- How the Six Domains Are Organized
- Current Outline vs. the Legacy Five-Domain Version
- Domain 1: AI Problem Definition
- Domain 2: Data Collection and Transformation
- Domain 3: Feature Engineering
- Domain 4: AI Algorithms and Models
- Domain 5: Application Integrity and Deployment
- Domain 6: Maintaining and Monitoring AI in Production
- Exam Format, Scoring and Fee Mechanics
- Sequencing Your Preparation by Domain
- Who Benefits From This Credential
- Frequently Asked Questions
- The current English ITS-AI objectives list six domains, from AI Problem Definition through Maintaining and Monitoring AI in Production.
- Pearson publishes no percentage weights per domain, so treat all six as examinable and avoid trusting weight claims.
- The English content update took effect June 30, 2026; older five-domain outlines are legacy and incomplete.
- Passing requires a scaled score of 700 on a 1-1000 scale, not a raw 70 percent.
How the Six Domains Are Organized
The IT Specialist - Artificial Intelligence certification, issued by Pearson and delivered through Certiport (a Pearson VUE business), is built around six official exam objective domains. They follow the lifecycle of an AI solution: you define the problem, gather and shape the data, engineer features, select and evaluate models, integrate and deploy the application, and then keep it healthy in production. Once you see the domains as one pipeline, the exam stops looking like six unrelated topic lists.
The six domain names, exactly as they appear in the objectives, are:
- AI Problem Definition
- Data Collection and Transformation
- Feature Engineering
- AI Algorithms and Models
- Application Integrity and Deployment
- Maintaining and Monitoring AI in Production
One important framing point: these headings are the official exam objectives, not a preparation curriculum. Pearson tells you what a candidate is expected to know; it does not prescribe how to learn it. If you are still orienting yourself, our overview of what ITS-AI certification is covers the credential at a higher level, and the ITS-AI study guide turns these objectives into a preparation plan.
Current Outline vs. the Legacy Five-Domain Version
Candidates frequently get tripped up by outdated outlines. The revised objectives document (copyright 2026, five pages, ending at objective 6.5) is the current English scope. Certiport's exam content update archive lists the English update date as June 30, 2026, following an announcement on April 29, 2026. Rollout dates for other languages differ, so do not assume the English date applies to your language.
The conflict to be aware of: Pearson's general IT Specialist resources page still links a legacy copyright-2025 outline with five domains. That older document should not be treated as complete current English coverage. The revised outline separates topics that the older structure bundled together.
| Aspect | Legacy outline (copyright 2025) | Current outline (copyright 2026) |
|---|---|---|
| Number of domains | Five | Six |
| Feature engineering | Not a separate domain | Its own Domain 3 |
| Deployment domain name | Different framing | Application Integrity and Deployment |
| Production coverage | Limited | Dedicated Domain 6 covering drift and decommissioning |
| Status for English exam | Legacy; do not rely on it | Current scope |
Domain 1: AI Problem Definition
Domain 1: AI Problem Definition
This domain tests whether you can decide if, and how, AI should be applied before any data is touched. It rewards judgment over memorization.
- Translating a business need into a well-scoped AI problem
- Assessing feasibility: do you have the data, the signal, and the resources?
- Recognizing when a non-AI approach is the better answer
- Understanding human interaction with AI systems and where people stay in the loop
The revised objectives call out feasibility and human interaction explicitly, so expect scenario questions where the correct answer is "this problem is not suited to a machine learning approach" or "this system needs human oversight at a particular decision point." Study how problem types map to task families (classification, regression, clustering, generation) because that mapping drives everything downstream.
Domain 2: Data Collection and Transformation
Domain 2: Data Collection and Transformation
Data is the raw material, and this domain covers getting it, cleaning it, and shaping it into something a model can consume.
- Identifying data sources and collection methods appropriate to the problem
- Handling missing values, outliers, duplicates and inconsistent formats
- Transforming data: scaling, encoding, splitting into training and test sets
- Recognizing data quality and bias issues that undermine a model before training begins
Expect applied questions: given a described dataset problem, which transformation is appropriate? Be ready to reason about why data leakage happens and how a clean train/test split prevents it. This domain pairs naturally with Domain 3, and many candidates study them back to back.
