Generative AI has become a familiar topic across almost every industry, and the media sector is no exception. Many teams feel an expectation to explore new tools or test emerging capabilities so they do not appear to be falling behind.
This sentiment is often expressed through questions such as “We need to be using it” or “Where can we use it?”. When activity starts from that position, AI can quickly become a solution in search of a problem rather than a purposeful way to improve how teams work or make decisions.
Recently, Anna Mleczko, Head of Media & Advertising at Future Processing, sat down with Emma Wicks, Director of Customer Analytics & Data Science at The Telegraph.
During their discussion, Emma’s shared insight into her approach, which begins with identifying the real issue at hand and only then considering whether generative AI is the right tool to address it.
For media leaders, the question becomes less about how to use generative AI and more about where it can reliably help. The most effective applications tend to emerge in areas that are already well understood but have long been difficult to optimise, where the right intervention can strengthen decision making and reduce the day-to-day friction that slows teams down.
Creativity under constraint
Creativity in media organisations is often expressed through how teams approach everyday problem solving. With evolving audience behaviours, shifting platforms and a wide mix of legacy processes, progress frequently relies on people who can connect ideas across different areas of the business and rethink how established workflows operate. This type of creativity is not about producing something novel, but about recognising patterns and spotting opportunities to find practical ways of making improvements.
It is also an essential part of how AI is used effectively. Generative AI introduces new possibilities, although its value depends on the ability to frame the right questions and understand where it can meaningfully support existing work.
Emma Wicks explains:
This attitude reflects a broader reality across the industry, where resourcefulness shapes what innovation looks like in practice.
When teams approach challenges with this mindset, AI becomes a means to strengthen processes that have long been difficult to optimise, rather than a piece of technology to deploy simply because it is available.
Get recommendations on how AI can be applied within your organisation.
Explore data-based opportunities to gain a competitive advantage.
The problem-first approach to using AI
AI adoption in many organisations begins with questions such as “Where can we use it?” or “What can we automate?”. These questions often come from a genuine desire to innovate, but they can push teams towards technology-led initiatives that are only loosely connected to real operational needs.
Most AI failures in media come from starting with the technology rather than the problem. The result is familiar across the industry: isolated proofs of concept, tools that do not integrate into daily workflows and projects that struggle to demonstrate lasting value.
I always start with the problem first. If generative AI is the way to solve it, great. If it is not, we will not use it.
A more sustainable approach starts with understanding the underlying problem. Instead of looking for opportunities to apply AI, teams focus on identifying the specific challenges that hinder progress or limit insight.
This perspective reduces the pressure to adopt AI for the sake of visibility and instead encourages teams to concentrate on areas where new capabilities can genuinely help. For media organisations, it also provides a safeguard for investment by ensuring that resources are directed towards improvements that support editorial, commercial or operational goals rather than experimental work that may not lead anywhere.
By making problem definition the first step, AI becomes part of a broader toolkit, applied selectively and with clear purpose rather than a technology that drives strategy on its own.
Read more about AI on our blog:
Use case 1: improving image metadata with AI
Problem: No structured way to analyse image characteristics at scale.
AI Capability: Generative AI auto-tags attributes such as people, faces, layout and composition.
Outcome: Clearer insight into which images perform best across platforms.
Image choice has a noticeable influence on how articles perform across platforms, yet many media organisations still lack the detailed metadata needed to understand why certain images work better than others. Most publishers rely on simple counts of images or basic engagement metrics, which offer little insight into the visual features that shape performance. In the absence of this structure, teams often depend on intuition rather than evidence.
This leads to a familiar pattern. When editors ask which image styles perform best, the task often becomes a manual review of thumbnail collections and an attempt to spot informal trends. As Emma noted:
We were literally just eyeballing thumbnails... now we can see whether a person looking at the camera works better.
Manual review can produce broad impressions, but it cannot scale or reveal the subtleties that influence behaviour across platforms.
Generative AI provides a practical alternative by automatically tagging images with attributes such as the presence of people, facial expression, colour balance and layout. When combined with performance data and analysed through machine learning models, these tags make it possible to identify which visual elements contribute to stronger results on some platforms but not on others.
