Why your AI survival plan should start with a smile
Right now AI disruption in the media often feels like a game of Whack-a-Mole. Organisations, or individuals within them, respond when a challenge or a new tool pops up.
I analysed 725 cases of AI adoption by media outlets in 80 countries in recent years and discovered that most of them attempt to automate single repetitive actions or augment human work, but without strategic intent. (See Figure 1 for details).
On the scale of 1-5 of adoption complexity from lowest to highest, 71% of projects analysed are stuck at stage 1-2 (See Figure 2 for explanation of adoption complexity stages, adapted from an Economist podcast and originally used by the pharmaceutical company Moderna).
It means that these adoption experiments are not allowing media outlets to achieve major efficiency gains, and fully use the transformational opportunities brought by AI-based technologies.
Figure 1. AI adoption stage vs strategic intent in media
Figure 2. Explanation of AI adoption stages
1AccessTool exists but not in regular production use, team has access2AdoptionTeam or department using the tool regularly3ProficiencyMeasurable KPIs tracked; iterating on outcomes4Ways of WorkingAI rebuilt into editorial workflows; roles changed5ReorganisingNew business models, structural org changes, or entirely new products enabled
There is a better way of dealing with the AI disruption: smile.
I am not suggesting leaders adopt a spaced-out grin every time they hit an AI challenge. The idea behind the smile is to create a simple metaphor that would help them visualise the only two strategic priorities that I think really matter, and a process that underlies both.
One of these priorities is to reinvent their value proposition. Journalism needs to do a lot of soul-searching to understand what it gives the audience that is so unique, so precious and so visceral in people’s lives that no amount of robot intelligence can replace it, now or in the future. And then reconfigure the teams to keep delivering on it ruthlessly.
The second priority is to automate everything else, using the violently erupting technology. Every operational aspect that does not fit into the human-led value creation box needs to be outsourced to machines.
To enable both tracks, leaders need to bake learning and innovation into their companies in ways that are uncommon for the journalism industry. But neither developing human excellence, nor technical R&D are any longer optional if the media want to thrive – or even survive.
Double down on unique.
There is a peculiar zeitgeist in the industry. Leaders believe that their outlets have a fighting chance despite compounding disruptions, while journalism as a whole less so. I think that mostly the opposite is true. For the outlets to survive they need to go back to the essence of journalism, the core value it delivers, and reinvent and reassemble themselves around it for the AI era.
But journalism will remain.
It’s sometimes referred to as the second oldest profession in the world (although it seems to compete for the spot with spies, politicians and gigolos). The reason it might be up high in the ranking is that it serves deep human needs, and has certain features that AI can’t touch.
Attempts at identifying those features are ongoing. “I think we are still going to have journalists because we are going to [need] editorial opinion and relationships and things that don’t seem well-proxied by models,” says Sarah Guo, an AI investor.
Daniel Hulme, Chief AI Officer at WPP, an advertising industry giant, talks about “asking great questions” and creating “authentic content,” rooted in empathy, grasp of the big picture and an ability to creatively personalise it.
Within the industry the consensus is that media need to deliver more original investigations and on-the-ground reporting, contextual analysis and explanation and human stories, while cutting back on service journalism, evergreen content and general news.
Ethan Mollick, an AI researcher and author of the book “Co-intelligence” in his research suggests that we need to analyse bottlenecks for what AI cannot and will not be able to do (at least in the foreseeable future) to arrive at sustained competitive advantages.
A system, however intelligent, cannot confront an official or a corporation based on a tip-off from a trusted source. It cannot feel the crinkle of a frosty leaf underfoot and a tingle on the tip of its nose on a spring morning walk. It cannot witness, it can recap.
What AI cannot do clusters around three distinctly human domains.
The rest needs to be automated.
Figure 3. What is AI actually used for?
Stanford professor Erik Brynjolfsson says that every time a new technology comes along, you need to rethink how the economy is run: “If you simply pave the cow paths and put the same technologies on top of the old way of working, you don’t really get the business benefits.”
To achieve real efficiency companies need to break all jobs up into tasks, and outsource the tasks that can be automated. Existing workflows (and business models) need to be reassembled around the new human/machine job divides.
There are media organisations trying this: After training reporters in AI use and developing a proprietary tool, UK’s Newsquest has managed to achieve mind-boggling productivity gains: from four stories to 30 on average per reporter per day.
