Skip to main content

Proof

We run it on our own business.

Every number on this page was pulled from the kanban and fleet config on bailian as of Nov 19, 2026 — the same infrastructure that runs this site. No estimates.

Case study #0 — Tacavar

The Tacavar pipeline

Situation → Automations deployed → Outcome

Situation

We had the same recurring-work problem we sell against: blog cadence, SEO monitoring, research aggregation, lead follow-up. None of it was getting done reliably by humans.

Automations deployed

A multi-profile agent fleet on cron: signal-collector → signal-enricher → writer → editor → verifier → distributor, plus SEO strategist and analyst profiles. Specialist profiles pick up cards on the board; cron dispatches the cadences.

Outcome

6,700+
jobs completed
110+
this week
34+
agent profiles
160+
auto-recovered

Reliability improved as the fleet was dialed in: crash rate fell from 86% in May to 17% in June. Average runtime per completed job is ~6 min. The verifier profile alone has 2,100+ completed runs — verification is structural, not a tagline.

pipeline.bailian
live · daily cadence
croncadenceboarddispatchcollectorsignalswriterdraftseorankverifiergatepublishshipescalate→ telegram
entry / verifiedagenthuman escalationflow particles indicate active work moving through the pipeline

Case study #0 — in practice

The pipeline, unedited.

A real content job moving through the board, and the daily digest the system produces. Nothing staged.

Kanban task detail view: a content job ('Create Therapeutic Vaccine Platform Analysis') showing the tenant (research-brain), agent worker log, and verification steps as it moves through the pipeline.
↑ A content job end-to-end: tenant, worker log, verification.
The Tacavar Daily Intelligence Digest on Telegram: an auto-generated digest with citation counts, signal totals, and top-scored opportunities — produced and delivered by the agent pipeline.
↑ Auto-generated daily digest — research, scored and delivered.

Case study #1 — marketing

SEO & content pipeline

Situation

An SEO/content operation where blog cadence and rank tracking were the first things to slip when ops got busy.

Automations deployed

Weekly editorial calendar generation, per-site SEO refinement passes, rank tracking on a weekly cron, and drafts written → verified → published on schedule.

Outcome

219 posts in the live corpus, published and verified by the pipeline. We don’t publish invented organic-growth percentages here — we ship what’s verifiable.

Case study #2 — healthcare ops

Healthcare distribution operations

Situation

A biologics and specialty-products distribution operation drowning in document triage, intake, and recurring reporting.

Automations deployed

Document triage and routing, intake follow-up, and recurring operational reports — each a recurring card on the board.

Outcome

The recurring work runs on cadence instead of waiting for a human to remember it. (We describe the automations deployed rather than invent revenue figures.)

Run the same loop on your ops.