Insighter Digital
All work
2025 · Confidential — fintech

AI-augmented scraping pipeline replacing manual research

Daily extraction across 40+ sources with LLM normalization and a review queue. Cut a 3-person manual research team to one reviewer per week.

~30 hrs/week of analyst time recovered

The challenge

What the client was up against

A fintech research team spent ~90 hours a week — three full-time analysts — manually visiting 40+ data sources, copy-pasting into spreadsheets, and reconciling inconsistent schemas before the actual analysis could start.

Engagement

Scope, team, and stack

The shape of the work, not the marketing summary.

Timeline
14 weeks from kickoff to fully replacing the manual workflow
Team
1 delivery lead, 2 senior engineers, 1 fractional ML engineer
Tech stack
Python 3.12FastAPIPlaywright (headless Chromium)PostgreSQLAnthropic Claude (extraction + classification)Pydantic (schema validation)Dagster (orchestration)Sentry + structured JSON logsHostinger VPS + Tailscale
The work

How we built it

The brief

A fintech research team was spending most of every analyst's week manually visiting 40+ data sources, copy-pasting into spreadsheets, and chasing inconsistent schemas. They wanted scale without losing the analyst's judgment loop — the LLM was meant to do the boring extraction, not replace the human review at the edges.

What we built

  • Resilient scrapers — per-source extractors with retry/backoff, rotating user agents, structured-error reporting, and a kill-switch if a source's layout changes more than a configured threshold.
  • LLM normalization layer — pulls raw HTML/PDF/CSV into a single canonical schema; flags low-confidence rows for review rather than silently dropping them.
  • Review queue UI — one analyst sweeps the day's flagged rows in ~30 minutes instead of three people each doing a full day.
  • Audit trail — every cell's provenance traceable back to the source fetch + LLM call that produced it, so disputed numbers can be reproduced.

The hard parts

The scrapers themselves were straightforward; what was harder was the reviewability of the LLM step. An invisible normalization that silently mis-classifies a row would corrupt the downstream analysis with no way to know. Building the confidence scoring + review queue UI was where most of the project's discretion went.

We also ran an evaluation harness on every prompt change: a fixed corpus of ~500 historical rows with known-good outputs gets re-extracted on every prompt edit. A regression beyond 2% triggers a stop-the-line review before the change ships.

Outcome

Analyst time on data collection dropped from ~90 hours/week (3 FTE) to roughly 5 hours/week (one reviewer). The recovered capacity went straight into the analysis work that's actually the team's value-add. Source coverage has grown from 40 to 60+ since launch with no headcount increase.

Outcomes

What actually changed for the client

Quantified where we can, plainly stated where we can't.

  • Analyst time on data collection dropped from ~90 hrs/wk to ~5 hrs/wk
  • 3-person manual team replaced by 1 reviewer + automated pipeline
  • Every output row has a provenance trail back to source + LLM call
  • Schema drift now surfaces as a flagged review item, not silent data loss
Why us

What we bring to every engagement

The constants across our work — regardless of stack, size, or vertical.

  • Senior people only

    No bait-and-switch from your lead engineer in the pitch to a junior on day one. The people you scope with are the people who build.

  • Honest scoping

    We tell you what is in scope, what is not, and where the risks are — in writing, before you sign anything. Surprises kill projects.

  • Weekly demos, not big reveals

    You see real, running software every week. If a direction is wrong, you find out in week three, not week thirteen.

More work

Other case studies

A few more projects worth looking at.

  • 2025 · Insighter Digital (in-house)In-house

    A CRM purpose-built for life-insurance comparison funnels

    A multi-tenant CRM with lead-routing, Google Ads OAuth integration, agent dashboards, and conversion analytics. Powers daily operations for our partner agencies.

    Daily active across multiple agency tenants

    • Engineering
    • Data & Platform
    • Commerce
    Visit live
  • 2024 · Confidential — DTC brand

    Shopify replatform with headless checkout

    Migrated a legacy WooCommerce store to headless Shopify with a custom Next.js storefront and bespoke subscription logic.

    LCP 1.1s → 0.4s, +18% mobile conversion

    • Commerce
    • Engineering
FAQ

Common questions about working with us

If you do not see your question, just ask. We will tell you straight whether we are the right fit.

Have a project? Let's scope it.

A 30-minute call, no slides. Tell us the outcome you need and we'll tell you what it takes to ship — honestly.