Frequently Asked Questions (FAQs)

No to both. AI direction is a business capability, not a technical one — it belongs to anyone who understands their own work deeply and has real work to point AI at.

The diagnostic that starts your journey asks nine questions about your work. Not one of them mentions AI. That is deliberate: what makes someone good at directing AI is judgment about their own domain, not familiarity with any tool. Foundation assumes no prior experience, and concepts are explained through your own work examples rather than jargon.

Whatever your current AI habit is — daily user or complete beginner — it is a starting point, not a requirement. An existing ChatGPT habit is not a shortcoming, and it is not a head start either. Extraction can be trained into direction.

What you do need is three things:

Basic digital fluency. No coding, no programming, no technical AI background.

Access to one suitable generative-AI tool. ChatGPT, Claude, Copilot, Gemini or another — a free plan is often enough. The program doesn’t mandate a provider or include commercial AI licences, and your organisation’s approved-tool rules always apply. Your coursework and submissions live in the CAIDA Academy, which runs in the browser with nothing to install.

Genuine work you are accountable for. Every assignment and capstone is a real deliverable from your own job. Employed, consulting, freelance, entrepreneurial and nonprofit work all qualify. A fictional case study does not — at least one capstone must show real workplace use, which is the one requirement that cannot be met by someone with nothing real to point the method at.

That last condition is the only genuine barrier, and it has nothing to do with technical skill.

Attendance earns nothing. The credential is granted only on demonstrated work, and a capstone can come back Not Yet Demonstrated — which is exactly why passing means something.

What gets assessed is your own work. Every domain ends in a capstone that is a real deliverable from your job: a live research question, a real dataset, a decision you own, a communication you must deliver, a workflow you actually run, or a tool your team needs. Nothing is simulated. And what’s examined isn’t only the artifact — it’s the direction trail behind it: what you delegated, how you framed it, what context you supplied, what you validated before relying on it. A polished artifact by itself proves nothing.

Every capstone is scored on the same eight abilities. Delegate, Frame, Context, Plan, Steer, Validate, Land, Capture — each rated independently on a four-level scale: 0 Not evidenced, 1 Developing, 2 Demonstrated, 3 Strong. All eight must reach at least Demonstrated. A weak ability is never averaged away against strong ones; a 0 or 1 anywhere triggers clarification, guidance or resubmission.

Three further conditions have to hold, and none of them can be satisfied by good writing: the domain’s real-use requirement is met, the evidence is authentic and verifiable, and no Responsible Direction red line is left unresolved. A capstone that crosses a red line cannot pass until the issue is corrected, whatever the scores elsewhere.

A human decides, always. Your capstone is pre-evaluated offline by your trainer, then confirmed in a mandatory live viva. AI may help summarise; it cannot pass or fail anything. Trainers complete CAIDA onboarding, work from common rubrics, and calibrate against shared anchor exemplars so the standard holds across batches. Every viva asks a question that is hard to prepare for: when should this result not be used?

And the credential needs more than two passed capstones. At least one must be in actual workplace use, and at least one reusable Skill must have been successfully run by someone other than you, from your written instructions, without you standing beside them. That last test is the sharpest one — it is where capability stops being personal and becomes transferable.

Two domains earn the credential. How many you actually take is a judgment about your role, not a target the program sets.

There are six, and they are peers. Research & Knowledge · Analytics · Problem Solving & Decision Making · Influence & High-Stakes Communication · Workflow & Productivity Systems · Business Applications. No core, no electives, no prerequisites, no fixed order. Take any two, in either order, or two in parallel.

Your free nine-question diagnostic recommends where to start. It asks about your work — not about AI — and your personal AI Direction Report identifies the two or three domains where directed AI is most likely to give you leverage first, explained in your own words and drawn from your own answers. Each recommendation is tiered: Best fit now, Best fit soon, or Possible fit. It is developmental guidance, not a competence measure, and the choice stays yours.

Why two, and not one. The eight Direction Abilities are the same in every domain. What differs is how the work goes wrong. Research fails as a plausible synthesis nobody traced. Analytics fails as a method that cannot support the claim made of it. Communication fails as fluency mistaken for argument. Workflow fails as waste automated rather than removed. Learning to validate a research brief does not teach you to notice when your own decision criteria were written after the choice. The ability is the same; the way of being wrong is not — which is why one domain demonstrates a competence and two demonstrate a discipline.

Why more than two, for some people. A consultant researches, analyses, and then has to persuade someone. An operations lead whose recurring work also needs a small shared tool spans two domains before starting. Each further domain is added to the same credential, under the same permanent ID, with earlier achievements intact. Some take two and add more as their role evolves; others commit to all six at enrolment as CAIDA Complete. Both are ordinary routes — more domains means wider, not higher.

They are not all the same length. Four run four weeks. Analytics runs five, opening with a primer week. Business Applications runs six, with two build labs. All six share the same four-week arc.

About three hours a week while a domain is running — one two-hour live session at the weekend, plus roughly an hour applied to work you were going to do anyway.

