A working collection of automated live builds.
A webhook (or an on-demand EDGAR check) takes a raw 10-Q, 10-K, or earnings-call transcript and returns a structured FP&A brief: YoY growth, margins, guidance direction, ranked risks, and a scored sentiment read — no analyst in the loop.
A webhook receives an inbound lead, Claude scores it 0–100 against a fit & intent rubric, tiers it hot, warm, or cold, and drafts the follow-up itself — a Slack/email alert for hot leads, a nurture email for warm or cold ones.
A notetaker bot hands off the transcript the moment a call ends, and every action item comes out automatically — task, owner, due date, priority — with a drafted Slack/email DM to the owner, or a triage alert to the organizer if nobody claimed it.
A code node computes variance and a baseline forecast for every budget line item deterministically; Claude explains the likely driver, refines the forecast with qualitative context, and drafts a severity-toned message to the owner — alert, watch, or a quiet on-track note.
A code node computes a naive trend baseline for every line item deterministically; Claude decides whether trend, seasonality, or a known future event should drive the projection, builds a base/upside/downside band, and drafts a forward-looking summary for the stakeholder — no actual required, no severity threshold to trip.
A webhook receives a contract's full text, Claude flags the risky clauses across liability, termination, IP, SLA, and venue, scores overall risk 0–100, and drafts the tier-appropriate action itself — a legal escalation, a counterparty negotiation email, or an internal approval note.
A webhook receives an invoice and its matched purchase order, Claude flags price, quantity, and line-item discrepancies, decides whether to auto-approve, hold, or dispute, and drafts the tier-appropriate action itself — an AP ledger note, an internal review flag, or a vendor dispute email.
A webhook receives an incoming support ticket, Claude reads it for tone and business impact, tiers it escalate, standard, or self-serve, and drafts the tier-appropriate response itself — an urgent internal page, a support reply, or a self-serve answer pointing to the right help-center article.
A Google Sheets trigger picks up new budget rows straight from a live spreadsheet; a code node computes variance and severity deterministically, Claude explains the driver and drafts the owner message, and the result writes back into a Variance Report tab in the same Sheet — no Data Table, the spreadsheet itself is the system of record.
New automation pipelines get added here as they go live — this space is reserved, not empty.