STEPHEN ASH

Defining the AI Visualisation Pipeline

The Challenge:

The generative AI landscape is heavily fragmented. While most off-the-shelf models can easily produce a polished, standalone image, they frequently fail when subjected to the rigors of actual event production—struggling with spatial continuity, scale, and lighting logic. For an agency to adopt AI effectively, it is critical to know not just how to prompt, but which specific physics engine to use for different phases of a project.

The Approach:

To define a reliable production pipeline, I conducted a rigorous, multi-stage stress test across five major AI architectures (including Nano Banana, Flux, GPT, and Seedream). The models were tasked with building a complex event space iteratively: establishing the room, populating it with people to scale, adding a foreground audience, and finally executing a dramatic camera perspective shift to test their 3D spatial awareness.

The Impact:

The R&D successfully mapped specific AI models to distinct production phases:

• Phase 1 (Ideation): Models like Flux generate beautiful, highly atmospheric visuals but drift mathematically. They are perfect for early-stage mood boards where selling the emotion is more important than architectural accuracy.
• Phase 2 (Execution): Models like Nano Banana possess a rigid understanding of 3D space, making them the go-to tools for locked designs that require strict continuity across multiple camera angles.
• Volumetric Insight: The test also revealed that video generation models possess a vastly superior understanding of stage lighting. Because they are trained on temporal physics, video tools render beams, washes, and atmospheric haze with a level of realism that static image generators cannot achieve.

Global Broadcast Delivery (Moody's Credit Trends)

The Challenge:

During the pandemic, Moody's Investor Services needed to pivot their flagship "Credit Trends" roadshow into a fully virtual format. The brief was highly complex: seamlessly connect and broadcast from 11 different countries over an 8-week period, maintaining flawless production value for a demanding financial audience.

The Approach:

I delivered this as part of a tight, three-person core team, working alongside a Logistics Manager and an Event Manager.

• Dividing the Scope: While they handled global logistics and talent coordination, I took the lead on the technical broadcast architecture.
• Decentralised Broadcasting: Rather than relying on a single, centralized studio, I managed remote satellite teams to build a resilient, multi-site network.
• Risk Mitigation: I implemented rigorous technical rehearsals to ensure fail-safes were in place across multiple international time zones.

The Impact:

The series was delivered flawlessly, providing Moody’s with a resilient broadcast platform that kept them connected to their global market when physical events were impossible. The collaborative framework we built was recognized within the industry, winning Media and Broadcast Event of The Year at the 2022 C&IT Awards.

Directing AI Narrative (The Aphantasia Project)

The Challenge:

The generative video space is heavily dominated by disjointed, 20-second "AI slop" clips. I wanted to prove these tools could execute genuine, engaging storytelling. I chose a deeply personal subject: my experience with Aphantasia (the inability to visualise mental imagery) and how my highly vivid dreams blur the lines between reality and memory.

The Approach:

This project served as a sandbox to understand the practical limits of AI video generation and build a reliable production workflow.

• Character Consistency: To bypass strict AI guardrails (especially Google's restrictions on uploading faces), I used my own likeness as the protagonist and generated AI representations for my family members.
• Tool Orchestration: I used the production to test which specific image and video generators excelled at different tasks, learning exactly how to prompt for consistency, lighting, and animation hand-offs.
• Post-Production Pivot: I initially attempted to edit the film using an AI-native editor, but the massive file sizes made it unworkable. I pivoted back to DaVinci Resolve to cleanly assemble, grade, and finalise the cut.

The Impact:

While the rapid evolution of AI video means the renders from 12 months ago look a bit rough by today's standards, this project was a crucial foundational exercise. It taught me the realities of AI workflow architecture, prompt limitations, and the necessity of combining generative tools with traditional post-production pipelines. The final film is now live on YouTube.

Training Flux LoRAs for Live Event Aesthetics

The Challenge:

While off-the-shelf AI image generators are incredibly powerful, they fundamentally misunderstand the physical realities of live events. Generic AI struggles to accurately generate industry-standard trussing, lighting fixtures (like moving heads), screen aspect ratios, and realistic staging geometries.

The Approach:

To solve this, I curated specific datasets of past event builds, lighting rigs, and custom staging. Using the Flux architecture, I trained bespoke LoRA (Low-Rank Adaptation) models to teach the AI the exact visual language of high-end corporate and concert production. This essentially gave the AI an "industry-specific" education.

The Impact:

The custom models enabled the rapid generation of highly accurate, brand-compliant creative concepts that were actually buildable. This allowed for real-time visual iteration during client workshops, ensuring the creative concepts remained structurally and technically viable from day one.

The "One Team" Approach to Venue Management

The Challenge:

Historically, there is often deep friction between an in-house venue technical team and an external incoming production agency. This "us vs. them" mentality leads to logistical bottlenecks, safety risks, and lost commercial revenue for the venue.

