Richard Fox

Fox writes that Pixelz wants AI to become “more boring”: predictable, dependable and reliable, so teams know what they put in and have a good idea of what they’ll get back, rather than crossing their fingers with every generation.

• The gap, they write, is between generating something impressive and creating something a team can rely on in production.
• Randomness, assumptions and hallucinations stand in the way.
• The proposed fix is to build workflows around the model, not just improve prompts: use clear inputs and defined use cases, review and post-produce outputs, and add guardrails to rein in unpredictability and make results easier to repeat at scale.

Fox describes these stages as part of Pixelz AI Studio and invites readers to a session during E-commerce Content Creation Week, where the studio is opened up from input to final asset. The post is signed Richard.

Read original

AI adoption need not mean reinventing an entire content operation, they write.

• Often, the better business case is a use case that fits what a team already does and solves a problem it already has.
• A smaller use case is not necessarily less ambitious, just more useful.
• Fox also describes an AI use case matrix that weighs the opportunity against what it takes to introduce it, rather than ranking AI by how advanced or exciting the technology seems.

The examples are deliberately unglamorous: using AI to relight existing photography, creating another asset from images already being shot, or finding one small part of the content workflow where AI can remove work without changing everything around it. Fox links to a longer piece on building a less complex AI business case and says the post includes further examples.

Read original

Yet that middle is what determines whether an asset is usable: choosing the right inputs, regenerating, refining, retouching and ensuring the final asset is ready for the product detail page. Showing the chain helps a team distinguish a demo from a production process.

To show the process rather than just describe it, Pixelz built a fully AI-generated product detail page and made it interactive. Hover points reveal the input images, generations and retouching behind each asset, and a link invites readers to explore.

There’s a mild tension between this argument and the idea that AI should become boring: the interactive breakdown is itself a showcase, designed to make an unglamorous process interesting to look at. The positions are compatible—the focus is on the process, not the spectacle—but they differ slightly on how much presentation matters.

Read original

The stated purpose is to understand how AI is being used and perceived inside organisations, returning to the gap between hype and adoption. Findings were promised for that year’s E-commerce Visual Trends Report, positioning the account as both a commentator and a collector of practitioner evidence.

The survey offers three explicit answers—active implementation, still testing or not convinced—takes a few minutes, is anonymous and links directly from the post. Ask for the results, response volumes and report contents; they aren’t included in the posts.

Read original

They present public recognition as the remedy, with criteria that go beyond output volume: improving how a studio works, raising creative standards, helping a team adopt new technology or simply being the person everyone relies on to make things happen. This also signals what the discipline values beyond AI.

The post announces that nominations are open for the returning “Top 30 E-commerce Content Creation Professionals” list, described as one of Pixelz’s favourite projects. Self-nominations are explicitly encouraged, and the post links to the nomination form.

Read original

The idea is simple: reliability matters more than spectacle.

The post makes a deliberately counterintuitive case: Pixelz wants AI to become "more boring," because boring means predictable, dependable and repeatable — you know what goes in and roughly what comes back.

• Fox names the gap directly: there is a difference between generating something that looks impressive and creating something a team can rely on in production, and the gap is filled with randomness, assumptions and hallucinations.
• The proposed fix is workflow rather than better prompts — clear inputs, defined use cases, review, post-production and guardrails — and Fox says this is what has been built into Pixelz AI Studio.
• The post closes by inviting readers to a session during "E-commerce Content Creation Week" where the studio is opened up from input to final asset, and includes a registration link.
• It signs off personally as Richard.

For brands planning a campaign

A natural anchor for a webinar or content collaboration on practical AI adoption: “make AI boring” works as both a talk title and a session premise, and the input-to-asset walkthrough lends itself to a live demo. The walkthrough sits behind the registration link, so ask to see the actual studio before committing to a format.

It addresses the question many teams have: where do we start if we do not want to rebuild our whole content operation?

Fox contrasts the AI use cases that dominate the conversation — campaigns, models and generated video — with a quieter layer of small-scale adoption that still pays off: relighting existing photography, producing an extra asset from images already shot, or removing work from one step in an existing content workflow.

• The principle: AI adoption need not mean reinventing an entire content operation.
• Often, the stronger business case is the use case that fits what a team already does.
• The post describes an “AI use case matrix” that weighs opportunity against the effort of introducing a tool, rather than ranking AI by how advanced or exciting it seems.
• It links to a piece on building a less complex AI business case and says examples follow, but includes none.

For brands planning a campaign

A joint session built on this framing — small, adoptable use cases with an honest effort column — would resonate with teams that have budget but no AI roadmap. Ask to see the matrix and working examples before promising a framework.

This example shows the production chain a brand would actually be buying.

