Work · Production engagements · 2025–2026

Three
systems,
running.

We keep the work-list short on purpose. Every case here is in production with a named owner on the client side, numbers reviewed monthly, and the system still running today.

Live since Q2 2025
Sectors Construction · Aerospace · Logistics
Geographies UK · Italy · Netherlands
Review cadence Monthly accuracy + override review
−74% Median cycle-time reduction
2.1× Median throughput uplift
97.4% Decisions accepted without edit
3 Engagements in production

Case 01 · Tender automation · Live since May 2025

Baldan Group started responding to
the tenders it used to walk away from.

ClientBaldan Group
SectorConstruction · 240-person contractor
SystemTender & RFP automation
Time to live5 weeks
GeographyVeneto, Italy

The situation

01 Four-person bid team was responding to one tender a week — with another three declined every week for lack of bandwidth.
02 Twelve years of prior proposals were scattered across SharePoint, local drives, and old laptops. Answers existed; finding them didn't.
03 Commercial director wanted more top-line. CFO wanted fewer weekend shifts. Both were right.

What Algebra runs

A Ingest agent: parses the PDF, extracts every requirement, builds the compliance matrix.
B Retrieval layer over the full proposal archive, ranked by recency and authority.
C Drafting agent writes each response in Baldan's house voice, cites the source, flags gaps.
D Review queue for the bid manager — accept, rewrite, or escalate each answer in one surface.
−82% Cycle time
3.4× Tenders responded to
+31% Win rate
94% First-pass coverage

"We went from responding to one tender a week to three, with the same bid team. The first one we won was a contract we wouldn't have had time to write six months ago."

Marco Baldan · Chief Commercial Officer · Baldan Group

Case 02 · Cash-to-order · Live since Aug 2025

Novus Aero closed the month three days
earlier — every month, without hiring.

ClientNovus Aero
SectorAerospace components · 480 staff
SystemCash-to-order reconciliation
Time to live7 weeks
GeographyBristol, UK

The situation

01 Finance ops team spent six working days every month matching bank feeds to invoices across three ERPs and two customer portals.
02 Exceptions queue grew faster than it could be cleared. Mis-applied cash was showing up in aged receivables that weren't actually aged.
03 Close was slipping past day 7. Month-end reporting was always one week stale.

What Algebra runs

A Ingest: pulls bank feeds, remittance PDFs, portal data, and open invoices into one ledger.
B Match: deterministic + probabilistic matching with confidence scores on every pairing.
C Exception triage: proposes resolution with evidence (the payer reference, the short-paid amount, the prior pattern).
D Finance analyst approves from a single queue. Everything else is auto-posted.
−71% Analyst time on matching
3.1 days Faster close
98.2% Auto-match accuracy
£0 Additional FTE required

"Our team used to spend the first week of every month just reconciling. Now they're closing books on day four and spending the rest of the time on analysis the board actually reads."

Priya Shankar · Finance Director · Novus Aero

Case 03 · Email triage · Live since Oct 2025

Meridian Log gave its agents back
three hours a day.

ClientMeridian Log
SectorFreight forwarding
SystemEmail & ticket triage
Time to live4 weeks
GeographyRotterdam, NL

The situation

01 Customer-ops inbox took ~1,800 emails a day — quotes, status checks, customs queries, disputes. All mixed together.
02 Agents spent the first 90 minutes of each shift triaging before answering anything.
03 SLA on urgent customs issues was slipping because they were buried behind routine status requests.

What Algebra runs

A Classifier: assigns every inbound to one of 14 categories with an intent confidence score.
B Enricher: pulls the shipment ID, order status, and customer tier from the TMS into the ticket.
C Drafter: writes a first-pass reply for the routine categories (status, tracking, POD).
D Escalation: urgent customs + disputes auto-route to the specialist queue with enriched context.
3.0 h Agent time saved per day
−68% First-response time
91% Drafts accepted as-is
99.1% SLA adherence on urgent

"The honest measure isn't the dashboard. It's that my agents stopped dreading Monday mornings. The inbox shape that used to wreck their week now just — isn't the shape anymore."

Ilse Kuipers · Head of Customer Ops · Meridian Log

Your name,
our next case.

We take on a small number of engagements each quarter. If there's a fit, we'd rather build than pitch.

Book a call → See the systems

Show us your workflows. We will show you where AI can run it.

Book a workflow review