Crew 01: live, business development
The crew finds each opportunity and defends every word of it.
Six narrow scouts run the weekly sweep of government grants, public tenders, corporate intelligence, regulatory triggers, industry events and academic collaborations. An editor dedupes, scores and fills the EN and 简中 gaps. A verifier corroborates every claim with fresh independent searches, checks that every link is live and tests every deadline for plausibility. A human BD owner then promotes the verified pipeline into the CRM. We built it and we run it, and the live instance feeds WaterDoctor's BD pipeline every week.
Coverage
Six lanes, two working languages and a curated source registry.
Source lanes
Grants, tenders, corporate intelligence, regulatory triggers, events and academic collaborations. Each lane has one narrow scout, with its own thematic brief and source list.
Markets
Singapore. China at national level plus Jiangsu, Zhejiang, Shanghai, Guangdong, Fujian, Hainan, Shandong and Guangxi. Hong Kong and Taiwan. ASEAN: Vietnam, Indonesia, Thailand, the Philippines and Malaysia.
Languages
English and 简体中文 end to end. Local-language sources are read in their own language, and every opportunity has parallel EN and CN titles, summaries and eligibility notes.
Workflow
Six scouts, one editor, one verifier and one human owner.
scouts.agent (×6)
Source-specialist sweep
Each scout owns one source lane and one set of thematic queries. Grants tracks SFA, Enterprise Singapore, A*STAR, MARA 渔业发展补助资金, MOST 蓝色粮仓, NSFC, provincial DoSTs and ADB blue-economy TA. Tenders tracks GeBIZ plus the CN, ID, PH, TH, VN and MY procurement portals. Corporate, regulatory, events and academic each work the same way against their own brief.
editor.agent
Dedupe, score and translate
Pulls every scout's findings, dedupes by source URL across lanes, scores fit, urgency and value on a 1–5 scale, fills in any missing EN or 简中 side, and drops anything below the fit bar. It doesn't fabricate anything or paraphrase into the source data.
verifier.agent
Corroborate, then sign off
Each surviving opportunity gets a fresh independent search for live corroborators, a HEAD/GET on every candidate URL, and a plausibility test on the deadline. The verdict is VERIFIED, FLAGGED or REJECTED, and it stays on the record.
BD owner (human)
Promotes to the pipeline
No opportunity reaches CRM follow-up unsigned. The owner approves, edits, sends it back or kills it. The crew does the legwork and a person owns the pursuit.
Source rigour
False precision gets rejected.
Two corroborators, or a tier-1 primary
Every opportunity carries its primary source plus a list of corroborating sources. VERIFIED requires either two live independent corroborators or a tier-1 primary that's still live the week we send it.
Title sanity is the cheapest reject
An empty title, or a title that equals the URL, is REJECTED before a single corroborating search is spent.
Deadline plausibility, every time
A missing deadline is FLAGGED for the human, and so is a deadline more than five years out. Evergreen-looking links can't pass as live grants.
Auto-retry on a suspicious reject rate
When more than 70% of a batch comes back REJECTED, the rejected subset re-runs once before being saved, so a flaky grounded search doesn't end up as quality data.
A registry instead of a crawl
Sources are tier-graded and curated. The crew can propose a new source, which lands in a Proposals queue, and it's scanned only after the operator approves it.
Eval harness
How every opportunity
gets graded.
Every opportunity is graded before sign-off on corroboration depth, source-tier authority, deadline plausibility and sector fit against the brief. Anything below threshold goes back to the verifier or the originating scout, and stays out of the CRM. The crew runs weekly, the registry is reviewed monthly and the brief is reviewed quarterly.
Brief us