. “Agents can now work twenty four seven.”

A case study from Brazil

While Whitehead focused on the strategic framework, Pradella shared how Grupo RBS is already applying AI agents inside its product and marketing operations.

Grupo RBS, the largest media group in southern Brazil, reaches 11 million people monthly across television, radio, newspapers, and digital platforms.

The company turned to AI after its UX team experienced a major reduction in staff, Pradella said: “Our team structure and our capacity was reduced by 40% … and we needed to gain scale and efficiency.”

At the same time, demand for product development and marketing campaigns continued to grow.

With only seven UX specialists responsible for more than 20 digital products and the entire subscription funnel, the team began using AI agents as additional team members, as Pradella described them: “We see AI as a new teammate, a new coworker that works with us every day.”

Scaling marketing and product workflows

The RBS team first tested more than 40 AI tools before redesigning its processes to integrate AI directly into daily workflows.

“We shifted from that unstructured usage to a structured, controlled production for campaigns with AI agents,” Pradella said.

A crucial step was creating extensive documentation to train the agents: “AI only scales with strong documentation,” she said.

Today the company uses agents across several functions including:

• Summarising meetings and operational updates.
• Brainstorming product ideas and campaign concepts.
• Validating design consistency and accessibility.
• Testing user flows and product features.
• Generating marketing copy for subscription campaigns.

One major application has been copywriting automation for subscription marketing. Because the organisation has only one UX writer supporting all campaigns, the team developed specialised agents trained on brand voice and campaign rules.

“We have multiple agents that work with us in all our subscription funnel marketing campaigns,” Pradella said.

These agents generate campaign content for channels such as e-mail and WhatsApp based on prompts describing the offer, audience, and campaign context.

Faster campaigns without losing quality

The impact has been significant. RBS now produces hundreds of marketing messages with AI assistance while maintaining performance.

“Last quarter we produced more than 300 e-mail campaigns with AI agents and more than 200 WhatsApp message campaigns,” Pradella said.

During promotional periods, production became three times faster compared with the previous year.

“Our campaigns’ KPIs remain stable when comparing our human-created campaigns and our AI-assisted campaigns,” she said.

Instead of writing every campaign from scratch, the UX writer now focuses on refining AI output: “Our copywriters shift from writing from zero to direction and curation of the AI agents’ production.”

A new role for humans

Despite the productivity gains, both speakers emphasised that humans remain essential.

Whitehead argued that managing AI agents requires clear goals, rules, and oversight.

“We actually need to treat these people in the same way that we treat humans,” he said, referring to agents. “Exact duties, clear KPIs and goals, and there can be no ambiguity.”

Pradella echoed that point, saying the technology works best when embedded into structured workflows: “AI works best when it becomes a system structure with processes, with models, not an unstructured usage.” 

Ultimately, she added, success depends less on the technology itself and more on how teams learn to use it:

“The secret isn’t the tools themselves but the team’s maturity in using them to generate real impact to our users.”

Photos by Robert Downs Photography.

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