AI insights — the behaviour-watching advisorOWNER: ENGINE — AI/orchestration (insight-engine), Folo-first (Mav’s map, 2026-08-09) — FG mounts the same engine as its “AI watchdog”; one surface, per-product insight content. Guard-rail: the advisor watches behaviour and drafts changes — it never ships one. Low-risk changes are one-click Apply (audited, reversible); anything material is Propose for review — routed to a human and the roadmap. Confidence and evidence are mandatory on every card; no evidence, no suggestion.
folo.link scope (§17 + §17e): this page READS the engine’s per-product insight streams — every card is chipped with the platform whose behaviour produced it. An Apply works in place on every skin: the write always routes via the owning product’s admin API (one write path, audited in its log) — the deep-link into that product’s admin is only the quiet secondary.
folo.link scope (§17 + §17e): this page READS the engine’s per-product insight streams — every card is chipped with the platform whose behaviour produced it. An Apply works in place on every skin: the write always routes via the owning product’s admin API (one write path, audited in its log) — the deep-link into that product’s admin is only the quiet secondary.
Ranked, evidenced, two-tier actioned — the engine suggests; a human or an audited low-risk apply decides.
Demo dataLive on Folo
Open insights
7
ranked by severity × confidence
DEMO DATA · source: insight-engineHigh severity
3
need a decision this week
DEMO DATA · source: insight-engineApplied this week
1
low-risk one-click · audited
DEMO DATA · source: audit logProposed for review
2
routed to human + roadmap
DEMO DATA · source: roadmap boardHighINS-0412Confidence 91%Friction Folo behaviour
Age-gate retry loop is dropping verified users
Evidence
- 1,940 sessions in 28d hit the age-gate twice within 10 minutes
- 62% of second-attempt users had already passed verification this device
- Drop-off after the second prompt is 3.1× the funnel baseline
Suggested change: honour the existing verification token for 24h on the same device before re-prompting; falls back to a full check on any risk signal.
Effect of Propose: logs a proposal row and files a job card to Folo’s human owner + the roadmap board — nothing changes for users until Folo ships it.
Effect of Propose: logs a proposal row and files a job card to Folo’s human owner + the roadmap board — nothing changes for users until Folo ships it.
material change · routed to Folo’s human owner + roadmap · sends nothing on click
HighINS-0409Confidence 87%Monetisation Folo behaviour
Checkout abandons spike when the wallet balance is nearly enough
Evidence
- 28d: 412 abandoned unlocks where wallet covered ≥80% of the price
- Median gap $1.40 — top-up minimum is higher than the gap in 88% of cases
- Same users complete at 2.4× rate when the gap is under $1
Suggested change: offer an exact-gap top-up at checkout instead of the fixed minimum. Touches the payment flow — human review required.
Effect of Propose: logs a proposal row and files a job card to Folo’s human owner + the roadmap board — the checkout changes nothing until reviewed and shipped.
Effect of Propose: logs a proposal row and files a job card to Folo’s human owner + the roadmap board — the checkout changes nothing until reviewed and shipped.
material change · routed to Folo’s human owner + roadmap · sends nothing on click
MediumINS-0405Confidence 82%Retention Folo behaviour
Expired-card subscribers churn silently instead of being nudged
Evidence
- 231 subscriptions lapsed in 28d on a card-expired failure, no retry prompt shown
- Comparable platforms recover 30–40% of these with a single in-app nudge
Suggested change: enable the existing dunning banner for card-expired states only. Copy already exists; flag is off. Low-risk and reversible.
Effect: card-expired subscribers see the dunning banner on their next Folo session; nothing is sent to anyone; reversible by the same switch.
Effect: card-expired subscribers see the dunning banner on their next Folo session; nothing is sent to anyone; reversible by the same switch.
applies in Folo via its admin API · audited there — one-click only because Folo wired this switch · low-risk · reversible
MediumINS-0398Confidence 79%Creator Fanogram behaviour
Creators who post within 48h of approval earn 3× in month one
Evidence
- Cohorts 2026-03 → 2026-07: first-post-within-48h creators reach first sale 2.8 days sooner
- 41% of approved creators still have zero posts at day 7
Suggested change: add a first-post checklist step to creator onboarding with a day-2 reminder. New comms to creators — human review.
Effect of Propose: logs a proposal row and files a job card to Fanogram’s human owner + the roadmap board — no creator is contacted by the click.
Effect of Propose: logs a proposal row and files a job card to Fanogram’s human owner + the roadmap board — no creator is contacted by the click.
material change · routed to Fanogram’s human owner + roadmap · sends nothing on click
LowINS-0391Confidence 84%Growth Folo behaviour
Search-landing visitors bounce on the slow hero image
Evidence
- LCP on /discover is 4.1s at p75 — hero image is 1.8MB uncompressed
- Bounce on that route is 1.6× the site baseline for search-sourced sessions
Suggested change: serve the already-generated AVIF rendition from the CDN. Asset exists; a header flip. Low-risk and reversible.
Effect: /discover’s hero serves the AVIF rendition once the CDN cache window (~5m) rolls; users see the same image, faster; revert re-serves the original.
Effect: /discover’s hero serves the AVIF rendition once the CDN cache window (~5m) rolls; users see the same image, faster; revert re-serves the original.
applies in Folo via its admin API · audited there — one-click only because Folo wired this switch · low-risk · reversible
✦
The two-tier pattern is the contract (Folo reference wording, adopted verbatim): every card ends in either “Apply change” (low-risk, one-click, audited) or “Propose for review” (material, routed to human + roadmap). There is no third button, and neither button contacts a user. Write path (§17-2 + §17e): an apply is always the owning product’s write — one-click exists only where that product wired the switch — and it works in place on every skin: folo.link owns no platform state, so the write routes via the owning product’s admin API and audits in its log; “open in <product> admin ↗” is the quiet secondary, never the only path.