Tom Cranstoun | Professional Profile

Agent Experience (AX) audit pioneer and founder of Machine Experience (MX).

I started as an apprentice telecommunications engineer on the west coast of Scotland during the Cold War, working with security clearance on the communications networks that served the NATO installations there. Accuracy, resilience, and trust were operational necessities. I studied Communication and Computing at Glasgow Technical College, graduating with distinction, and worked on crossbar and electronic telephone exchanges, which I saw as specialised computers routing information at scale.

I've been building content systems since 1977, starting with assembler code, long before "CMS" was a term. Worked through IBM mainframes and PL/I. Co-authored Superbase, a database and content management system that predated the CMS category. Worked on the BBC News newsroom around AP's ENPS, and built .NET systems that fed off it and complemented it. The newsroom taught me provenance, sourcing, and the importance of context, compliance and accuracy. It's why I read AI hallucination as a failure of sourcing before it's a fault in the model. Over a decade with Adobe AEM, recent years with Edge Delivery Services.

From Edge Delivery Services to Machine Experience

Working with EDS taught me something unexpected: the structure that makes content work for AI agents is mostly what everyone needs. Hidden state, ephemeral notifications and incomplete information break AI agents. The same patterns fail humans with disabilities, cognitive load, or non-ideal conditions. And by 2024 machines had become a permanent audience: once published, content travels into search engines, retrieval systems, knowledge graphs, corporate AI, and agents its creator doesn't control. Those insights became Machine Experience (MX). I say "machine" rather than AI, LLM, or agent because each of those names a temporary implementation. The implementations will change; the audience will remain.

I help organisations make better strategic decisions about Adobe Experience Manager and Edge Delivery Services in this new reality. After working on content systems for BBC, Twitter, Nissan-Renault (hundreds of websites, where I led 15 architects each heading a team), Ford, MediaMonks, and others, I've learned that successful implementations come from asking the right questions before building anything. That matters more now, as AI agents change how web experiences are consumed.

Recent Adobe Experience Manager implementations demonstrate this approach: the Generate Variations feature cut banner creation to days where it had taken weeks, whilst maintaining human strategic oversight, delivering many variations with much higher click-through rates. Success came from agent-ready foundations, semantic structure, explicit state, machine-readable metadata, that let AI handle pattern generation whilst humans controlled messaging and brand alignment.

My work centres on what I call "clarity infrastructure": systems that make state explicit, feedback persistent, and information complete. Using Cloudflare's global edge network and Adobe EDS, I've implemented this principle at scale, enriching HTML with explicit state attributes, enforcing semantic structure, providing machine-readable Schema.org data, and putting critical information in the served HTML before JavaScript runs. This creates agent-ready foundations that work for CLI agents, API agents, browser agents, and every human user through universal design patterns.

Clarity Infrastructure at Scale

The business urgency is real: Amazon, Microsoft, and Google all launched agent commerce in early 2026. First movers in each sector who build genuinely agent-ready systems will capture agent-mediated transactions while competitors struggle with silent failures. But here's the efficiency multiplier: agent compatibility and accessibility improvements are identical work. Every pattern that helps agents, semantic HTML, explicit state, persistent errors, also helps screen reader users, keyboard users, and anyone in non-ideal conditions.

I work with teams facing complex AEM and Edge Delivery Services decisions, whether evaluating EDS adoption, planning AI agent integration, or reviewing architectural approaches. My focus is strategic guidance that prevents expensive mistakes and builds internal capabilities. The BBC, Twitter, and Nissan-Renault implementations weren't successful because of technical complexity. They worked because we developed frameworks that helped distributed teams make consistent decisions independently. That principle shapes everything I do.

My approach combines practical implementation experience with deep understanding of AI system internals. I write extensively about the statistical foundations of AI agents, how next-token prediction produces both capabilities and hallucinations, why linguistic tokenisation creates functional inequities, and how weighted averaging determines which HTML patterns agents can reliably process. This technical depth informs architectural decisions: knowing that agents perform statistical pattern-matching rather than "understanding" content explains why explicit state attributes and semantic structure matter more than visual design.¹

Consultancy Engagements

I take on interim consultancy roles and advisory engagements where strategic experience makes the difference:

• Plan Reviews - identifying gaps between intention and reality before implementation begins, particularly around agent compatibility
• Architecture Strategy - developing frameworks that balance corporate control with team flexibility while ensuring agent-ready foundations
• AI Integration - ensuring automation enhances rather than complicates workflows, with focus on clarity infrastructure
• Team Mentoring - building strategic thinking capabilities that outlast any single project
• Audit - where things went well, and where they could be improved

I have established first AEM practices from scratch. Strategic decisions prevented platform crashes and delivered significant cost savings. Teams gain capabilities to maintain and evolve solutions independently.

The engine behind MX

Over the last three years I've built the system that runs CogNovaMX: an AI-operated content company on the open MX standard. An AI chief of staff works inside a repository that describes, checks and heals itself. A partner who gives me access to their GitHub repository gets a Palimpsest, which talks to it through signed, encrypted cogs, and a project of their own, managed the same way as ours; one partner's codebase runs her already. Analytics from a partner's project only ever improve Maxine, never work against the partner. A partner can also purchase any number of dedicated fleet machines, each an Apple Mac, that run Maxine for their projects. For regulated industries, CogNovaMX leases an air-gapped Maxine fleet controller: the machine that runs the whole estate's Maxines inside the customer's own walls, with no open internet access. There's a charge for each additional Maxine added to that private fleet. CogNovaMX onboards the company's own knowledge to the controller Maxine, so she runs on what that business knows. The partner programme is in alpha until late January 2027. A registry of verified documents, REGINALD, has been in alpha since May 2026. A web audit measures the gap between what a machine receives and what a browser shows. And Intent CMS is content that manages itself: the fields the platform needs travel inside the file, and publication strips them so only what a reader needs remains. The Palimpsest reads the cogs Intent CMS publishes, so a site on Intent CMS is one the Palimpsest can use in full. It's the working proof that the method in the books holds, and what it proves goes back into the standard.

Industry Perspective

As a member of Boye & Company's CMS Experts Group and regular industry speaker, I stay connected with emerging trends while grounding recommendations in proven approaches. My work is a practical reference model for what agent standards are converging on: interoperable, multi-vendor systems where clarity serves everyone. For a decade CMS circles knew me as "The AEM Guy", a name earned architecting Adobe platforms. Now I'm The MX Guy, and the work follows one line borrowed from Steve Krug: don't make the machine think. I still prefer being known for helping teams make sound strategic decisions that prepare for agentic workflows using MACH principles of modularity, openness, and composability. I co-founded The Gathering, the independent standards body for Machine Experience, and Tityra Ltd, which shows organisations whether machines can actually understand, trust, and use what they publish.

Since 1977, I've been solving content distribution problems across every generation of technology, from assembler code through Superbase, BBC systems, Adobe AEM, and Edge Delivery Services. All variations of the same fundamental challenge: content that works for different consumers. Now those consumers include AI agents, and the patterns I've been refining for nearly five decades apply more than ever.

I work exclusively through Digital Domain Technologies, focusing on engagements where experience and objectivity matter most. Available for interim consultancy roles, advisory projects, and strategic reviews, not seeking full-time positions.

If you're evaluating Edge Delivery Services for agent readiness, planning major AEM changes in an AI-native world, or need architectural guidance that prevents problems before they're expensive to solve, let's talk about how strategic partnership might help.

Strategic advantage comes from having the right frameworks in place before you need them, frameworks that recognise "agent-ready" means accessible, observable, and universally comprehensible. That's where experienced advisory makes the difference.