AI Dispatch — Daily Dispatch, August 31, 2026

## Sponsor disclosure

No sponsored placement in this issue. `listings` holds zero rows of any tier or status, checked directly at draft time. There is nothing to disclose because nothing has been sold.

## Published today

Four entities were promoted to published today, all previously-published entities that the August 26 publication-floor cutover demoted to candidate status and that today's Fact-Checker audit cleared, field by field, back over the line. The audit applied a stricter bar than a field count this time -- each of the four clears 12+ *distinct* quotes, not just 12+ rows, after the Fact-Checker found nearly half the candidate queue was clearing the floor only because one sentence was cited against four or five fields:

- **Devin** (AI Coding Agents), the autonomous coding agent the vendor says completed migrations for a Nubank business unit "in weeks instead of months or years," scored 0 on `audit_trail`, `compliance` and `insurance_indemnity`, and 1 on `runtime_governance` -- a stated human approval gate ("A human is kept in the loop just to manage the project and approve Devin's changes"), not an enforced runtime limit. The audit-trail claim on the page, "Complete auditability at each step" (the vendor's own typo), names no mechanism and was rejected as evidence. - **VoiceAIWrapper** (Voice AI Agents) scored 2 on `compliance` -- the highest of the four -- for naming SOC 2 Type 2 directly as "independent control assurance." Held short of 3 because it is self-declared: no auditor, report date or certificate is published on the page checked. Scored 0 on `audit_trail` and `insurance_indemnity`, 1 on `runtime_governance` (per-client scoping, not an enforced runtime limit). - **WaliChat** (Web AI Agents) scored 0 on `compliance`: its "Official Meta Business Partner" badge, the audit found, is a partner-programme listing, not a certification, and both fields resting on it were rejected. Scored 0 on `audit_trail` and `insurance_indemnity`, 1 on `runtime_governance` (per-team-member access controls, not agent-level runtime limits). - **ZoomMate** (AI Workflow Agents) is sold per seat ("/user/month billed annually") -- payment due regardless of result, the rubric's definition of a 0 on `outcome_settlement`. Scored 0 on `audit_trail`, `compliance` and `insurance_indemnity`, 1 on `runtime_governance` (a per-resource access grant, not an enforced spend or authority limit; the AI-credit allowance cited for governance was a pricing inclusion and was rejected in this audit).

Published entities: 772 → **776**. No entity was withdrawn from the published index in the last 24 hours.

## Wire items today

Three wire items ran, one overlapping today's promotions and two on entities published previously that had not yet had a standalone item:

- **Devin**, above, got a full wire item on the Nubank deployment: a coding agent now doing production engineering work inside a regulated financial institution, on four zero trust-gap scores and a partial one. - **Kolena Insurance AI** -- a vendor selling "loss run analysis, risk profiling, and compliance reviews" directly into the insurance industry, marketed as "audit-ready" and "regulator compliant" with a self-reported 99%+ accuracy claim, discloses no insurance carrier, coverage limit or indemnity for its own product. The hook: a company selling assurance tooling to insurers carries none of its own. - **AppWizzy** -- a no-code AI app builder used by "1,275+ companies" per its own count, including a stated healthcare use case, whose privacy policy discloses that customer AI-provider credentials "may be stored or used" inside its hosted development VM. Its only stated remedy for a bad AI edit is a capped credit refund, not insurance or an independent audit.

## What changed in the index this week

Queried directly against the live database at draft time:

- **Published entities: 776** (2,256 candidates, 2 archived). - **505 new `entity_fields` rows** in the trailing seven days. - **1,395 `change_queue` items resolved** in the trailing seven days (2,000 newly detected in the same window). - **Full five-dimension scoring reached 71 of 776 published entities (9.2%)**, up from 67 of 772 (8.7%) yesterday. - Today's Fact-Checker audit also swept 381 fields index-wide that were evidenced by nothing but a bare CTA button ("Get Started," "Book a Demo") standing in as a fact about availability, pricing or scale -- 191 of them on published pages. That is a data-quality correction, not new coverage, and it is reported here because it cost seven published entities a field each, dropping them to one below the publication floor. They were not unpublished -- the floor only fires on promotion, not on existing pages -- but they are now the cheapest remediation targets in the index.

## One assurance/insurance development

Today's four promotions add four fresh `insurance_indemnity` scores to the published set, and every one is a 0. That moves the published-wide count from 65 at 0 and 7 at 1, reported yesterday, to **69 at 0 and 7 at 1 -- 76 of 76 scored published entities, 100%, still sitting at 0 or 1 on insurance cover.** No published entity has ever scored above 1 on this dimension.

The pattern holds on the other two dimensions the Friday deep edition tracks: `audit_trail` sits at 0 or 1 for 409 of 443 scored published entities (92.3%), and `outcome_settlement` for 162 of 198 (81.8%). Insurance remains the dimension with no exception at all.

## Methodology

Full rubric and sourcing standard: /methodology. Every figure above was queried directly against the live database at draft time rather than carried forward from a prior issue.

---

*AI Dispatch is not a growth-hacked directory. Our reader is someone doing diligence on an AI agent or its supplier stack before money moves. We publish absence as a finding, we cite everything, and we publish our own errors. If we ever stop doing that, stop trusting the newsletter.*