The post argues that conversations about AI visuals focus on prompts and final images — the wow moments — while ignoring everything in between.

• Fox describes that middle layer: choosing the right inputs, regenerating, refining, retouching and ensuring a final asset is ready for a product detail page.
• To demonstrate the process, Pixelz built a fully AI-generated product detail page and made it interactive, with hover points revealing the input images, generations and retouching behind each asset.
• The post links to the page; ask to review the artefact before using it.

For brands planning a campaign

For brands seeking proof of process rather than proof of concept, this is the example to show them.

• An interactive breakdown could also work as a co-marketing format, embedded or recreated for a joint landing page.
• Before using a screenshot or recreation in a partnered asset, verify that the interactive page still works and that its inputs, generations and retouches match the description; only the post’s description was reviewed.

A short pulse check asks e-commerce professionals about their experience with AI-generated visuals, offering three options: actively implementing, still testing or not convinced yet.

• The stated aim is to understand how AI is genuinely used and perceived inside organisations, with findings to be shared in that year’s E-commerce Visual Trends Report.
• The survey is anonymous, takes a few minutes and is linked in the post.

For brands planning a campaign

A survey-and-report collaboration offers a lower-commitment entry point than a webinar: it needs a question and distribution, not a rehearsed demo, and produces a shareable asset.

• Treat the audience as self-selected e-commerce practitioners.
• Ask for the sample size and how findings were reported before starting a joint data project; the post provides neither.

The post announces nominations for the “Top 30 E-commerce Content Creation Professionals” list, a returning project and one of Pixelz’s favourites.

• It recognises behind-the-scenes content work that often goes unnoticed, though it keeps production moving.
• Suggested nominees include people improving studio operations, raising creative standards, helping teams adopt new technology or making things happen for others.
• Self-nominations are explicitly encouraged, and the post links to a nomination form.

For brands planning a campaign

A nominations call is a ready-made organic-touchpoint format: it invites audience participation rather than passive consumption, making it a fit for a brand seeking association with talent recognition rather than product promotion.

• It also shows the account publishes community and people-focused posts alongside product content, which matters when assessing whether a placement would feel native.
• Ask for the list and its selection criteria before committing.

One post puts it bluntly: AI should become "boring"—predictable and dependable—because results remain random, and impressive work is not always reliable. The answer, they argue, is process rather than a better model: clear inputs, defined use cases, review, post-production and guardrails that make output easier to reproduce at scale.

Describes fitting AI around existing production stages rather than replacing them, with clear inputs, defined use cases, review, post-production and guardrails.

For brands planning a campaign

A practical AI adoption webinar can build on this argument: focus on "how do we get repeatable output," not "what can this model do." A session on inputs, review checkpoints and post-production for e-commerce assets fits the material they already publish and is easy to title and promote.

Examples: Source 1

Examples include relighting existing photography, creating an extra asset from images already being shot and removing work from one step in a content process without rebuilding the whole operation. One post argues that AI adoption need not mean reinventing content operations: a smaller use case can be more useful, not less ambitious.

Frames a use-case matrix around the opportunity and the effort required to introduce it, rather than how advanced the technology seems.

The posts argue against ranking use cases by size: smaller ones can be more useful, not less serious.

For brands planning a campaign

This is a strong hook for a B2B e-commerce content production audience because it lowers the perceived cost of getting started.

• A webinar or joint session could rank small use cases by the disruption they require, using scenarios the audience already faces.
• It also gives producers a clear internal argument: choose a use case that fits existing work.

Examples: Source 1

It describes a fully AI-generated product detail page made interactive, with dots revealing the input images, generations and retouching behind each asset. The aim is to show how assets are chosen, regenerated, refined and retouched until they are ready for the PDP.

Uses an interactive artefact to reveal the process: hover states show the input, generation and retouching stages behind each final asset.

Treats the eye-catching generated image as the least interesting part, shifting attention to the process behind it.

For brands planning a campaign

This offers a format for a collaboration, not just a topic.

• A session or co-produced piece showing intermediate versions, rejected generations and retouching would position the producer as someone who understands the unglamorous work—credibly, for an operator audience.
• It could also produce an interactive demo.

Examples: Source 1

It argues that behind-the-scenes content work often goes unnoticed, despite keeping things moving. The criteria include improving studio operations, raising creative standards, helping a team adopt new technology or being someone others rely on.

For brands planning a campaign

This offers a reason to approach them beyond a single webinar.

• A producer could support or co-promote the nomination drive, contribute judges or case studies, or create content around the final list.
• It reaches practitioners beyond decision-makers.

Examples: Source 1

It asks whether their team is actively implementing, still testing or unconvinced, with findings promised in an annual e-commerce visual trends report. The post frames it as a pulse check on how organisations actually use and perceive AI, reflecting the recurring gap between hype and adoption.

