LoginSign Up
← Back to all sparks
A

Alhena AI

AI-ASSISTANTS
Velocity5.0

AI shopping and support concierge for ecommerce (formerly Gleen AI).

alhena.ai ↗

Alhena is building the scoreboard for shopping agents it also competes in.

agentic-commercebenchmark-researchai-visibilityretail-aicategory-definition
◆Current state
The feed has consolidated around one piece of original research: a 2026 stress test running fifteen live AI shopping agents through real storefronts as ordinary shoppers. The headline numbers repeat across several posts — all fifteen could answer questions, nine could sell, four could complete a return or order change, and one remembered the shopper on a return visit. Around that sit vertical censuses of who is actually live in health and wellness retail, an attribution model for measuring agents, and comparison pages against AI visibility platforms including Profound.
◆Where it's heading
Alhena is defining the category's measuring stick and choosing metrics where most competitors fail — acting rather than answering, and remembering across sessions. Publishing a dated census that separates shipped assistants from announced intent serves the same purpose: it establishes Alhena as the arbiter of what counts as live. The vertical focus on supplements and wellness, with its FDA claims boundary and subscription economics, looks like a deliberately chosen beachhead rather than broad retail coverage.
◆Prediction
Expect the stress test to become a recurring dated benchmark with more agents and more verticals, and for the act-and-remember gap it identifies to be positioned as what Alhena's own product closes.

◆Recent moves

  1. 1d ago

    Do AI Shopping Assistants Remember You? Only 1 in 15 does

    The memory cut of the fifteen-agent stress test: only one assistant recognized a returning shopper. It isolates persistence as the rarest capability in the category, but it is research publishing rather than a product change.

    View source ↗
  2. 4d ago

    Why Can't My AI Agent Complete a Return? Inside the answer-to-act gap

    The returns cut of the same study: all fifteen agents could explain a return, four could actually process one. The widest gap in the dataset and the sharpest argument in the series.

    View source ↗
  3. 6d ago

    The State of Agentic CX in 2026: Why AI Shopping Agents Answer in Unison but Act Alone

    The parent report carrying the full scoreboard — fifteen can answer, nine can sell, four can act, one remembers. The anchor the other posts are cut from.

    View source ↗
  4. 20d ago

    Who's Actually Live: AI Assistants in Health & Wellness Retail (July 2026)

    A dated census of live AI assistants in health and wellness retail, separating shipped products from announced intent with evidence for each. Category-arbiter content, not a release.

    View source ↗
  5. 25d ago

    Measuring AI Agents for Wellness Brands: Benchmarks and an Honest Attribution Model

    A three-tier attribution model separating engagement, attribution, and incrementality, alongside a 4.68% conversion figure for LLM-referred wellness traffic. Measurement framing that argues the common benchmark measures a traffic source rather than an agent.

    View source ↗
  6. 25d ago

    The Wellness Brand's AI Agent Playbook: Knowledge, Guardrails, and Subscriptions

    A deployment playbook for wellness brands covering knowledge grounding, the FDA claims boundary, subscription flows, escalation, and measurement. Practitioner content supporting the vertical beachhead.

    View source ↗