AI answer monitoring · bolddesk

Headline

No data was collected. No MCP servers were connected for this run, so the openrouter ask_engine tool was unavailable. All 42 rows (7 prompts × 6 engines) are marked not-checked, and there are no mention, citation or accuracy numbers. Note that a complete run for BoldDesk for the same month (2026-10) is already published: reports/publish/bolddesk/ai-answer-monitoring/20261109-2041_ai-answer-monitoring_bolddesk_29e6. That run remains the current baseline. This run should probably be discarded rather than published.

Score

No score. The checklist does not define one, so overall_score is null.

Top 3 actions

1. Don't publish this run as a data point. Discard it, or publish it only as a record of the failed attempt. It doesn't replace the 2026-11-09 baseline for 2026-10.
2. Connect the openrouter MCP server before the next run (see _reference/mcp-and-tools.md), then rerun for month 2026-11 in the first week of December so it gives a real month-over-month comparison against the baseline.
3. Follow up the open urgent item from the baseline: Gemini claimed ISO 27001 and PCI DSS compliance for BoldDesk (P7-gemini). This run couldn't check whether that is still happening. Confirm that the product marketing lead and the COE Team were told, and that the proposed security-page fix is tracked.

Findings

AI answer monitoring: bolddesk, 2026-10

Run by: Claude | Location: n/a (no engine calls made) | Date: 2026-11-12

Planned prompt text (same as the baseline, unchanged): P1 "What are the best AI help desk / AI customer support software?"; P2 "What is the best AI help desk / AI customer support software for an enterprise / midsize company?"; P3 "BoldDesk vs Zendesk: which should I choose?"; P4 "What are the best alternatives to Zendesk?"; P5 "What is BoldDesk, how is it licensed and who uses it?"; P6 "How to automate tier-1 support tickets with AI" (GTM plan AI-monitoring prompt 5); P7 "Is BoldDesk secure and compliant for enterprise use?".

Summary

• Mention rate: not checked (0 of 42 rows collected). There is no change vs last month to report. For reference, the published 2026-10 baseline was 12 of 28 API rows.
• Citation rate: not checked. The baseline was 11 of 28 API rows.
• Wrong or outdated facts: not checked. The baseline logged 5: P3-claude, P5-gemini, P5-claude and P7-gemini were Wrong, and P3-gemini was Outdated.

Gaps and fixes

No new gaps, because no answers were collected. The gap list in the baseline run (20261109-2041_..._29e6/summary.md) is still the open list.

Not checked

• ChatGPT, Gemini, Perplexity and Claude, all prompts (P1–P7), 28 rows: no MCP servers were connected, so the openrouter ask_engine and list_engines tools weren't available. I didn't use the running agent as a stand-in for the Claude row because it isn't a clean session (it has repo context and memory of the baseline) and has no web search, so its answer wouldn't be comparable. I didn't use web search as a substitute either, because it doesn't reproduce any engine's answer.
• Google AI Overviews and Microsoft Copilot, all prompts (P1–P7), 14 rows: neither has an API and both must be checked by hand. This run was unattended.
• Month-over-month comparison: no data from this run. The only earlier run is for the same month (2026-10), so it isn't a "last month" comparison anyway.
• Phase 2 (gaps, fixes, hand-off, weekly-review deltas): nothing new to hand off.

Proposed changes (need approval)

This run proposes no new changes. The 8 proposed changes in the 2026-11-09 baseline run are still open as far as this run can tell. The most urgent is correcting the ISO 27001 and PCI DSS claims behind P7-gemini.

Data sources

• MCP servers: none were connected, and no engine calls were made (0 ask_engine calls, no API cost).
• Repo files read: CLAUDE.md, .claude/skills/ai-answer-monitoring/SKILL.md, demand-gen/search/ai-answer-monitoring.md, reports/schema/*, products/bolddesk/gtm-plan.md (AI-monitoring prompts and competitors), and the published baseline run reports/publish/bolddesk/ai-answer-monitoring/20261109-2041_ai-answer-monitoring_bolddesk_29e6/.
• Raw files: raw/run-status.json records the engine availability, the planned prompts and a pointer to the prior run.
• Choices made in this unattended run: same prompt wording, top competitor (Zendesk) and P6 choice as the baseline, so a rerun stays comparable. The metric period is the run's date_range (2026-10-15 to 2026-11-11), and previous is null because the baseline is the same month, not the previous one.