Malcolm Peralty, based near Toronto, brings roughly 20 years of WordPress experience to his role at Pressable. He began as a full-time blogger on early WordPress, moved into project and agency work, spent time in the Drupal/Acquia ecosystem, and now serves as a technical account manager (TAM) focused on long-term site strategy rather than sales or first-line support.
Role and focus
– As a TAM, Malcolm helps customers plan how sites should evolve over months and years. He advises on architecture, performance, plugin selection, end-of-life risks, and service tiers. Rather than answering quick support tickets, his work is consultative: sizing resources, recommending optimizations, and sometimes advising a plan downgrade after successful improvements. TAMs also act as a bridge to Pressable’s WP Cloud teams when deeper infrastructure work is required.
Pressable’s platform approach
– Pressable runs its own WP Cloud instead of building on top of public clouds like AWS or Google Cloud. Owning the stack allows tighter control over hardware and resource provisioning.
– The platform targets a spectrum of WordPress sites: highly cacheable brochure sites that need little hands-on work, and complex, mostly-uncached sites such as WooCommerce shops or LMS installations that require consultative management under load.
– Pressable prioritizes resiliency and real-world concurrent performance over single-user speed benchmarks. While some providers look faster on a single request, they may fail under heavy concurrent traffic; Pressable optimizes for predictable behavior under load.
Performance and optimization challenges
– A recurring issue is educating customers on trade-offs. Page builders, certain plugins, and integrations (for example, Facebook for WooCommerce) often break cacheability or create heavy server-side compute.
– Pressable combines server-side design with customer guidance on architecture and plugin choices. One notable infrastructure decision is a worker model that maps one worker per VCPU, creating dedicated lanes for requests. That contrasts with the more common high-density worker models and yields more predictable concurrency behavior.
– Many sites remain fully cacheable and perform well out of the box. Problems arise when user-specific dynamic pages (shopping carts, course progress, personalized dashboards) prevent effective caching and require more CPU and worker capacity.
Infrastructure priorities
– Malcolm calls out several technical investments Pressable is pursuing:
– Improved database replication and real-time streaming to cut failover latency and avoid lost transactions during peaks like ticket drops or product launches.
– Virtual cluster and multi-datacenter strategies that make geographically distributed deployments appear local, improving resiliency and failover behavior.
– Better human-readable auditing, logging, and forensic tools so teams can quickly trace who changed what and when without relying solely on backups.
AI and the Managed Control Panel (MCP)
– Pressable is developing an MCP: an AI-driven control layer that exposes control panel actions via natural language. The vision is that anything clickable in the control panel could eventually be performed by AI on the user’s behalf.
– Example workflows: a developer could ask the MCP to spin up a sandbox, push new code, sync production files and DB to staging, and notify the team when ready. An admin could request a list of sites with a particular plugin version and have the MCP update that plugin across many sites.
– Because the control panel is API-driven, connecting AI tools to actions is straightforward technically. The harder work is defining clear semantics, inputs, outputs, and expectations so the AI acts safely and predictably.
Safety, guardrails, and human oversight
– Malcolm stresses safety-first design. Backup systems, simlinked core files, and other architecture choices limit catastrophic outcomes. For instance, WordPress core files are protected via simlinks so a mistaken command can’t wipe the core; plugins, uploads, and DB remain recoverable from backups.
– MCP rollout will be phased and include human approvals for high-risk operations. System prompts, verification steps, and human-in-the-loop checks are part of the plan to prevent destructive automated actions.
– Early adopters are likely to be agencies and developers who can troubleshoot issues; the tool delivers the most value at scale when managing dozens or hundreds of sites rather than for single-site hobbyists.
Observability, audit logs, and rollback
– AI-driven changes make comprehensive, readable audit logs essential. Logs must capture steps taken and enable clear, reversible actions. Pressable already retains hourly DB backups and daily filesystem snapshots; AI workflows may require more granular backup and rollback capabilities.
– Malcolm uses AI to build diagnostic scripts that run WP-CLI and other checks, and anticipates AI summarizing findings, proposing fixes, and—if permitted—executing changes.
New costs and malicious activity
– AI agents and crawlers create new resource costs. Malcolm described an incident where an AI bot repeatedly toggled items in shopping carts, spawning uncached sessions and heavy load. Bots can then adapt to evade blocks.
– The industry must decide who pays for extra indexing, traffic, storage for audit logs, and other AI-induced resources. Some customers will resist extra costs, so hosts must balance features with detection and mitigation of abusive traffic.
User experience and relationships
– AI tooling could shift how customers interact with hosts: less reliance on a traditional UI and more natural-language orchestration through IDEs, local agents, or developer tools.
– Malcolm advocates ‘human-first, AI-enhanced’ features. Agencies want faster delivery and the ability to prototype in real time—AI could enable spinning up a sandbox during a client meeting. There will also be a market for environments that avoid AI entirely, though those offerings may face scaling challenges.
Technical opportunities ahead
– Promise areas include WebAssembly and ephemeral in-browser WordPress playgrounds for fast experimentation, smarter caching strategies, compressed efficient audit logging, and improved DB replication.
– AI can help surface meaningful anomalies by ranking signals, reducing the manual effort of sifting through raw logs.
Contact and takeaway
– Pressable: pressable.com
– Malcolm Peralty: peralty.com; active intermittently on LinkedIn and in WordPress Slack communities.
Malcolm’s work at Pressable underscores a central tension: making WordPress accessible while managing complexity introduced by real-world, dynamic sites. Pressable aims to host simple, cacheable sites reliably while offering consultative support for resource-heavy, uncached workloads. AI-powered tooling like the MCP promises to automate routine, multi-site tasks and speed development workflows, but it also raises challenges around safety, auditing, cost, and bot traffic. Pressable’s approach combines phased rollouts, human-in-the-loop safeguards, robust backups, and continued investment in observability and infrastructure resiliency.