For brands planning a campaign

This opens a co-research opportunity: contribute questions, share the survey with a producer's audience or co-publish the results. Survey responses could also surface audience objections for a session aimed at sceptical teams.

Examples: Source 1

The post confirms Fox speaks for the business; it does not describe Fox's specific role or title.

View example

Both accounts treat the generated image as only part of the production problem.

Ladefoged's posts frame the issue through event work and a longer post-production history, including a FLOW London keynote and industry agenda setting. Fox's focus is adoption choices, workflow guardrails, and smaller use cases that fit existing operations.

Read original · Original 2

That overlaps with Fox's arguments about reliability and production scale.

Rasmussen's posts centre on stage appearances, panel discussions, and delivering the FLOW London event. Fox's focus is process design, use-case selection, and explaining AI adoption to teams still deciding what to introduce.

Read original · Original 2

Both accounts target brands and content teams evaluating production technology.

Mulhem's posts cover 3D and AR product twins, phone-based capture, and building Voxelo. Fox's focus is AI post-production, workflow reliability, and content operations—not 3D capture or a product launch.

Read original · Original 2

The FLOW: London agenda put Ladefoged in the opening keynote, looking back at the major shifts of the past 12 months and revisiting last year's trends and predictions, then kept them on stage as MC for most of the day.

• Separately, Ladefoged writes directly about the quality question a partner would want on the agenda: a lot of obviously AI-generated e-commerce content fails not because of the generation but because of what comes after it — adjusting lighting and composition, keeping the product accurate in colour, texture and size, and smaller retouching choices such as adding grain.
• Generating an image and having an e-commerce-ready image remain two different things.
• If the goal is a big room, an industry-state-of-play framing and a host who can carry a full day, this is the closer match.

The tradeoff is demo shape.

• Richard Fox's posts describe a keynote and quality argument, plus a posted event agenda.
• They publish no transcript, recording or slide deck, and do not cover the linked article.
• They also offer no walkthrough of inputs, regeneration, review and guardrails on a named product.
• If the partner needs a repeatable, productised working session with clear inputs and post-production steps they can co-publish, Fox's posts describe that chain directly rather than at keynote level.
• Ask for the transcript, recording or slides, and a product walkthrough.

Examples: Source 1 · Source 2

The FLOW: London posts show an event that sold out more than six weeks early, with keynotes, workshops, demos and eight roundtable conversations.

• Rasmussen states plainly that people bought before the full agenda was known — useful evidence when a partner wants co-marketing attached to an audience that already commits.
• The Visual 1st post documents Rasmussen on a panel in San Francisco, "Visuals for Commerce - Turning Visuals into Revenue Engines," alongside Caimera, Cloudinary and Titles, covering how GenAI is taking over parts of the traditional photoshoot and the questions that follow: is the product accurate, is quality consistent, does the output meet e-commerce requirements, and can it be done for 5,000 products rather than the five that made the demo?

Richard Fox's posts focus on event scale, audience trust and stage delivery; the production substance is a summary of an approach rather than a demonstrated process.

• They publish no recording, slides, registration numbers or follower breakdown, and do not cover the conference and event pages linked from the posts.
• If the partner needs to see the middle of the process — input selection, a rejected generation, the retouch pass and the guardrails — and publish it as a co-produced asset, Fox's posts cover those steps more concretely.
• Ask for the recording, slides, registration numbers and follower breakdown.

Examples: Source 1 · Source 2

Mulhem writes from 20+ years of watching e-commerce product pages and argues that a 3D layer has finally become practical: capture on video with any camera, reconstruct it as an interactive 3D model and drop it on the product page.

• Product photographers are the unlock because they already know lighting and spend a day in the studio; the same shoot can produce photos for the hero campaign and 3D for the interactive layer.
• Mulhem is building Voxelo to give photographers that path.
• A later post announces Voxelo 2.0 going live after six months of customer feedback, names co-founders Ben McKay and Roman Bromidge, and describes one video becoming a full 3D digital twin, AR-ready, with no 3D skills and no expensive studio.
• That is a better match when the buyer's own product is 3D, AR or product-visualisation software, or when the audience is product photographers rather than content operations teams.

Voxelo describes the launch in its own announcement; it is not an inspected demo.

• The posts publish no audience or commercial-availability data.
• They focus on one product and one category rather than AI adoption inside an existing content operation, so a partner whose audience needs review steps, guardrails and a repeatable process would get a different session from the one Fox's posts already describe.
• The fit depends on whether the campaign angle moves to 3D and AR.
• Ask for audience and commercial-availability data.

Examples: Source 1 · Source 2