Miroma AI Summit 2026

Every session from the day, with a short recap and the recording. Recordings play here if you’re signed into Microsoft 365. If one won’t load (common in Safari and private windows), use the SharePoint link under it.

1. 09:30

   Opening and welcome

   Ben Pilgrim

   AI is the tool. People are the difference.

   ▶ Load the recording

   Trouble playing? Open in SharePoint ↗

   Ben set out how the AI team thinks about AI across the group: through two lenses, competition and compliance. He framed the day around a paradox: AI is simultaneously overhyped and underhyped, and the team's job has been to work out, objectively, where it genuinely helps agencies and where it doesn't.

   • Competition: predictions that platforms would make agencies redundant haven't come true, so the question is what keeps agencies valuable and competitive.
   • Compliance: not just legal, but making work that is genuinely interesting and original rather than AI-generated volume.
   • The shape of the day: panels and external talks on each lens, agency showcases, and two practical sessions from the AI team on using AI well and building with it.

   "There is a lot of paradox held within AI… it is both simultaneously overhyped and underhyped."

   "Compliance… it's also… how we do things that are genuinely interesting, genuinely groundbreaking and not just AI generated volume."

2. 09:45

   Context First

   Tess McKean · AI Team

   Why AI outputs disappoint, and what to build so they don't.

   Build your agency brain →

   ▶ Load the recording

   Trouble playing? Open in SharePoint ↗

   Why AI outputs disappoint, and what to build so they don't. Tess opened with the same RFP run twice: a basic prompt with no context produced a generic response any agency could have written; the same prompt using Multiple's RFP skill produced a response grounded in their real case studies and policies, with every point cited.

   • AI is a prediction engine. It predicts what comes next from what it has seen. On well-worn paths it's excellent; on your niche, specialist work it guesses, confidently. The fix is context, not cleverer prompts: "context is not something you type, it's a layer you build."
   • Tools you can use today: connectors (work from live data in Drive, Figma, Meta Ads and more), skills (Claude interviews you and packages how you do a repeat task), projects, and design systems (your brand guidelines built into every output).
   • Build an agency brain. A curated set of folders on your own shared drive, with a map that tells AI where everything lives. Curation beats volume, it reduces risk, and it isn't locked into any one AI tool. It is groundwork, not thinking: you still own the direction, the standards and the review.

   "Every tool, every workflow and every agent you build is only going to be as good as the context sitting underneath it."

   "The model is the same for everyone… Having AI on its own doesn't set you apart any more. What nobody can copy is what's yours. Your client knowledge. Your wins. The way you work. And the trust you've built with your clients."

   "AI will reshape almost every role, but it won't shift accountability. Whatever the AI does, you own the direction, set the standards and review the work."

   Turn a repeat task into a skill → Build your agency brain →

3. 10:45

   Agency showcases

   Multiple · Spot Co · Miroma · Sold Out

   What each agency has put into practice.

   Browse use cases from across the group →

   Share your own AI win →

   ▶ Load the recording

   Trouble playing? Open in SharePoint ↗

   The Multiple Brain

   Rebecca, Multiple

   Agencies are so busy marketing their clients that they forget to market themselves. After one summer week in which Multiple released campaigns for two clients, started work for a third and told nobody, they built the Multiple Brain.

   • Capture: their resource-planning tool flags when a campaign wraps and prompts the owner to write it up, through one structured form (brief, results, content, tags).
   • Remember: every campaign becomes a tagged record in their Drive, which Claude can search.
   • Create: skills turn any record into a case study, website copy, Instagram and LinkedIn posts, a mailer or RFP copy, always with human review before anything goes out. Pulling past case studies together for an RFP, in their presentation branding, used to take about three days and now takes a couple of hours.

   "Agencies endlessly forget to market themselves. We're so busy doing it for everyone else, we never actually do it for ourselves."

   ▶ Load the recording

   Trouble playing? Open in SharePoint ↗

   Spot Co ad design tools

   presented by Tess for Spot Co

   • HTML calendar builder: email-safe showtime calendars for show e-blasts, on the client's brand via a design system. It replaced around two hours of hand-coding per email.
   • QR code generator: clean, styled codes that stay scannable with a logo in the middle, instead of InDesign's dense codes for long links.
   • HTML5 banner animator: drop in a flattened mock-up and its layers, and it positions and animates them.

   "They had a really clear goal and objective of what they wanted to build. So they gave Claude all of that context…"

   More detail, with screenshots, in the Impact stories on the Use Cases page.

   Brief to creative concept

   ▶ Load the recording

   Trouble playing? Open in SharePoint ↗

   The Miroma Brand Hub

   ▶ Load the recording

   Trouble playing? Open in SharePoint ↗

   Brief to creative concept, and the Miroma Brand Hub

   Aidan Shephard, Miroma Group

   Aidan uses Claude to condense long briefs into something digestible, and to generate concept imagery that previously meant hours in tools like Cinema 4D. It is good enough for concepts, pitches and internal comms, but not production-ready: final delivery still goes through the traditional craft route. He also showed the rebuilt Brand Hub, made with Claude after an earlier WordPress version didn't land: a resource library, agency portals, an events calendar and an onboarding area, checked by a developer for security.

