Skip to main content

What we found and fixed for ecommerce brands. Learn more about what we found and fixed for ecommerce brands

Changelog

What's new in Morsegrid.

Releases ship every other week. Bigger ones make it here, with the why behind each change.

  1. ·v2.6Release#

    Builds: typed deploys for AOPs

    Every change to an Agent Operating Procedure is now a typed Build with diffs, reviewers, and a one-click rollback. The old free-form save flow has been retired.

    • Side-by-side diff view across prompts, tools, and policies.
    • Required reviewers per environment with role-based approvals.
    • Rollback any Build to any prior Build in a single click, even across model upgrades.
  2. ·v2.5Improvement#

    Watchtower: custom rules and SLO alerts

    Watchtower now supports custom evaluation rules written in plain language, plus per-AOP SLOs that page the on-call when breached.

    • Plain-language rule authoring with auto-generated test cases.
    • Per-AOP SLOs for resolution rate, escalation rate, and CSAT.
    • PagerDuty and Opsgenie destinations alongside Slack.
  3. ·v2.4Improvement#

    Voice: barge-in tuning and accent packs

    We rebuilt the voice activity detector and shipped tuned acoustic models for ten additional regional accents. Median time-to-first-token on voice is down 18%.

    • Configurable barge-in sensitivity per AOP.
    • Ten new accent packs, including IE, IN-South, ZA, AU-regional.
    • Latency dashboards now show TTFT per accent pack.
  4. ·v2.3Release#

    Snowflake and Postgres connectors are GA

    Both warehouse connectors graduated from beta. Connections are read-only by default, support row-level scopes, and ship with a query budget per agent.

  5. ·v2.1Improvement#

    Experiments: outcome-based promotion

    Experiments now promote winning variants automatically against an outcome metric you choose, with statistical guardrails and a kill switch on regressions.

    • Per-cohort traffic splits with stratification by channel.
    • Pre-registered metrics with multiple-comparisons correction.
    • Auto-rollback if any guardrail metric regresses by more than the configured threshold.
  6. ·v2.0Release#

    Suggestions: AI-drafted knowledge proposals

    Suggestions reads from your real conversations, identifies knowledge gaps, and drafts new help-center articles or AOP edits for your team to approve.

    • Gap detection grouped by topic, channel, and persona.
    • Drafted articles in your house style with citation backing.
    • One-click promotion into Notion, Zendesk, or Intercom.
  7. ·v1.9Release#

    EU data residency is generally available

    EU customers can now pin all conversation processing and storage to our Frankfurt region. The choice is per-workspace and applies retroactively to new conversations.

That's everything we've shipped in public. Read the playbooks for the why behind a lot of it.
Free Assessment

How systemized is your team?

Get a personalized Systemization Score across seven categories, plus the three highest-leverage areas for your next 30 days.

Take the Free Assessment ~10 minutes. Results instantly.