Prompting and Crafting the Gen AI-Driven Recruitment Campaign Giving Network Rail a Cinematic Edge
When purpose-driven creative agency 23red (part of Capgemini) and Future Crea+ives (the AI division of Common People Films) unveiled Network Rail’s latest recruitment campaign this week, it marked more than a new set of OOH and print executions. By embedding generative AI at the heart of production, the team demonstrated how human-AI collaboration can streamline workflows, reduce budgetary pressures, and still deliver beautifully cinematic, slop-free imagery that resonates with both existing staff and prospective candidates alike.
The Challenge: Reimagining Recruitment in a Logistically Complex Environment
Network Rail is the publicly owned body responsible for owning, operating and maintaining the UK’s railway infrastructure – tracks, signalling, tunnels and stations – ensuring the safe and reliable running of train services across England, Scotland and Wales.
Recruiting rail maintenance workers traditionally demands extensive location shoots across live lines, comprehensive safety protocols and high production costs – often north of £200,000. As Kasia Williams, senior account director at 23red, explains, “A conventional shoot of this scale would typically entail at least two to three months of work, including four to six weeks of pre-production, a week on-site and another four to six weeks of post-production.” Even with a generous budget, filming on live railway lines can be fraught with permit delays, safety restrictions and scheduling conflicts.
Faced with these constraints – and the ambition to create a campaign that felt both authentic to Network Rail’s operational reality and “playful yet cinematic” – 23red knew that relying solely on traditional production methods would not only inflate costs but also hamper timelines. Instead, they chose to integrate gen AI alongside 3D modelling and advanced retouching, creating a “hybrid” process that preserved creative quality while accelerating delivery.
“Recruitment campaigns often talk as much to your existing staff as they do to potential new ones, so it was important that we reflected the reality of their job, even with the inclusion of metaphoric objects crucial to the idea,” says Tristan Cavanagh, 23red creative director. “Common People x Future Crea+ives worked really hard to be able to iterate the gen AI outputs whilst also maintaining consistency in what was liked from previous versions, but we had to communicate constantly to ensure we had a good understanding of what could be achieved at each stage.”
Ramy Dance, co-founder and creative director at Common People Films / Future Crea+ives is naturally an AI proponent. But for him, there is a crucial balance to be struck. “The secret sauce to achieving successful results with AI is a hybrid approach and knowing when to use it and when not to,” he says. “That’s why an open and collaborative approach is key. Working closely with the awesome team at 23red we quickly identified which scenes should be AI and which we would tackle using other traditional techniques. For us, the tech should remain invisible, it’s about great ideas. The work should leave you saying ‘wow’ and not ‘how?’”
Building the Hybrid Workflow: Four AI-Powered Steps
Rather than relegating AI to simple concept-visualisation, the team adopted a four-step workflow that wove AI tools into each phase of production:
- 1. Background Generation
Using Adobe Firefly alongside Photoshop, gen AI crafted highly accurate railway backgrounds, embellishing them to achieve that cinematic quality. These backgrounds established context – think multi-track junctions rather than a single lonely line – to reflect Network Rail’s distinct safety protocols (for example, always depicting a second worker when someone is on a ladder). - 2. Object and Prop Creation
3D-modelling experts built essential props such as carts, integrating them seamlessly with AI-generated environment elements. - 3. Character Generation (Workers)
A customised LORA model was trained on real Network Rail personnel imagery – licensed from the client’s own media library – to ensure uniforms, PPE and workwear were true to life. These 3D-derived figures were then run through gen AI prompts to introduce slight gestures and expressions, giving each character authenticity without the need for talent sourcing or on-set direction. - 4. Compositing and Retouching
Finally, seasoned retouchers at Common People combined all elements – backgrounds, props and characters – much like a high-end photographic edit. AI was used in each step, but the team still needed traditional methods like compositing and 3D creation by Future Crea+ives, or sometimes a mix of both.
Balancing AI Efficiency with Creative Rigor
Although gen AI cut both time and costs, 23red and Future Crea+ives were adamant that AI never replace human creativity. As Kasia points out, “The AI was used as a creative tool enhancing workflows and creativity, not replacing human expertise. By reducing the need for physical shoots, particularly on a live railway, AI demonstrated clear benefits in efficiency and cost-effectiveness – which aligned with Network Rail's operational priorities.
