Customer assistants
Customer assistants that deliver measurable ROI measurable ROI
More than a decade of customer-facing conversational AI in production behind every assistant we deliver.
Trusted by enterprise teams in NZ and beyond
Most customer-facing AI doesn't make it past the demo
Your customers ask the same questions a thousand different ways, and the right answer depends on what's true today. Plenty of AI agents and chatbots demo this; fewer reach production, stay accurate, and sound like you. While the experiment stalls in pilot, the contact centre keeps burning energy on the repetitive questions.
What good looks like in customer-facing AI
The customer assistants that reach production and stay there share three traits. They are the things generic chatbots and AI agents can't fake.
-
Connected to your systems
Real-time integration with student records, CRM, support stack, or whatever your business runs on. Personal answers, not generic ones.
-
In your voice, not a vendor's
Trained on your content, your tone, and your edge cases. Holds the brand your team has built.
-
Updated by your team, not a developer
Content authors update the assistant directly as information changes. No release cycle, no support ticket.
A strong foundation built to support what comes next
-
The right architecture first
We start with the experience your customers need, the systems behind it, and the architecture that will support what comes next. The right foundation makes the next layer cheaper, not harder.
-
First layer in production
The opening use case goes live to real customers, measured against real KPIs like Coverage and Containment. One channel done right beats five half-launched.
-
Layers that compound
New channels, new queries, new integrations stack on the foundation without rebuilding it. Five years on, the system grows with the business, not against it.
Inside the University of Auckland's student assistant
-
70,000+
conversations served in year one
-
90%+
containment, answered without a human agent
-
5 years
live in production, zero major incidents
The University of Auckland's student assistant handles enrolment, exam, and study queries for 40,000+ students across five campuses. It integrates with student systems to answer personal questions ("when is my next exam?") and is kept current by the Student Contact Centre team, not a developer. ElementX has supported the deployment from initial scoping through more than five years of evolution, including the transition from earlier NLP into newer conversational architectures.
More than a decade of conversational AI in production
The University of Auckland is one Customer Assistant deployment, shown in depth. We've built similar customer-facing AI assistants for organisations in education, insurance, infrastructure, finance, retail, healthcare, and government for more than ten years. The discipline is the same throughout: AI assistants that reach production, that hold your brand and your business logic, and that your team owns after handover.
- Education
- Insurance
- Infrastructure
- Finance
- Retail
- Healthcare
- Government
Questions that come up before every deployment
Four questions every buyer works through before they commit. Short answers, in plain English.
-
What's the difference between a customer assistant, a chatbot, and an AI agent?
A chatbot follows scripted rules. A customer assistant uses conversational AI to handle the same question a thousand ways, integrated with your business logic. An AI agent autonomously executes multi-step tasks. Most customer-facing AI in production sits in the assistant category; where the work shifts to “act on my behalf,” that's Agentic Systems territory.
-
How long does a customer assistant take to reach production?
Six months is the typical PoC-to-production window for an enterprise customer assistant. UoA's student assistant went from scoping to live in around six months. The variable is your data and systems readiness, not the AI.
-
Can a customer assistant integrate with our existing systems?
Yes. That's where production AI lives or dies. UoA's assistant pulls live student records to answer “when is my next exam?” personally, scoped to what each student can see. Generic chatbots can't do that.
-
Who owns the customer assistant after deployment?
You do. Content authors on your team update the assistant directly as information changes. No developer required, no release cycle. That's how UoA's Student Contact Centre team keeps it current through every enrolment season.