But new workflows can be tuned to other goals than production bulk, and should be built for multiple business functions. In journalism, discovery and use will most certainly be affected more than production. But right now disproportionate effort goes to editorial experiments.
In the pool of 725 AI adoption cases I analysed, editorial production and workflow cases account for 66%. This is many times higher than audience experience experiments.
More complex still, media will need to go beyond the current ways of understanding workflows to make a huge leap from adopting tools for current tasks, to reorganising completely (see Figure 3).
Figure 4. Where media are vs where they need to be with AI adoption
As a first step it would be worth considering reshaping companies into small, cross-functional teams that would drive experiment and innovation. They would need to include engineers, editorial product people and audience experts to succeed.
While experimenting, organisations need to get used to conducting true cost/benefit analysis of automation (and analyse their future projections based on an assumption that technology companies will eventually start charging for true cost of compute), before irreversible automation decisions are made.
Education and innovation enabling tech and journalism
At a recent discussion of media managers in exile about AI disruptions an accomplished founder of several niche media complained that it’s very difficult to expect journalists to learn all the time: “We can talk about it all we like, but they will resist, and they truly have no time.”
But another one chimed in, saying that “learning needs to be akin to basic hygiene, sort of like hand-washing after the toilet.”
Managers globally say that their teams are experimenting with AI tools, but this process is often disorganised and varies widely between individuals, often depending on personal enthusiasm for the new technology.
Instead, it needs to be managed. Leaders need to set up microlearning opportunities to achieve knowledge transfer inside teams, as well as organise deep dive trainings. For example, a regular team meeting can have a 10-minute slot where a certain prompt or a tool is presented, and deep-dive workshops can be set up at a pace that suits a specific organisation.
Some outlets are already going down this path. The Guardian recently announced that they rolled out a mandatory AI course for the whole staff that “goes beyond dos and don’ts – explaining how AI works and the science behind it.” They expect to expand this program in the future.
Experimentation also needs to be tied to practical outcomes and specific strategic goals to make sure the whole organization is working towards automating processes and overall transformation, not tackling individual tasks.
During the same management discussion another person said that the gap between journalism and technical jobs is narrowing, and suggested it needs to get narrower still.
She is not unreasonable to expect that job descriptions will evolve: it’s already happening in other highly-disrupted industries. For example, coding roles are rapidly changing to include more project management and supervision duties, effectively pushing people to adopt more skills and embrace a bigger-picture view of their jobs. This expansion is also needed in journalism.
Aside from that, journalism needs to relearn good old reporting and storytelling in a process that a recent WAN-IFRA story aptly labeled as “saying goodbye to air-conditioned journalism.”
In recent decades a greater focus on social media rewarded skills related to succeeding in the multiplatform environment, often pushing human connection online. Brief formats contributed to poorer ability to see the story beyond retelling the main talking points. Proper storytelling gave way to lengthy podcast-like dialogue formats.
At a recent workshop I conducted in Warsaw I watched six seasoned journalists and editors – some with over a decade of experience – practically crippled with an assignment of writing a single paragraph based on direct observation. They kept defaulting to factual statements and analysis, struggling to lean into empathy to tell a story.
But, to break through machine slop, journalists need to captivate the audience with the story, and cultivate complex relationships to achieve it. Also, to supervise any machine-produced content humans need the kind of editorial judgement that comes with years of mistakes and rewriting.
The makers of major AI models like Claude and ChatGPT say that their models either already are, or will soon be, operating in a new mode: recursive self-improvement. Human journalists must do the same.
The danger right now isn’t that news outlets will not make changes – AI disruption is taking care of that. It’s that they do half the smile. They might automate aggressively but skip the harder work of building what machines cannot replace. Or they perfect the human craft while leaving operations sloppy. Both tracks need to run at once. Otherwise, it’s not a smile, it’s a grimace.
About the author:
Katya Gorchinskaya is a media and strategy consultant with three decades of experience in editorial leadership, corporate governance, entrepreneurship, and institutional transformation. She works globally with media and educational institutions to modernise journalism, media education, and business models.
Over the course of her career, Katya has worked as a CEO and editor with leading media organisations in Ukraine (such as Hromadske TV and the Kyiv Post), and as a journalist for international titles, including Forbes, the Guardian, and the Wall Street Journal. She holds a double MBA specialising in leadership and technology from Ecole de Ponts, Paris (Dean’s List), and EADA, Barcelona. She is currently writing the first modern textbook on western-style journalism for Central and Eastern Europe. LinkedIn profile.