For media organisations, strengthening metadata offers immediate and transferable value. It supports clearer insight and more consistent testing, which results in more informed decisions about image strategy. By improving data foundations, teams can achieve reliable gains without relying on high-visibility AI features.
£1B+ in bookings for the UK’s largest independent broadcaster with a new ad management platform
Use case 2: AI for more accurate article classification
Problem: Outdated article classification taxonomies with inconsistent tagging and heavy manual rules.
AI Capability: Generative AI assigns mid-level subjects more accurately than rigid code.
Outcome: Clearer insight, improved reporting and stronger personalisation.
Article classification remains a long-standing challenge for many publishers. Taxonomies often evolve gradually, becoming brittle and inconsistent over time, leaving teams with a mix of broad categories that are too general and granular tags that are too detailed to support reliable analysis. The result is a classification system held together by complex rules, extensive conditional logic, and continuous editorial effort simply to keep it functioning.
Instead of relying on fixed rules or thousands of inconsistent tags, AI can help create mid-level taxonomies that better reflect how content is actually produced and consumed. These categories offer enough nuance to reveal meaningful patterns in audience behaviour while remaining structured enough to maintain consistency. At The Telegraph, this shift replaced a system built on lengthy logic statements. Emma gladly explained:
We used to have 13 pages of code... now AI categorises articles in a more consistent way.
For the wider industry, the benefits are clear. More reliable categorisation strengthens reporting and insight generation, improves topic planning and reduces friction between desks by offering a shared understanding of how content is organised. It also enhances personalisation and recommendation systems by providing more accurate signals about article topics.
The broader lesson is that valuable AI use cases often begin with strengthening core foundations. By modernising taxonomies, media organisations can reduce editorial overhead, unlock clearer insight and create a more effective base for future AI applications.
Building a culture of safe AI exploration
The ability to use AI effectively is shaped as much by organisational culture as by the technology itself. Tools evolve quickly, and their value depends on how confident teams feel when experimenting with them and how willing they are to share what they discover. A low-pressure environment encourages this kind of behaviour, allowing people to learn through practical use rather than formal instruction alone.
One useful model is to create regular opportunities for teams to exchange ideas and demonstrate what they have found. At The Telegraph, this takes the form of monthly sessions that range from simple discoveries, such as a feature that improves the tone of an email, to early prototypes built through experimentation. As Emma observed, “It is not all on me to find everything”, a reminder that reflects the importance of distributing curiosity and responsibility across the whole team.
For media organisations, several practices help create the right conditions:
- Encourage experimentation by giving teams permission to test tools without expectation
- Support better prompt craft through simple guidance and shared examples
- Offer safe spaces where people can try ideas without concern for mistakes
- Highlight useful discoveries so others can learn from them
Together, these practices help AI adoption grow from genuine relevance rather than pressure, allowing new ideas to take hold naturally across the organisation.
Where AI creates real value in media
Much of the public discussion around AI in the media still focuses on automated content creation, particularly the idea of AI writing articles. Although this attracts attention, it overlooks areas where AI can make a more consistent and measurable contribution. The stronger opportunities lie in improving the infrastructure that supports journalism rather than attempting to replace the work itself.
Enhancing metadata, strengthening insight generation and streamlining workflows offer clearer value. These improvements help teams understand how content performs, identify bottlenecks and create space for journalists to focus on reporting and analysis. Selectivity is essential when using AI.
The real risk is not that AI will replace newsrooms, but that organisations may concentrate on visible experiments while overlooking quieter improvements that raise quality every day. By prioritising the foundations of content production, media companies can develop a more stable and effective approach to AI.
Summary
Generative AI delivers the most value when it is used to solve well-defined challenges rather than adopted by default. A problem-first approach helps teams understand where AI can make a meaningful contribution and where established methods remain more effective.
Strengthening metadata, improving insight generation and giving teams low-pressure opportunities to explore new tools all support this approach, even if these areas attract less attention than content-focused experiments.
Emma emphasised the importance of knowing “where we want to use it and where we do not”. Generative AI is not inevitable, yet it can be transformative when applied with purpose. The organisations that benefit most will be those that invest in foundations rather than shortcuts and prioritise improvements that genuinely support their goals.
Driving revenue and shaping the future of Media with technology
Contact us to see how we can help you transform your media business.
