Trust & standards

How we train and certify our team.

The people who build your AI employees are trained against a written curriculum, tested with timed exams and hands-on builds, and held to a published design standard. This page is the program in full, so you can see the bar behind every conversation your customers have.

Certification levelsThree, plus an invitational tier
Curriculum24 chapters
Design standardTwelve principles
Hands-on proofPractical tests, minimum 3 team builds
Singapore's first respond.io Certified Strategy Partner

The certification ladder

Certification runs in three levels. Each is a timed, closed exam with an 80% bar to pass, and each level has to be earned before the next one opens. Level 2 adds a hands-on requirement: a real build, reviewed and approved, before the exam can even be taken.

Zelix Certified, Level 1 badge Level 1

Zelix Certified

WhatsApp Business Foundations

Timed exam, 80% to pass. The grounding every teammate needs before touching a customer conversation.

Zelix Advanced, Level 2 badge Level 2

Zelix Advanced

Advanced AI Agent Development

Practical build plus timed exam. A real agent build is submitted and approved first, then the timed exam. Requires Level 1.

Zelix Expert, Level 3 badge Level 3

Zelix Expert

Enterprise AI Agent Architect

Timed exam, 80% to pass, at a longer time limit for deeper material. Requires Level 2.

A fourth tier, Zelix Black Elite, is an honorary recognition reserved for master builders and is by invitation only. The three exam levels are the working ladder every builder progresses through in order.

What the training covers

The curriculum is 24 chapters, written and kept current against Meta and respond.io source material, and grouped onto the three levels. It moves from how the messaging platform actually works, to building and shipping AI employees, to the deeper platform engineering behind larger systems.

LVL 1
Foundations · 10 chapters

How the platform actually works

How WhatsApp Business works end to end: message types and the template rules, opt-outs, pricing and cost, quality ratings and rate limits, and the cadence that keeps a number healthy. The shared grounding every teammate is held to.

LVL 2
Professional · 10 chapters

Building and shipping AI employees

The build skills: connecting a number, modelling contacts and data, running click-to-WhatsApp ads with proper attribution, and designing, writing, and testing AI agents, including the design principles and the release gate every build passes before it goes live.

LVL 3
Expert · 4 chapters

The engineering behind larger systems

Platform depth: the advanced respond.io surface, integrations and automations, giving agents live data over HTTP requests, and full multi-channel growth builds that span more than one system.

Our full curriculum opens to the public shortly.

Exams and practical tests

A certification is earned, not attended. The mechanics are deliberately strict, because the point is to prove someone can do the work, not sit through it.

  1. Timed, randomised exams. Every attempt draws a fresh, randomised set of questions under a countdown, with an 80% bar to pass. No two sittings are the same paper.
  2. A hands-on practical for Level 2. Before the Advanced exam opens, the builder submits a real agent build. It is reviewed against our standards and has to be approved. A written score alone does not unlock it.
  3. A cooldown between attempts. There is a 24-hour wait between attempts, so a pass reflects understanding rather than repetition.
  4. Valid for one year. Certifications carry a one-year validity, so the credential reflects current knowledge as the platform and our standards evolve.

Who builds your agents

A certificate is the floor, not the finish. Before anyone on our team builds for a client, they have proven it on real work, under builders who have already shipped enterprise systems in production.

Level 2 certified, at minimum

Every team builder holds at least Zelix Advanced (Level 2), so the person building your agents has already passed both the exam and the hands-on practical.

Three builds of their own

New builders complete a minimum of three real agent builds themselves before they are trusted with a client account.

Reviewed by proven seniors

Those builds are reviewed by senior builders who have each launched at least three enterprise builds in production.

Shadowed by a Level 3 architect

Every team builder works alongside a Zelix Expert (Level 3), so there is always an enterprise architect on the build.

The standard every build follows

Training is only half of it. Every AI employee we ship is held to a written design standard and to clear commitments on how we handle your data, both published in full for you to read.

AI Agent Design Principles

Twelve principles across orchestration and trust and safety: agents that stay in one voice, answer from real records, disclose that they are AI, keep a human reachable, and confirm before anything counts.

Read the standard ›

Security & data handling

How we handle your data, access, and secrets: PDPA-aligned practice on platforms that carry recognised certifications, and a clear escalation path when something needs a person.

See our commitments ›

A trained team on your customer conversations

Every AI employee we build is designed, tested, and run by people held to this program and this standard. Tell us what your customers keep asking, and we will show you what a certified build looks like.

Build your team
This page describes the internal certification and training program we run for our delivery team. Certification levels, exam formats, and validity are the real mechanics of that program. Related reading: the design principles, security and data handling, and AI Employees.