STACK AI Lab Book a Demo
Case studies

Three deployments, three layers of the stack replaced.

Each ran on the same terms: nothing upfront, a two-week pilot, settlement against measured results. Client names are published as consent lands — the processes, systems and scope below are the real record.

01 — Acquisition
AI Acquisition Agent
B2B services · the outbound tool stack cancelled outright
02 — Revenue
CRM Agent
Distribution · off a per-seat CRM and onto a database they own
03 — Operations
Operations / ERP Agent
Manufacturing · quote-to-invoice running without re-keying
Case 01

AI Acquisition Agent

A B2B services firm with nine people in sales, running outbound across four stitched-together subscriptions, stuck in the low double digits of meetings a month.

Timeline
DiagnosticDay 0
First sequences liveDay 11
Full volumeDay 26
First settlementQuarter 1
The problem

Purchased prospect data nobody cleaned, sequence templates untouched for three years, replies read a day late. Half the team's time went into moving records between tools, and the actual conversations were the thing getting squeezed.

What we built
A target profile derived from who had actually bought, matched continuously against public data.
A first message written per account, with one-click rep approval before anything sends.
Reply triage and same-minute answers, with anything commercial routed to the owning rep.
Meetings booked straight into the calendar with a briefing attached.
Software cancelled
Prospect data subscription Sequencing tool Inbox warm-up tool CRM engagement add-on
Case 02

CRM Agent

A distribution business, twenty-odd CRM seats paid for three years running, while the sales team still tracked deals in spreadsheets and the forecast was rebuilt by hand every month.

Timeline
Migration and de-dupWeek 1
One team on the pilotWeek 2
Whole team switchedDay 30
Old seats not renewedRenewal date
The problem

The CRM was always behind, because updating it was extra work for the people it measured. Management's reports didn't match reality, so nobody trusted them, so nobody updated the CRM — the loop had been running for years.

What we built
Email and calls logged against the right deal without anyone touching a form.
Stages advanced from real activity, with stalled deals surfacing themselves.
A forecast the system produces on a fixed definition, traceable down to the deal.
All of it on a database the client owns, with no per-seat meter running.
Software cancelled
Per-seat CRM licences Forecasting add-on Outsourced CRM admin
Case 03

Operations / ERP Agent

A manufacturer seven years into its ERP, still re-keying between quote, order and invoice, and still closing the month on overtime.

Timeline
Document parsing pilotWeek 2
Quote to order connectedWeek 3
Reconciliation and invoicingDay 30
ERP kept as system of recordOngoing
The problem

The ERP wasn't the problem — the gap between it and reality was, and people were filling that gap. PDFs read by hand, exceptions chased by hand, and every process change queued behind a change request measured in quarters.

What we built
Supplier and customer documents parsed and written into the ERP as fields, not attachments.
Three-way matching handled automatically, with only the exceptions escalated.
Stock and purchasing rules the operations team can change themselves, same week.
The month-end pack produced on its own, without a BI seat in the chain.
Software cancelled
ERP add-on seats Power BI viewer seats Document-capture subscription Connector automation plan
On the numbers

We don't publish numbers we can't stand behind.

The processes, systems and delivery timelines above are the real record. The revenue and savings figures go public only once the client has signed off on both the definition and the disclosure — the same standard we invoice against: measurable, traceable, agreed by both sides.

If you need the full figures before deciding, we'll walk you through a complete review of the closest case under NDA.

Which one of these is yours?

Bring one process and one set of software invoices. Ninety minutes later you have a written pilot scope with a baseline and a definition of success.