Services AI Audit Case Studies Blog Contact
Case Studies

What AI automation actually looks like in practice.

Five honest demonstrations of how Kanaky Tech turns repetitive business work into AI agents — from opportunity audits and proposal drafting to lead qualification, operations and internal knowledge. Plus the real systems we already run in production across the Pacific.

In short

Kanaky Tech is a Pacific AI automation and agent-development studio that builds AI agents to automate real business processes — audits, proposals, lead qualification, operations and internal knowledge. The examples below are framed honestly: some are internal frameworks and demonstrations we built and tested in-house, and others are anonymised representations of client work covered by NDA.

A note on honesty. Some engagements are covered by NDA, so client identities are withheld. Internal demonstrations are labelled as such. We do not publish invented client names, fake testimonials or star ratings. Where a figure is not the one verifiable proof point below, it is labelled illustrative, target or internal benchmark. The single hard number we stand behind: an AI visual-production pipeline took a 50-product campaign from roughly 25 hours to about 20 minutes. You can read more about how we work in our AI opportunity audit and our approach to business process automation.

Demonstrations
Internal framework · real

AI Opportunity Audit Framework

Problem
Most businesses know AI matters but have no idea where to start. They face a wall of tools and hype with no objective way to decide what is worth automating first.
Solution
A structured audit that maps a business's actual workflows and scores each one for automation opportunity — by time spent, repetition, error risk and feasibility.
Implementation
A four-stage method: discoveryworkflow mappingopportunity scoringroadmap. No code is shipped until the priorities are clear and agreed.
Results
A prioritised roadmap of ranked quick-wins — the highest-impact, lowest-effort automations first — so the business has a sequenced plan instead of guesswork.
ROI
Clarity and a sequenced plan before any spend. The audit itself is the deliverable: you leave knowing exactly what to build and in what order. See the AI opportunity audit.
Anonymised / internal demo

Proposal & Quote Automation

Problem
Teams spend hours writing repetitive proposals and quotes — the same structure, the same boilerplate, re-keyed for every prospect.
Solution
An AI agent that drafts a full proposal or quote from a short brief, pulling in pricing, scope and standard terms in the company's own voice.
Implementation
Templated generation with a mandatory human review checkpoint before anything is sent — the agent drafts, a person approves.
Results
Draft time cut from hours to minutes (illustrative), with consistent formatting and fewer copy-paste errors across proposals.
ROI
Faster turnaround and more quotes sent in the same working day — capacity is freed without adding headcount.
Anonymised / internal demo

Lead Qualification Agent

Problem
Slow, inconsistent lead follow-up. Inbound enquiries sit unanswered or get handled differently depending on who picks them up.
Solution
An agent that scores and routes inbound leads against clear rules and drafts a first reply so no enquiry goes cold.
Implementation
Intakescoring rulesCRM and notification integration, so qualified leads land in front of the right person immediately.
Results
Faster first-response times and consistent qualification, regardless of when the lead arrives (illustrative).
ROI
Fewer dropped leads. More of the pipeline that already exists actually gets worked.
Anonymised / internal demo

Operations Agent

Problem
Repetitive ops and admin work plus manual reporting — data re-entered by hand, status chased over email, reports rebuilt every week.
Solution
An agent that automates data entry, status updates and report generation, keeping systems in sync without manual intervention.
Implementation
API, N8N and MCP integration with existing tools, with human checkpoints on anything that changes state or goes to a client.
Results
Recurring manual hours removed from the weekly cycle (illustrative), with fewer transcription errors between systems.
ROI
Team time returned to higher-value work. Read more on business process automation.
Internal demo

Internal Productivity Agent

Problem
Knowledge scattered across drives, chats and inboxes — staff re-answer the same questions and new hires take weeks to find their footing.
Solution
A private internal-knowledge agent that answers questions over the company's own documents, on demand.
Implementation
A private, controlled environment with no training on your data — the agent retrieves over your documents rather than absorbing them into a public model. See private AI systems.
Results
Instant answers and less context-switching — staff stop interrupting each other for the same lookups (illustrative).
ROI
Faster onboarding and less time lost hunting for information that already exists.
Live, in production

Beyond demonstrations, these are real systems Kanaky Tech has shipped and that you can verify. They are why we can talk about AI agents and automation from experience rather than slideware.

01

mcp-datagouv-nc

The first Model Context Protocol server exposing New Caledonia's open-data portal to AI agents — it lets AI assistants query government open data directly. Real, shipped, and on our GitHub.

02

Kanak Languages Dictionary

A live AI dictionary for Kanak languages — a working product in the field, built to put Pacific languages and AI to work together.

03

Visual-production pipeline

The one hard number we stand behind: an AI visual-production pipeline took a 50-product campaign from ~25 hours of manual work to ~20 minutes. It powers the Keou AI platform (keou.systems).

We serve businesses across the Pacific, on the ground in both hubs — AI agents in New Caledonia and AI agents in Auckland. Pasifika AI is in development.

FAQ

They are a mix. Some are internal frameworks and demonstrations that Kanaky Tech has built and tested in-house, such as the AI Opportunity Audit framework and live systems like mcp-datagouv-nc and the Kanak Languages Dictionary. Others are anonymised representations of the kinds of automation we build, because several client engagements are covered by NDA and the client's identity cannot be disclosed.

Many engagements are covered by a non-disclosure agreement. Rather than invent names or fabricate testimonials, Kanaky Tech withholds client identities and labels those examples as anonymised. The live systems we link to — mcp-datagouv-nc, the Kanak Languages Dictionary and the Keou AI platform — are public and verifiable.

Where a client has agreed to act as a reference, Kanaky Tech can arrange an introduction during the engagement conversation. Some clients are under NDA and cannot be referenced. We prefer to demonstrate capability directly through a free AI opportunity audit and live working systems rather than rely on testimonials alone.

One proof point is real and verifiable: an AI visual-production pipeline took a 50-product campaign from about 25 hours of manual work to about 20 minutes. Any other figures on this page are clearly labelled as illustrative, target or internal benchmarks, and are not dollar ROI attributed to named clients.

Let's find your quick-wins.

Start with a free AI opportunity audit. We map your workflows, score the opportunities and hand you a sequenced plan — before you spend a thing.

Book a free AI audit Contact