Kasia continues, “Network Rail’s client team had a clear understanding of AI as a collaborative tool, not a job replacer, and saw its potential to drive innovation, reduce costs, and elevate creative production within complex environments like the railway network.”
Ramy underlines this: “We've got people with over 20 years of experience in production and post-production who really know their stuff. They're experts at using all the latest tools to get the best outcome for every project, every time.”
In practice, the hybrid process halved the typical production timeline – from an estimated 10–12 weeks down to roughly five or six.
“It’s undeniable that AI can find efficiencies across the board,” says Ramy. “It certainly helped the budget and timelines on this project. It’s important to note that it’s not always a saving on the creative part though. There are still so many talented and experienced people involved who use traditional processes when creating the work. We like to think of ourselves as a company of efficiencies, helping to reduce cost and timeframes alike where we can. It’s case by case and really depends on what challenge you're facing and how best to answer it.”
Nevertheless, the team resisted any temptation to let AI shortcuts degrade authenticity. For every AI-generated shot, Network Rail’s in-house experts reviewed safety protocols, uniforms and infrastructure details to ensure nothing felt “off.” “No railway in England has just a single track for instance,” Tristan notes. “So we needed to include a second track in the composition, despite making the images busier. When we felt it’d be great to have someone on a ladder for the flowerpot execution, we had to ensure a second worker was also shown securing it. So from an agency side you have to be as flexible and collaborative as possible, and from the client side they needed to ensure they were running any proposed visuals past internal teams with practical knowledge of what would be normal procedure.”
Ensuring Ethical and Legal Integrity
Generative AI often raises questions around copyright and training data. To mitigate risk, 23red and Future Crea+ives committed to exclusively using gen AI platforms indemnified for commercial usage – principally Adobe Firefly with an enhanced licence. All AI training was conducted on proprietary Network Rail imagery or properly licensed 3D assets. “With open AI you’re never sure where the reference material is drawn from,” Kasia explains. “So everyone needs to ensure that you’re using closed models with images you own the rights to.”
The 23red and Future Crea+ives also implemented rigorous quality-control checkpoints: every AI output was scrutinised for potential hallucinations (inaccurate or implausible details generated by the model). Human experts then reined in any anomalies via traditional tools. The result is a suite of images that feel both inventive and grounded, marrying gen AI’s flair with industry-standard production values.
Lessons Learned
Both 23red and Future Crea+ives walked away with new skills and processes. For 23red, the emphasis is on ethical gen AI usage – developing new ways of thinking, protecting IP and building robust feedback loops. As Tristan notes, “You’re feeding back in a new way – almost thinking in a prompt-oriented way with lots of specificity. You also need a rigorous quality control at the final stage, to look for any hallucinations that might’ve crept in along the way.”
Future Crea+ives, meanwhile, has confirmed that AI-assisted OOH and print campaigns are now squarely in their wheelhouse. “We’re very comfortable in our ability to deliver OOH and print campaigns to specification using AI and a hybrid approach, and this success gives us confidence that we can unlock endless creative possibilities moving forward,” says Ramy. “We were quietly confident about delivering this brilliant idea from the start, but we aimed to push our technical and creative boundaries to ensure exceptional work. Our open and collaborative partnership with 23red was instrumental to achieving this.”
What This Means for Future Campaigns
For Network Rail, the success of this recruitment campaign opens doors to more AI-enabled projects that might have been dismissed due to budget or logistical barriers. “With a closed model, where you’re using owned images and working with trained professionals it’s now got to the stage where it’s a genuine option for certain ideas,” Tristan reflects. “For Network Rail, with their restrictions around shooting on the railway, it’s given them the confidence that realistic portrayals of rail infrastructure and staff are possible without complex logistics. It’s not going to be right for some clients and briefs, but like any new tool, if the people using it are aware of the ethical parameters you need to adhere to, and have a good idea of what’s possible, it has the ability to make an idea viable that you might’ve discounted before.”
