Malcolm Peralty joined the Jukebox Podcast to discuss managed WordPress hosting at Pressable, his background, and how AI is reshaping hosting workflows.
Background
Malcolm has been in the WordPress space for about 20 years—using WordPress since 0.72, originally as a full‑time blogger, then moving into project management, agencies, and a stint in the Drupal/Acquia world as a technical account manager. He returned to WordPress to work at Pressable as a technical account manager (TAM).
What a Technical Account Manager Does
At Pressable, TAMs act as long‑term strategists for clients rather than salespeople. They advise on site architecture, plugin choices, performance optimizations, migration tiers, and future planning (one to two years out). TAMs sometimes handle pre‑sales sizing and act as escalation points when support issues need deeper, longer investigations.
Pressable’s Approach and WP Cloud
Pressable runs its own WP Cloud rather than hosting solely on public clouds like AWS or Google Cloud. That gives them control of hardware and the ability to tune infrastructure for WordPress-specific workloads. TAMs work with WP Cloud teams to present WordPress-driven performance insights and suggest hardware or configuration adjustments, while respecting the hardware team’s expertise.
Performance, Caching, and Realities of WordPress
Malcolm emphasized nuance in performance claims. Some hosts rely on aggressive caching, Redis, or fast CPUs to make a site seem fast under single-user tests, but those approaches can fail under real load. Key distinctions:
– Cacheable vs. uncached sites: Brochure/marketing sites that are cacheable perform well without handholding. E-commerce, LMS, or highly dynamic sites often become uncached for many sessions and require different resources and planning.
– Plugin and builder bloat: Heavy themes and page builders (Elementor, Divi, etc.), plus plugins that break cache, drive resource use. TAM conversations focus on trade-offs and alternatives.
– Worker model: Pressable uses a one‑worker-per‑one VCPU model (dedicated worker lanes) versus many workers per VCPU used elsewhere, improving resilience and consistent performance under load.
– Education: Customers often prefer not to dig into technical details, so part of the TAM role is translating performance trade-offs into business terms.
Infrastructure Trends and Challenges
Malcolm sees several areas of innovation and challenge ahead:
– Database replication and latency: High‑availability setups need better, faster replication. Even a couple of seconds of lost transactional data (during a product drop, for example) can be catastrophic. Pressable focuses on near‑real‑time streaming replication to limit data loss.
– Virtual clusters and multi‑region resiliency: Technologies that present multiple data centers as a single local server help with failover, availability, and reduced complexity of syncing.
– WebAssembly and ephemeral WordPress: Playgrounds that run WordPress in the browser (via WebAssembly) are interesting for demos and sandboxing.
– Better logging and auditing: Current logs aren’t always easy to audit or use to trace user actions; better, compressed auditing and AI summarization could help identify and fix issues faster without storing prohibitively large datasets.
Pressable MCP: AI Control Panel
Pressable is building an MCP—an AI‑powered control panel—to let users manage and deploy WordPress sites via natural language and AI assistants. Key capabilities and design principles discussed:
– Natural language actions: Users could ask the MCP to spin up sandboxes, push code, sync production files, update plugins across many sites, or create staging environments—using plain language.
– APIs underneath: The control panel is API driven, so AI agents can call the same endpoints as the UI. Exposing those endpoints to AI is technically straightforward; the harder part is making the AI understand intent, inputs, and effects.
– Human‑in‑the‑loop and guardrails: MCP will include confirmations and safeguards to prevent destructive actions. Pressable maintains backups (hourly DB, daily file system), simlinked core to protect WordPress core files, and encourages prompts that require confirmation before destructive tasks.
– Target users: Initially aimed at agencies and multi‑site managers who benefit from automation at scale—e.g., “run this update across 1,000 sites”—rather than single site owners.
– Integration with developer tooling: Examples include triggering sandbox creation from an editor or CI workflow, pushing code, and syncing uploads—all orchestrated via AI.
– Auditability and rollbacks: Emphasis on robust audit logs and backup/restore workflows so actions taken by AI can be reviewed and reversed if needed.
AI, Trust, and the Human Relationship
Malcolm acknowledged concerns about replacing human support with AI. Pressable’s stance is human‑first, AI‑enhanced: features should be useful to humans first, then made available to AI. Agencies are already pushing for AI to reduce time‑to‑delivery; hosts that don’t offer these capabilities risk being left out of that workflow.
There are social and trust considerations: customers may feel less personal connection if everything is AI automated. Conversely, AI may free up human agents to focus on higher‑value support for lower‑tier customers who need human handholding.
Operational and Cost Issues from AI
AI introduces operational load and new costs:
– AI bots and scraping: Automated agents can generate heavy traffic patterns (e.g., repeatedly adding items to carts), creating uncached sessions and resource drain. Bots can quickly evolve to evade simple blocks.
– Audit and storage cost: Detailed auditing, backups, and logs—especially if used to enable reversible AI actions—consume storage and compute. Determining how to pass those costs to customers fairly is an industry challenge.
– Maintenance expectations: Malcolm joked that humans won’t maintain all AI‑generated code—AI will—but that drives questions about security, performance, and long‑term upkeep.
Future of Interfaces and Workflows
Malcolm and Nathan discussed how the WordPress UI might recede for many users as AI interfaces proliferate. Workflows could move from the WP admin to AI assistants, editors, or other tools that speak to hosting APIs. That will create diversity in how websites are built and maintained, complicating support and standardization—but it also opens faster, more flexible ways to deliver proof‑of‑concepts and iterate with clients.
Practical Protections
Pressable’s practical safeguards include:
– Backups and simlinked core to avoid accidental deletion of core files.
– Recommendations for system prompts and human confirmations before destructive operations.
– Using AI to generate scripts and reports for performance audits, then having humans interpret and act—or allow AI to act under controlled workflows.
Contact and Links
For more on Pressable, visit pressable.com. Malcolm’s site is peralty.com. WP Tavern episode notes and links are at wptavern.com/podcast.
The interview highlighted the complexity behind managed WordPress hosting, the balance between human expertise and automation, infrastructure innovations (replication, virtual clusters, worker models), and how AI—if applied with care—could speed workflows while creating new operational and trust challenges.