The Approach:

Throughout my venue management career (spanning prestigious locations like One Great George Street to Manchester Central), I implemented a strict "One Team" philosophy. Rather than acting as a mandatory hurdle, I positioned the in-house technical department as an indispensable, collaborative partner. This meant offering total transparency, sharing expert knowledge of the building's unique quirks, and proactively solving rigging and power challenges before the incoming trucks even arrived.

The Impact:

This approach transformed the in-house technical offering into a major commercial asset. By building trust and removing friction, incoming agencies were far more likely to utilise our in-house equipment and crew, directly increasing the venue’s capture rate for high-value technical contracts.

Project Meta – Building This Digital Portfolio

The Challenge:

To transition into senior agency roles, I needed a premium, highly visual digital portfolio that accurately reflected my technical and creative capabilities. However, my background is in live event production and 3D rendering, not web development. I had never designed or coded a website from scratch in my life.

The Approach:

Rather than relying on generic template builders, I decided to treat the portfolio as a live R&D project, orchestrating a full suite of AI tools to act as my development and design team.

• Strategy & Copy: Partnered with Perplexity as an interactive sounding board to refine my career narrative and aggressively edit the CV text.
• Site Architecture: Utilised "Vibe Coding" via Google Anti-Gravity to generate the raw HTML, CSS, and site structure entirely through natural language prompting.
• Visual Assets: Generated the cinematic video backgrounds using AI text-to-video models, and produced the bespoke case study imagery using Nano Banana and advanced image generators.

The Impact:

From zero web development experience, I successfully coded, designed, and deployed a modern, interactive digital CV. Beyond just showcasing my past work, the site itself acts as a living proof-of-concept. It demonstrates how effectively orchestrating modern AI workflows allows technical leaders to rapidly execute complex, multi-layered projects outside their traditional domain.

Automating the Quote – The N8N / AI Prototype

The Challenge:

The event quoting process is notoriously slow, requiring a project manager to manually break down a client brief, distribute requests across multiple sub-disciplines (audio, lighting, scenic), and compile disparate price lists into a single, cohesive budget. I wanted to see if this entire workflow could be automated.

The Approach:

I designed and prototyped an automated quoting pipeline using N8N. The workflow started with a digital client intake form. An AI "Project Manager" agent analysed the brief, fragmented the requirements by discipline, and routed specific prompts to distinct "Company AI" agents. These secondary agents cross-referenced deterministic data (inventories, day rates, and standard equipment lists) to generate estimated sub-quotes. Finally, the system compiled these into a master budget, explicitly flagging bespoke or complex requirements for human review.

The Impact & The Reality:

While the prototype successfully executed the complex prompt-chaining and generated moderately accurate deterministic quotes (equipment and basic crew schedules), it highlighted a crucial industry truth. Live events possess too many non-deterministic variables—venue quirks, specific client demands, and creative nuances—for a fully autonomous quoting system to be reliable.

Ultimately, this R&D project proved incredibly valuable. It demonstrated that while AI is excellent at automating the heavy lifting of inventory and basic mathematics, high-stakes account management still inherently requires a seasoned human to interpret the grey areas and finalise the commercial strategy.

The Trusted Advisor – A Methodology for Client Retention

The Challenge:

In the high-stakes corporate event sector, clients are frequently sold "one-size-fits-all" packages, or presented with creative concepts that look incredible on paper but fail completely in execution. Agencies often lose clients not because an event failed, but because the client felt unheard, stressed, or disconnected from the core message they were trying to deliver.

The Approach:

Over two decades of managing £500k+ portfolios, I have refined a client management methodology built entirely on active listening, strategic honesty, and bespoke ideation.

• Objective Mapping & Active Listening: Before discussing logistics, budgets, or deliverables, I focus entirely on the why. I work closely with the client to understand their core objectives and identify what their "key moments" need to be, ensuring we never sell the same generic show twice.
• Solution-Focused Ideation: I am unafraid to be honest when a creative concept is operationally unviable or misaligned with the budget. However, I never present a roadblock without simultaneously collaborating with the creative specialists around me to pivot toward a brilliant, achievable alternative.
• Systemised Cohesion: I act as the calm centre of the storm. By utilising strict organisational frameworks and building highly realistic budgets from day one, I ensure that the internal agency team, external suppliers, and creatives all pull in the exact same direction.

The Impact:

This methodology removes the friction and anxiety from event planning. By making complex creative processes feel simple, adapting my communication to the client’s level of understanding, and ensuring the final event actually delivers on their core message, I foster deep trust. This approach is directly responsible for transforming single-project clients into multi-year, highly profitable retained accounts. They return because they know that both their vision and their reputation are in completely safe hands.