Per domain. One two-hour live online session per week. Four domains run four weeks and come to 12–14 hours in total. Analytics runs five, opening with a primer week, and comes to 14–16. Business Applications runs six, with two build labs, and comes to 18–20. All six share the same four-week arc of demonstrate, critique, stretch and confirm — only the weeks wrapped around it differ.

For the credential. Foundation plus two domains is roughly 26–36 hours depending which two, typically across two to three months. Taken in parallel, the fastest route is about four to five weeks for two four-week domains, five where Analytics is one of them, six where Business Applications is. Parallel study is permitted after Foundation, and you accept the combined workload of the two.

Foundation is one hour. A live online session at a scheduled time, normally one to two days before your first domain, plus about 25–45 minutes for its two-part assessment. Miss the session and a complete recording is available — but the recording replaces attendance only, never the assessment.

The between-session hour is the part worth understanding. It is not coursework. Every application task is a real deliverable from your own job, submitted as a structured Direction Record 48 hours before the next session. You are directing AI on work that was already on your desk — a brief you owed someone, an analysis that was due, a message you had to send. The hour is spent doing that work differently, not doing extra work. Your homework is your work.

The final week of each domain asks for a little more — one to three hours consolidating and preparing your capstone, which is due 72 hours before that session so your trainer can pre-evaluate it offline.

What it does not require. No career break. No technical retraining. No separate study material to find time for. Sessions run at the weekend precisely so that the working week absorbs the application rather than competing with it.

Yes, and the program is designed for it. Every capstone is real work from a real job, which means confidentiality was a design constraint from the start rather than an exception to be handled case by case.

You submit the minimum evidence needed for assessment — no more. What’s being assessed is the direction trail: how you framed the work, what context you supplied, what you validated, what you’d do differently. That trail is usually demonstrable without exposing the underlying material.

Confidential data may be redacted, masked, or replaced with a verifiable description. A figure can become “a variance of roughly this magnitude against a monthly baseline.” A client can become “a mid-size manufacturer in the same sector.” What matters for assessment is that the claim can be checked as a claim, not that the assessor can see the raw record.

Some things must never be uploaded at all. Raw sensitive datasets, credentials, and restricted client material. This is not a preference — unauthorised information use is one of the six Responsible Direction red lines, and a capstone that crosses one cannot pass until it is corrected, whatever the scores elsewhere.

Vivas never require confidential information to be shown to the batch. Where sensitive evidence needs examining, it moves to a private conversation with your assessor. The same applies to live critique sessions: records are anonymised, the trainer names the ability under discussion rather than the author, and workplace content is not displayed without your consent.

You keep ownership of your work throughout. Nothing identifiable is used in teaching or marketing without your explicit permission. Where an anonymised example appears in a demonstration, it appears because someone agreed to it.

A practical note on tools. The program doesn’t mandate an AI provider or supply licences — you work in a tool you already have access to, and your organisation’s approved-tool rules always apply. Deciding that a particular piece of work should not go near AI at all is itself an act of direction, and it is assessed as such.

If your work is sensitive enough that you’re unsure what would qualify, that conversation is worth having before you enrol rather than after.

You receive specific feedback naming the abilities or evidence that fell short, and one resubmission — included, normally within 14 days. It is a recoverable outcome, and it is a real one.

Not Yet Demonstrated means not yet. It is not a fail in the sense of a door closing. It names a gap: one or more of the eight abilities didn’t reach the standard, the domain’s real-use requirement wasn’t met, the evidence couldn’t be verified, or a Responsible Direction concern is unresolved. The feedback tells you which — not a score, but the specific thing that was missing.

There is a third outcome, and it is the most common near-miss. Pending Clarification means the work may well meet the standard but something needs establishing before that can be judged — usually evidence that wasn’t included or a step in your reasoning that wasn’t visible. It is often resolved in a short exchange rather than a resubmission.

The resubmission is included in your fee. Not an extra charge, not a repeat of the domain. You are strengthening the same piece of work, usually in a defined area, with the gap already named for you.

If you disagree with the judgement, there is a route. Your assessor first explains the rubric decision. If that doesn’t resolve it, a second calibrated assessor independently reviews the submission, the viva record and the rubric. Impartial review is a guarantee, not a favour.

Why the program is built this way. A capstone everyone passes proves nothing. If Not Yet Demonstrated were impossible, Demonstrated would be worth nothing to the person who earns it or the employer reading it. The standard has to be capable of not being met, or it isn’t a standard.

Most people who receive it are close. The trail is usually sound and one ability is thin — validation that checked the wrong thing, a result that never reached its real point of use, a Skill that nobody else could run. Those are fixable, and fixing them is where a good deal of the learning actually happens.

Three things, in descending order of how much they’ll matter to you: real deliverables already applied in your work, reusable Skills your colleagues can run, and a credential that records what you demonstrated.

