Trigger ConsultingTRIGGERCONSULTING

Work / 14 pilots shipped, 4 detailed here

Real automations.
Real hours back.

Four AI automation case studies from real small and medium businesses. Client names stay private. The numbers are real, the systems are still running. Full stories available on request.

Portfolio numbers01 / Logistics02 / Professional services03 / Retail04 / PropertyPatterns we seeWhat didn't scale

The portfolio in four numbers

14

pilots shipped

since 2024

280+

hours saved

per week, across clients

2.4 wk

average pilot

from kickoff to live

11/14

scaled past pilot

the rest are below

Selected case studies

Four pilots, four shapes of business

Each one started as a single painful job. Each one has been running in production for at least three months.

01 / Logistics

Invoice processing cut from 3 days to 4 hours

40-person freight & warehousing firm, Gauteng

The setup

Supplier invoices were arriving in ops@ by email and PDF, captured by hand into Xero, then matched against POs in a separate spreadsheet. Two people, three days a week.

The pain

Roughly 140 invoices a week. Late payments, missed early-payment discounts, and an accounts team that hated Mondays.

What we built

An automated intake pipeline: email is parsed, the invoice extracted, fields validated, PO matched, and the entry posted into Xero. Anything ambiguous (vendor mismatch, total off by more than 1%, missing PO) lands in a review queue with the original PDF and the parsed fields side by side.

Tools

Azure Document Intelligencen8nXeroSlack

Timeline

2 weeks pilot, 1 week stabilisation

Results

~8 hrsof admin saved per week
  • 94%straight-through (no human touch)
  • R 41kin recovered early-payment discounts, quarter 1
“We used to talk about invoices in days. We don't talk about them any more.”
Finance manager

What we'd do next

Auto-recon the supplier statements against what's been captured, so month-end stops being a project.

Request the full story →

Before: 3 days, hand-captured. After: 4 hours, mostly automatic.

02 / Professional services

Half of routine client queries answered without staff

Mid-size legal practice, Western Cape

The setup

Paralegals were spending hours each day answering the same questions: status of my matter, when's the engagement letter coming, what's the partner's diary look like. Real work was getting pushed to evenings.

The pain

Client satisfaction scores said replies were slow, the team said they were drowning. Both were right.

What we built

A custom assistant trained on the firm's own matter management and document templates. It answers routine status questions, sends out engagement letters from approved templates, and books partner intro calls. Anything that needs legal judgement is handed to a paralegal with the matter context already pulled.

Tools

OpenAIAnthropicMicrosoft TeamsIntercom

Timeline

3 weeks pilot, 2 weeks training rollout

Results

55%of queries handled automatically
  • 11smedian reply time, auto-handled
  • +14 NPSin the first quarter
“The clients who used to wait two days for an email now get a reply before they put the phone down.”
Senior partner

What we'd do next

Add document drafting (NDAs, mandate letters) from approved templates, partner-reviewed.

Request the full story →

Routine on the left, escalations on the right. The team owns the right.

03 / Retail

Weekly reporting went from half a day to automatic

Multi-store specialty retailer, Johannesburg + Cape Town

The setup

Every Monday someone exported sales from Shopify, stock from Lightspeed, cash from Xero, stitched it in Excel, sent a PDF by 11am. Sometimes 2pm.

The pain

Half a day, every week, on a report nobody read until Tuesday. And by then half the numbers were already stale.

What we built

A pipeline that pulls each system at 05:30, reconciles them, generates the report, flags anomalies (stock variances, return spikes, margin moves), and emails the deck before anyone logs in. The PDF still ships for the people who want it; the dashboard updates live for the people who don't.

Tools

n8nShopifyLightspeedXeroSlack

Timeline

2 weeks pilot

Results

0 hrsof manual reporting left
  • 06:00delivery, every Monday, no exceptions
  • 2 issues caughtin the first month a person would have missed
“I get the email before I get my coffee. That was never a thing before.”
Owner
Request the full story →

Three systems in, one Monday email out. Anomalies surfaced, not buried.

04 / Property & lettings

Tenant queries on WhatsApp answered before bedtime

Sectional title agency, 1,800 units under management

The setup

After-hours queries piled up in a shared WhatsApp number: levy questions, statement requests, maintenance reports. Replies came the next day, sometimes the day after. The good queries got lost in the noise.

The pain

Tenants frustrated, trustees frustrated, and the maintenance team only finding out about a leak when the geyser had already cost three ceilings.

What we built

An assistant on WhatsApp that handles routine queries (levy, statements, payment confirmations) instantly from the property ledger, and escalates anything mechanical, financial or legal to a person with the full thread already attached. Maintenance reports auto-create work orders and notify the right contractor.

Tools

WhatsApp BusinessOpenAISageWorkorder

Timeline

3 weeks pilot

Results

<1 minmedian reply, 24/7
  • 63%of after-hours queries closed without a person
  • 0 missedmaintenance escalations in 6 weeks
“The trustees stopped asking why we never reply. They started asking how we reply so fast.”
Managing agent

What we'd do next

Move levy collections and arrears reminders onto the same channel.

Request the full story →

Routine answered instantly. Anything mechanical or financial routes to a person.

Patterns we see

The same three shapes show up everywhere

The verticals change. The structure of the win doesn't. If your business has any of these, there's a pilot in it.

01

Inbox to system of record

Email, attachment, or form arrives. Extract, check, file in the place it actually belongs. Invoices, applications, orders, leads.

Examples

Logistics invoices · Sales leads · Loan applications

02

First-line conversational triage

Someone asks a question your team has already answered a hundred times. The assistant answers it. The interesting questions get a person with full context.

Examples

Client matter queries · Tenant queries · Customer support

03

Cross-system stitching

Your reports live in three places. Your customer profile lives in five. We pipe them together so the answer is one query away, refreshed automatically.

Examples

Weekly retail report · Customer 360 · Margin tracking

The honest part

Three pilots that didn't earn their keep

Eleven of fourteen pilots scaled past the first phase. Three didn't. Here's why, and what we learned. We'd rather show you these than pretend everything works.

  1. 01

    AI-generated proposal drafts (consulting firm)

    The drafts were technically fine. The partner's voice carried the deal, not the structure. We shipped a proposal template instead. No retainer.

  2. 02

    Inventory forecasting (small retailer)

    Too few SKUs and too much seasonality to beat the owner's gut. The model would have needed two years of clean POS data. We told them to come back in eighteen months.

  3. 03

    Voice-to-CRM after sales calls (B2B)

    Reps wouldn't use it. The friction of confirming the auto-summary was more than typing five lines. We killed it after two weeks.

Trigger Consulting

Let's find your first automation

A 30-minute call. We'll talk through where your time goes and whether AI can help. If it can't, we'll tell you that too.

Prefer email? hello@triggerconsulting.co.za