Case Studies
Discover how to use the productivity gains of AI workflows safely. Every case illustrates the zone split for a real-world implementation: which processes run in the AI Zone and which decision is made in the Human Zone.
Commodity Intelligence: From 6-hour cycle to 20 minutes
Energy-focused investment firm
Challenge
Analysts were manually downloading EIA weekly and monthly reports and the OPEC Monthly Oil Market Report, then cross-referencing figures against current portfolio positions and applying scenario logic by hand. By the time a synthesised view reached decision makers, the optimal window for action was already closing.
Approach
We built AI Zone adapters for each report source: structured extraction, portfolio position matching, and pre-defined scenario scoring. A human operator then triages the prioritised signal list via a structured interface, accepts or rejects signals, and assembles the final intelligence report for the portfolio team.
Outcome
The AI-to-triage handoff now completes in under 20 minutes. The intelligence report reaches decision makers swiftly after successful operator triage. Same day has become the expectation across all market insights. Coverage also improved: every data series is now processed, not just the highest-visibility figures.
Zone Split
- ·AI Zone: report fetching (EIA WPSR, STEO, OPEC MOMR), structured extraction, portfolio matching, scenario scoring, signal prioritisation
- ·Human Zone: operator triage of prioritised signals and report assembly; all portfolio decisions, position changes, and investment action
Accounts Receivable & Payable: 94% of invoices reconciled automatically
European event management firm
Challenge
The finance team managed invoices across four separate email accounts (client invoices, venue contracts, vendor payments, production costs) in different formats from counterparties with inconsistent reference conventions. Monthly reconciliation consumed two working days for two people and still produced occasional errors.
Approach
We built a read-only adapter across all four email providers that fetches every invoice regardless of format, extracting counterparty, amount, currency, due date, and reference codes into a unified schema. A manual step integrates bank transaction data. A matching workflow then determines payment status and prepares bookkeeping records. Mismatches and overdue dues are surfaced to a human operator for triage.
Outcome
94% of monthly invoices are now reconciled automatically. The operator exception review, covering only unresolved items, takes around 90 minutes, down from two working days. Duplicate payments and missed dues, previously caught only in monthly reviews, are now flagged in real time.
Zone Split
- ·AI Zone: read-only email ingestion across multiple providers, invoice extraction and normalisation, invoice-to-transaction matching, payment status assignment, bookkeeping record preparation
- ·Human Zone: operator reviews mismatches and overdue flags, resolves or escalates each exception; all decisions on disputed items including payment pursuit, credit notes, and counterparty communications
Inbound Lead Drafting: Hours of research to 10-minute review
Professional services consultancy
Challenge
Sales reps were spending up to two hours per inbound inquiry on research and drafting: locating relevant service documentation, matching the inquiry to past client work, and writing a contextualised response. Response times could take up to four days, and quality depended heavily on which rep picked up the lead.
Approach
We built a draft-only Gmail adapter without direct send ability by design. The workflow prequalifies inbound leads, matches each inquiry to relevant services via CRM and internal Notion documents, and creates a structured draft in the rep's Gmail account. Each draft is reviewed and sent (or discarded) by a sales rep. A feedback loop between drafts created and drafts sent refines the drafting tone over time.
Outcome
Reps spend an average of 8 to 12 minutes reviewing and editing drafts. Response times to inbound leads are consistently below 24 hours. Draft acceptance rate is high and continues to further improve as the feedback loop evolves the tone library and services to focus on.
Zone Split
- ·AI Zone: inbox monitoring (read-only), lead prequalification, CRM and Notion context matching, draft assembly
- ·Human Zone: rep reviews Gmail draft, edits, approves, and sends (or discards); all send decisions, final message content, and approval of feedback-loop updates to the drafting prompt library
Want to discuss how the zone model would apply to your organization?
Start the conversation