Deliverables from your own job. Not exercises. Every capstone is real work taken to its real point of use, so what you finish with depends on which domains you chose:

• an evidence-backed research brief, with material claims traced to source
• a reproducible analysis and a decision-ready insight product, independently validated
• a decision record that survives scrutiny — options, criteria, adversarial challenge, ownership
• a high-stakes communication, delivered and reviewed by its intended recipient
• a redesigned recurring workflow, run in practice with a measured before-and-after
• a bounded working tool that at least one other person uses to complete a real task

You produce the two that match your domains, not all six. Because they’re real, your manager and stakeholders can see them while the program is still running.

Named, reusable Skills. Not saved prompts — structured capability packages: purpose and intended user, required inputs and context, the direction steps, decision rules and human approval points, validation checks, worked examples, an owner and a version. They’re tool-agnostic by definition, so the same package can live as a Claude Skill, a Custom GPT, a Copilot setup, or a documented procedure. At least one must have been successfully run by someone other than you, from your written instructions, without you standing beside them. That’s the test that turns personal capability into team capability.

The credential. The Certificate in AI Direction and Application, accredited by Exemplar Global, part of the ASQ family. You become a CAIDA Professional. It has lifetime validity, and every certificate carries a QR code to a live record showing which domains you demonstrated and when. One credential, progressively updated — further domains are added under the same permanent ID, and all six adds CAIDA Complete.

And the thing that outlasts all three. The eight Direction Abilities transfer across every role, every domain, and every tool generation. Tools churn quarterly. What you learn here doesn’t expire with the next model release.

What any of this is worth inside your organisation depends on the work you point it at.

The tools will change. The abilities won’t. That asymmetry is the reason the program is built the way it is.

Nothing here is tied to a tool. You work in a leading LLM — ChatGPT, Claude, Gemini or Copilot — but what’s taught and assessed sits above all of them: deciding what to delegate, framing the outcome, supplying the right context, choosing the route, steering, validating before relying, landing the result, capturing what worked. Those eight abilities describe how a person directs work. Nothing in them depends on a menu, a feature, or a model number.

A new model doesn’t change the questions. When a more capable model arrives, you still have to decide what to hand it, define what good means, supply authorised context, and earn the right to rely on the output. More capable tools tend to make direction more consequential, not less — a confident wrong answer produced faster is a bigger problem, not a smaller one.

Your reusable Skills survive the churn too. A Skill is a capability package, not a saved prompt — purpose, inputs, direction steps, decision rules, validation checks. It’s tool-agnostic by definition, so the same package can be re-implemented as a Claude Skill, a Custom GPT, a Copilot setup or a documented procedure without rewriting the thinking behind it.

The credential has lifetime validity, and there’s a live record alongside it. The certificate permanently records what you demonstrated and when. Separately, your online record shows current-practice status: Current, Update Available during the six-month window after a material update is released, or Practice Update Overdue if that window passes. When the standard materially changes, CAIDA releases a Practice Update and you have six months to complete it. Every certificate carries a QR code to that live record.

Your credential stays valid throughout. The status reports currency of practice, not validity — those are two different things, and conflating them is what makes most certifications quietly worthless three years on.

And guidance doesn’t stop at certification. A monthly open clinic stays open to you as an alumnus, and it’s where new tool capabilities get translated into direction practice as they arrive. The abilities are permanent; the vehicles keep improving.

Then you may want CAISA — the Certified AI Solution Architect. It is a peer credential, a different craft, not a higher rank. Some people take one, some take both, and the order between them is not fixed.

The distinction is not “does it involve AI.” CAIDA includes a domain where you build working tools — Business Applications — so the line isn’t drawn at whether you build. It’s drawn at what the thing you build has to survive.

Four conditions define the boundary. Work stays in CAIDA scope when all four hold:

• it is buildable and maintainable by one participant
• its users are named individuals, not an open population
• it is safe without a governance artefact — no formal sign-off, policy or control framework required before it runs
• manual fallback can be restored by the participant if it stops working

Fail any one of these and the work belongs in CAISA. A tracker three named colleagues use, which you could replace with a spreadsheet tomorrow, is CAIDA. A system your department depends on, which needs a data-protection review and which nobody can restore by hand, is CAISA — the same underlying technology, a different discipline of accountability.

The presence of AI inside the tool at runtime is expressly not the boundary. A tool that calls a model is not automatically architecture, and a tool that doesn’t is not automatically simple. Scale, ownership, governance and fallback are what decide it.

Which suggests where you’ll land. If your instinct on hearing “build” is a system other teams will depend on, with an owner, a review and an escalation path, that’s CAISA. If it’s a working tool that removes a weekly irritation for people you can name, that’s CAIDA — and Business Applications is where it lives.

Neither replaces the other. Every architect still directs AI daily — framing, validating, deciding what to hand over and what to keep. Direction is the universal capability between merely using AI and architecting it, and it doesn’t stop mattering because someone’s title changes. If you’re drawn to both, CAIDA is usually the more natural first step, because it is the discipline you will use on the days you are not architecting anything.