   "It doesn't replace the creative. It just helps bridge that technical gap and speeds up process."

   ▶ Load the recording

   Trouble playing? Open in SharePoint ↗

   AI for client retention

   Freddie and Luke, Sold Out, with Ben Pilgrim

   Retention is the cheapest way to keep revenue, and for Sold Out, with clients of 30 to 40 years, every campaign is effectively new business.

   • Time saved goes upstream. Automating campaign set-up and reporting frees the team to spend time on the artist, the show and the client relationship, which is where they add value.
   • Real results: an annual review covering around 15 shows usually took two weeks of prep. This time the groundwork was done in advance and the final three priorities took about 20 minutes to pull together. The client then asked them back to present to their CEO.
   • Adoption is listened to, not imposed: they watch how tools are used, find out why people don't use them, and feed that back into the build.

   "We don't really hire agency people. We hire music fans."

4. 11:30

   If everyone has access to AI, why hire us?

   Panel · Adam, James, Toby, Nicole and Matt

   “We use AI” is not a differentiator. The proprietary knowledge, workflows and judgement built on top of the tools are.

   Clients & AIOur team & AIWays of working & AIWinning with AI

   ▶ Load the recording

   Trouble playing? Open in SharePoint ↗

   Panellists

   • AdamMultiple
   • JamesFold 7
   • TobyMFN
   • NicoleDewynters
   • MattMaker Lab

   Facilitated by James Heywood

   Clients now assume agencies use AI; it is table stakes. The panel agreed the differentiator is what sits on top of the tools: judgement, taste, craft, accumulated knowledge and trust.

   • Clients vary hugely. Some expect AI everywhere; some, particularly in live entertainment, are deeply sceptical. Being transparent and helping clients shape their own AI policy builds trust.
   • Knowledge is the moat. Long-standing agencies hold years of client and campaign knowledge that new, AI-first competitors don't. Putting it into an agency brain turns it into an advantage, not just a memory.
   • Speed matters too. Smaller, AI-first agencies can move faster. Established agencies win by matching that pace and pairing it with the knowledge and craft newer rivals don't have.
   • Pricing and people are shifting. Some procurement teams now expect year-on-year savings from AI, and agencies need to hold their ground on fair value. Freed-up time should go into higher-value work, and adoption works best bottom-up, by asking teams what they'd fix, not top-down.

   "In a world where AI is generating so much content… the human judgement, the human taste, the discernment to actually decide what is good… is only going to be magnified." (James Joice)

   "If you can put that knowledge into the brain, that is our point of difference."

   "For us, AI should be a friction. It shouldn't just be fluency. We use AI to spar and tell us when we are wrong, not just to go, 'here's an answer', really, really quickly." (Toby)

   "The act of writing is really the act of thinking."

   Capture what makes you different in an agency brain →

5. 13:15

   Springboards: The sea of sameness

   Pip · Founder, Springboards

   When every agency uses the same generic AI, they arrive at the same conclusions. Why AI pulls work towards sameness, and what keeps agency thinking divergent.

   Springboards on the AI Tools page →

   ▶ Load the recording

   Trouble playing? Open in SharePoint ↗

   A self-described "public service announcement" on AI's limits, from an AI company. Language models are fast, fluent and plausible, but their knowledge is "a mile wide and an inch deep", and they tell you what you want to hear.

   • Everyone gets the same answer. Ask any model for a random number and most of the room got 7, then 3, then 9. Ask for a running-shoe tagline and you get "Run your way", for any brand. Research across 25 models found the same convergence on open-ended questions.
   • Why it matters for agencies: agencies sell talent, a process and a form of insurance, not ads. When every agency uses the same models, pitches converge, novelty drops, and so does effectiveness. Pitch consultants are already seeing near-identical strategies.
   • The answer isn't to ignore AI, but to bring it into a genuinely creative environment of diverse people, collaboration and outside stimulus, rather than working solo inside a model.

   "AI won't kill advertising or agencies, but the sameness will."

   Get started with Springboards →

   Want a Springboards licence, or more seats for your team? Email the AI team at aiteam@miroma.com.

6. 13:45

   How does AI fit into creative workflows?

   Panel · Dave, OJ, Adam, Joe and James

   Where it belongs in the process, and where it doesn't.

