How publishers are actively testing agentic AI to hike productivity

Agentic AI refers to systems that can autonomously plan, take actions and adapt to achieve goals, often across multiple steps or tasks. Publishers say they are testing these tools cautiously, exploring ways it can be used to strengthen existing services and bolster productivity rather than reduce headcount.

Currently, Hearst is running single-modal tests, which involve an AI agent carrying out specific requests step by step — like a digital assistant following a to-do list. 

The tests are being spearheaded by Michael McCarthy, senior director of AI, sales, and business solutions at Hearst. McCarthy has begun testing by giving tasks to a computer use agent, a form of agentic AI that can move through different systems, open applications, browse the web, input data into software or manage files on your computer. 

“If you’ve used a computer use agent, you can just say, ‘go research a bunch of accounts for me, or I’m going to log you into my LinkedIn.’ And the key is, you can get up and walk away. You can be in a meeting or multitask, and it’s just going to navigate your browser,” he said.

McCarthy said he has now started experimenting with early versions of products that use AI agents to create image-based results and is testing their reliability before scaling. 

The tests form the latest part of McCarthy’s strategy to use generative AI to “reinvent the sales process.” So far, that’s involved encouraging sales teams to use generative AI tools to assist them in their administrative tasks like CRM updating, and account research, pre-call planning, creating media proposals, and also informing them how to have smart, consultative conversations with advertiser clients about their needs. McCarthy first started testing generative AI tools to do this in late 2023, and they were rolled out to all 500 Hearst Newspapers sales reps throughout 2024. 

The tools are based on data, pricing models, audience insights and historical data. They mine Hearst’s internal data for client information, pricing structures and rules for how to budget and scale campaigns, and generate sharp media plans.

These tools have already created efficiencies. For example, the average amount of time it takes a salesperson to complete account research using AI is two minutes, significantly down from 40 minutes without it, McCarthy said. Around 500 people in the sales teams now use the tools, up from seven in 2024, he added.

McCarthy said anecdotally that his AI tools have helped the sales executives show up better to customers and answer their objections more effectively, with several saying it helped them close six-figure deals, and led to an average 153% increase in average sale value, though he declined to reveal specifics.

The team is “genuinely more energized,” he said. “It’s a morale booster that is translating to performance… I really do believe in the power of good that these transformative tools have for our people, and I’m seeing it everywhere they’re being used.” 

The tools are also proving to be a valuable tool for recruiting fresh talent, he added. The company no longer needs to only look for candidates with digital ad sales knowledge, but hire based more on attitudes, and then use the AI to onboard faster. “I can teach someone about frequency and reach really easily, but we have 30 products within our portfolio,” he added. But using the generative AI tools, McCarthy has created a training and onboarding ecosystem which people can now query to answer their questions. “It’s a personalized kind of product knowledge on-ramp and it’s working really well,” he said.

Agentic AI is still very nascent, though the buy side of the ad market has been more aggressive in pursing opportunities for streamlining to date – leveraging it to automate workflows and optimize spend. 

The optimistic narrative around AI remains unchanged: that it will automate routine tasks, freeing people up to focus on more creative, strategic and high-value work — even as fears persist about its potential to displace jobs entirely. For sales employees, this looks like having more time to spend on areas AI can’t yet touch: building trusted, longstanding relationships with clients and upselling. “At the moment, you’ve got so many humans across our industry stuck in spreadsheets, stuck in workflows, stuck scurrying around your office chasing people for stuff,” said Hamish Nicklin, former CRO of the Guardian. “AI can do all that, and it doesn’t necessarily mean you’ve got an opportunity to cut costs,” he added. 

While AI may save time, those gains are often spread unevenly across roles — making it hard for CFOs to justify job cuts based on partial efficiencies, Nicklin stressed. Instead, the real value lies in using those savings to drive growth at little additional cost.

McCarthy added: “That’s one of the kind of key wins, if people are enjoying what they do more, the sales reps don’t feel overburdened, they’re liking what they’re doing and when they’re happy, then they’re driving more revenue.”

Thomson Reuters is actively using agentic AI capabilities to automate complex tasks like reading and processing documents and extracting data, across its tax, accounting and legal client-focused services, following its acquisition of U.S. agentic AI startup Materia in October 2024.

Reuters has benefited from the early testing of its parent’s generative AI tools, and is poised to embed agentic AI agents more into its workflows, according to Reuters’ head of AI strategy Jane Barrett, though details are yet to be finalized. “We’re absolutely thinking about agentic AI, how news might fit into news agents and other agents too, like if you’re a professional in any business, how might news fit into your agentic workflow,” she said. 

Meanwhile, Belgian-headquartered publisher DPG Media has already integrated agent-based AI across its workflows. The media group, which operates a range of newspapers, magazines, TV and podcasts across Belgium and the Netherlands, has encouraged all departments to test out agentic AI capabilities via its internal ChatDPG AI assistant.

A handful of employees have built AI agents and roughly 3,000 employees already ask the agents questions/ give them instructions, 1,500 on a daily basis, according to Stefan Havik, chief digital officer of DPG Media.

Adoption has been largest in news and advertising departments for workflow and efficiency, Havik said. Agentic AI is integrated within DPG’s order management system and ad server, so sales reps can draft an email from the ad server directly to the client with all the appropriate information. And it can also deliver pacing information — data that shows how ad sales are tracking against projected revenue goals — on the fly.

For Heart’s McCarthy, the ultimate goal is to roll out the same processes across other groups like Hearst Health, Hearst Magazines and Hearst TV. “AI is here to stay and it’s going to keep the bell curve rising and rising. So we want to make sure that AI literacy is being solved this year with all people,” he said.