Francisco Acasuso · AI Strategist & Builder

Today’s top opportunities in the Market Intel System sample workspace. Companies and assets are real and cited; signals, scores and people are illustrative.

Opportunity map in the sample workspace: who works with the target company, and the best route in, here through an existing client. Illustrative sample data, not client output.

Category: Client SaaS (weekly delivered market-intel product) Status: Delivered weekly in production in 2026 Time window: March – September 2026 Live URL: private (client dashboard with logins)

Purpose

The point is timing: find business opportunities at the right moment, so the client is in the conversation before competitors are. Weekly market intelligence for an offshore preservation services provider. It finds the buying moments in the sector (rigs going into lay-up, units coming off contract, ageing installations, maintenance campaigns), ranks the companies behind them, and tells the sales team who to call, why now and who to ask for.

Who it is for

The client's sales function, every Monday morning.

Stack

Pipeline: TypeScript on Trigger.dev v4 (scheduled Monday run, Sunday checks, weekly discovery jobs); started as a Python prototype
Data: Supabase, cumulative across runs
AI: Anthropic Claude for extraction, briefing, outreach drafts and the pre-send review; never in the ranking
Enrichment: contact and search providers, used for the top opportunities only
Dashboard: Next.js with logins and roles
Delivery: Excel workbook and e-mail summary
Observability: Langfuse tracing of model calls

My contribution

Led the product from the first client conversation to the weekly production run. Scoped the service with the client around one question: who to call this week, why now and who to ask for. Proved the sources and scoring in a prototype, then designed and implemented the production pipeline: more than 50 Norwegian and English sources, AI extraction that drops any signal its page does not support, company matching on identifiers instead of names, and a deterministic two-stage scoring model the client adopted. Built the safeguards that let a sales team act without checking first: corroboration across independent sources, an evidence check on every recommendation, a pre-send review by a second model, Sunday dry-runs, self-healing sources and a hard cost cap on every run. Delivered the Monday workbook, e-mail briefing and role-based dashboard, then turned the core into a platform behind the custom versions for other clients, the sector variants and the public sample demo.

Notable capabilities

More than 50 monitored sources in Norwegian and English: operators, public registers, fleet-status reports and trade press
AI extraction, with signals the source page does not mention dropped
Deterministic two-stage scoring with visible reasoning; tiers HOT, WARM and MONITOR
Company matching by key, domain and alias clusters; fuzzy name matching measured and removed
Norwegian-shelf scope from the Norwegian Offshore Directorate and Brønnøysund registers
Corroboration bar and an evidence check on every recommendation; withheld companies listed on the Source Audit
Pre-send review of the finished deliverable by a second model, a Sunday dry-run and self-healing sources
Per-run cost alert and hard cap, and one model policy for the whole pipeline

Outcome

Delivered every Monday in production in 2026. Every ranking explainable from its sources, unsupported recommendations kept out of the report, and running cost capped per run.

Evidence

Case study: Related case study

Skills demonstrated

trigger-dev-v4, typescript, supabase, claude-extraction, deterministic-scoring, entity-resolution, public-register-integration, evidence-checks, llm-pre-send-review, self-healing-sources, langfuse-cost-tracking, bilingual-nb-en, excel-delivery

Preservation – weekly market-intelligence service