   ▶ Load the recording

   Trouble playing? Open in SharePoint ↗

   Panellists

   • DaveFold 7
   • OJTwelve AM
   • AdamDewynters
   • JoeMultiple
   • JamesMaker Lab

   Facilitated by Ben Pilgrim

   • AI's biggest impact is in concepting: showing a client the idea rather than describing it. But that can pull time away from refining the idea itself.
   • Use AI to find the middle, then avoid it. Several panellists use AI's first answers as the map of what every other agency will pitch, and rule those out.
   • Discernment matters more than ever. With so much output, taste and judgement, and being able to defend why something is good, become the job. Audiences are also pushing back on work that looks AI-generated, and "perfectly imperfect" is gaining value.
   • The open question is juniors. How do new creatives build judgement when AI does the early work? Ideas from the panel: present the idea in two sentences and one image before any generation, and use AI for objective tasks rather than subjective ones.

   "We are able to make more stuff quicker than ever… what that has proven to me is that the need for discernment, good taste, judgement is greater than ever." (Dave)

   "By doing fewer things and asking them to do that exercise really precisely, I think you get to great creativity… 'cause it stops you just hitting the button and generating tonnes and tonnes of stuff." (Dave)

7. 14:45

   LTX: Designed to be built on

   Tara Filipovich · Product Lead, LTX

   Generative AI still asks you to prompt, generate and hope. How LTX's open model gives creative teams control back, and what people have built on it.

   LTX Studio on the AI Tools page →

   ▶ Load the recording

   Trouble playing? Open in SharePoint ↗

   LTX's video model is open: you can download the weights, run it locally and customise it, so your IP stays private. It generates synchronised video and audio, and understands how the world moves, not just how it looks.

   • Capabilities now: multi-keyframe control, retake (change one thing in existing footage), audio-to-video, long-form avatars of up to two minutes, extend, and lip-dubbing for localisation.
   • Closing the control gap with LoRAs: small add-ons trained on top of the model. A standard LoRA learns how something looks (a character, product or style); an IC-LoRA learns a function, such as relighting, colourising, removing objects, camera moves or upscaling. Many are ready to use in LTX Explore.
   • Where to start: LTX Explore (the playground), the open-source LTX Desktop app, and the LoRA trainer.

   "The hardest part is just deciding what you want to make, and having your own creative direction."

   Get started with LTX Studio →

   Want an LTX Studio licence, or more seats for your team? Email the AI team at aiteam@miroma.com.

8. 15:15

   AI at Maker Lab

   James Tugman · Maker Lab

   How Maker Lab builds in-house marketing teams that move fast and work smart.

   ▶ Load the recording

   Trouble playing? Open in SharePoint ↗

   Maker Lab builds in-house teams for clients, and AI has shifted their focus from who is on the team to how the end-to-end workflow runs.

   • Ideas and visuals faster: for a tourism client, Springboards gets past the blank page and LTX brings ideas to life, so clients can approve or reject quickly.
   • Fix the plumbing first: for a quick-service restaurant client, much of the win was structured briefing, triage and approvals rather than generative AI.
   • Production at speed: event quote cards went from about 7 hours to 45 minutes for 120 assets; point-of-sale materials went from 2 to 3 days to about 4 hours for 153 assets, with time kept for human review.
   • A 99% faster campaign: one of their makers built a pipeline that turns a brief into on-brand, compliant creative. Work that took about six weeks with an outside agency now takes about three hours, and it was built in about a month on everyday tools.
   • One system behind it all: a brief goes in, in any format. The system checks how that work has been done before for that client, then runs the right pipeline. The studio keeps adding new deliverable types.
   • No more chasing: meeting transcripts, the CRM and documents are connected, so the operations team can ask Claude what was promised to a client instead of chasing the sales team.
   • At scale: around 150 to 170 Maker Lab people work inside client teams, building with AI every day.

   "…the work around the work, which no one really likes."

9. 16:00

   Building our competitive advantage

   Tom Maycliffe · AI Team

   What the AI team has built, how we decide what to build, and where it's heading.

   Vote on what we build next →

   ▶ Load the recording

   Trouble playing? Open in SharePoint ↗

   How the AI team decides what to buy, extend or build, and what they’ve built so far.

   • Buy, extend, build. Buy where mature tools exist (Claude, LTX, Fireflies, Descript, Springboards). Extend them with connectors, skills and integrations. Build only where a solution is central to how we win and no tool fits. As models improve, the line keeps moving: some builds have already become Claude skills.
   • What's running: the AI Hub, the Maker Lab recruitment portal (more than 800 candidates processed), the out-of-home media system in testing with MX (40 campaigns built so far), and Meta campaign automation for Sold Out, which is now growing into a full digital media system with a client portal.
   • Where next: one media system for the group, rolled out agency by agency and deepened across planning, data and finance, plus tools across the new-business process. The room voted on the features they want first.

   "The number one metric for us is adoption."

   What should the AI team build next? →

10. 16:45

    Closing remarks and drinks

    James Heywood

    James thanked the speakers, panellists and everyone who made the day happen, and promised a summary and practical tips, including how to set up your own agency brain.

    "AI's really taken us all on a bit of an emotional rollercoaster… There's been a bit of fear, but hopefully after today there's a bit more inspiration… to tackle what's ahead."

    "The only thing that we do know is that it's not going to be how it looks today."