← Back to Support
Weekly · Support · Week of June 15, 2026

Support platforms push AI agents and CRM-wired feedback ops, while a US directive forces Hatz to pull Claude Fable 5.

ai-agentshuman-handoffcrm-integrationmcpdata-residencyfeed-quality
Generated 5h agoDrawn from 6 products

The week in customer-support

The customer-support sector's real movement this week is concentrated in a handful of platforms wiring AI agents, native channels, and feedback automation deeper into the support workflow — set against a sharp reminder that model rosters now answer to regulators. Hatz AI added Claude Fable 5 across chat, apps, agents, and workflows on June 10 as its most capable long-horizon model, then disabled it three days later to comply with a US government directive, keeping its other Anthropic models live. That add-then-pull is the most consequential single event in the sector this window: capability expansion and external regulatory shifts collided inside one model selector.

Elsewhere the direction is steadier. AI-mediated conversations with clean human handoff, omnichannel reach, and feedback-to-revenue plumbing are the recurring threads. Twilio opened private beta for Apple Messages for Business and took Agent Connect to GA; respond.io shipped live-call transfer from voice AI agents to humans; Canny kept wiring its Autopilot feedback layer to Salesforce and HubSpot opportunities; and HelpCenter.io moved its knowledge base toward being an answer engine. A large share of the sector's tracked feeds, however, are content-marketing blogs rather than changelogs, so the genuine release signal sits with a smaller set of products than the roster suggests.

Leaders

Hatz AI had the most eventful week, landing two sparks back to back. It added Claude Fable 5 across chat, apps, agents, and workflows on June 10, pitched for days-long autonomous tasks with sub-agent delegation, then disabled it on June 13 per a US government directive while leaving its other Anthropic models in place. Underneath that churn it continued building MSP governance depth — per-tenant integration and custom-MCP controls, tenant-template defaults, and phone-agent post-call workflows.

Twilio shipped the sector's other two sparks. Apple Messages for Business entered private beta, giving brands a native, branded channel inside Apple's Messages app rather than plain SMS, and Agent Connect reached general availability as a Python/TypeScript SDK for conversational agents with persistent memory and cross-channel orchestration. In parallel it advanced EU (Ireland IE1) data residency, with SMS residency hitting GA and Studio plus TaskRouter entering private beta — compliance plumbing running alongside the capability expansion.

respond.io continued pushing the conversation lifecycle onto AI agents with humans kept in the loop. Its Voice AI Agent can now transfer a live call to a human mid-conversation without dropping the caller — a step toward production-grade voice handoff — backed by auto-closing conversations with AI-generated summaries, contact-source attribution, and a Cal.com integration that surfaces meetings inside the inbox.

Canny posted no sparks but seven improvements, all converging on one bet: turning its feedback board into an AI feedback-ops layer wired to revenue. Autopilot now auto-links open Salesforce and HubSpot opportunities to captured feedback, the Ideas beta reached the Core plan, on-demand auto-grouping landed, and new Slack DMs notify account owners when an idea they care about ships. It is the clearest example this week of feedback being connected to CRM revenue signals.

HelpCenter.io took AI Answers to general availability, shifting the product's core job from hosting help articles to answering questions directly. It paired that with a HubSpot Help Desk integration that surfaces the knowledge base inside agents' existing workflow, plus smaller polish (Unsplash backgrounds, in-place embed editing, portable articles).

Wildcards

LiveAgent is the off-pattern case: after a heavy mid-May drop of an AI Agent Work Distributor, MCP tools, and OAuth 2.1 for claude.ai connector integration, its week is almost entirely hardening — a security fix, an API custom-fields 403 repair, and a tickets-history query that degenerated under load on a cold buffer pool. It is the cleanup phase that follows a big capability release rather than a new bet, and the scale fix is the substantive one.

Themes that compounded

  • AI-to-human handoff is becoming table stakes: respond.io's live-call transfer and LiveAgent's earlier AI Agent Work Distributor both keep a human in the loop rather than fully automating.
  • Feedback and conversations are being wired to CRM revenue: Canny auto-links Salesforce/HubSpot opportunities and DMs account owners on ship.
  • MCP and agent infrastructure keep widening across the sector — Hatz's per-tenant MCP controls, Twilio's Agent Connect SDK, and LiveAgent's MCP surface all point the same way.
  • Knowledge bases are turning into answer engines, with HelpCenter.io's AI Answers GA the cleanest example.
  • Model rosters are now exposed to external regulation: Hatz's three-day Claude Fable 5 lifecycle shows availability can reverse on a government directive.

Watch this week

The near-term signals are concrete. Hatz's customers needing long-horizon work will be steered to its remaining Anthropic models now that Fable 5 is disabled, and any further roster churn is worth watching. Twilio's EU residency should keep marching from beta to GA across more products, and Apple Messages for Business is positioned to graduate from private beta. Canny's broadening of Ideas to the Core plan suggests a push to make it the default feedback home, while LiveAgent's fix-heavy cadence implies its next visible feature wave on the AI/MCP surface hasn't surfaced yet. Note too that a large slice of this sector's tracked feeds — Supportbench, Social Intents, Spiceworks, ProProfs KB, Comm100, Knowmax, and Zoho Desk — are marketing or editorial blogs, not changelogs, so their silence reflects crawl-source choice, not stalled products.