Domain 3: Feature Engineering
Domain 3: Feature Engineering
Feature engineering is a standalone domain in the current outline, which signals that Pearson expects real depth here rather than a passing mention.
- Selecting informative features and discarding redundant or irrelevant ones
- Creating new features from existing data
- Representing text and other unstructured inputs as features, including natural language processing (NLP) preparation
- Reducing dimensionality while preserving useful signal
NLP appears in the revised coverage, so be comfortable with how raw text becomes model input: tokenization, vectorization concepts, and the trade-offs between simple and richer representations. The key exam skill is choosing the feature approach that fits the data type and the model you intend to use.
Key Takeaway
Domains 2 and 3 form the "data to features" bridge. If a scenario question describes a poorly performing model, ask first whether the cause is upstream in data quality or feature design before blaming the algorithm.
Domain 4: AI Algorithms and Models
Domain 4: AI Algorithms and Models
This is the model-selection and evaluation core of the exam, and it is where explainability and metrics show up.
- Choosing appropriate algorithm families for a stated task
- Training, tuning and validating models, and recognizing overfitting and underfitting
- Model-evaluation metrics: knowing which metric fits which problem and what each one actually tells you
- Explainability: why stakeholders need to understand model decisions and how that influences model choice
Metrics are a frequent source of avoidable errors. A candidate who knows the formulas but cannot say when accuracy is misleading (for example, on imbalanced classes) will lose points on scenario items. Practice reading a short description of a business risk and selecting the metric that reflects the cost of the errors that matter. If you want to calibrate how demanding this material is, see how hard the ITS-AI exam is.
Domain 5: Application Integrity and Deployment
Domain 5: Application Integrity and Deployment
The domain title reproduces the revised PDF's exact wording. It moves the model from experiment to a working, secure application.
- Application integration (objective 5.1): connecting a model to the systems and users that depend on it
- CI/CD concepts for automating build, test and release of AI-enabled applications
- Security considerations for AI applications and the data they handle
- Maintaining the integrity of what is deployed
This is the domain most likely to feel unfamiliar to candidates from a pure data science background, and conversely the one that software and operations people handle comfortably. Know what a deployment pipeline does, why automated testing matters for models as well as code, and which security risks are specific to systems that learn from data.
Domain 6: Maintaining and Monitoring AI in Production
Domain 6: Maintaining and Monitoring AI in Production
A deployed model is not finished. The last domain covers keeping it accurate, safe and eventually retired.
- Monitoring performance after release and defining what "healthy" looks like
- Detecting drift when real-world data diverges from training data
- Retraining and updating models responsibly
- Decommissioning: retiring a model or system in an orderly way
Drift and decommissioning are named explicitly in the current coverage, so both deserve dedicated attention. Questions often present a symptom (predictions getting worse over months) and ask you to identify the cause and the appropriate response. The document's final objective is 6.5, which confirms you have the full revised scope if your copy reaches that number.
Exam Format, Scoring and Fee Mechanics
The exam is a computer-based, proctored knowledge assessment. You can take it at a Pearson VUE test center or through OnVUE (online, with a system check and a live webcam proctor), or through the separate Certiport Authorized Testing Center (CATC) channel.
| Item | What the sources state |
|---|---|
| Questions | Estimated 45-50, a variable range rather than a fixed form count |
| Time limit | 50 minutes maximum of timed exam time, excluding administration and tutorials |
| Passing score | 700 on a 1-1000 scale (scaled, not a raw 70 percent) |
| Exam-only fee (MindhubPro) | USD 127 web/member pricing |
| CATC voucher with retake | USD 133, US-only; proctoring may cost extra; not usable at Pearson VUE centers or through OnVUE |
| Voucher expiry | 12 months after purchase, separate from credential validity; country-restricted |
| Credential validity | Five years for awards on or after 2025-06-18; earlier awards remain nonexpiring |
Two cautions. First, do not confuse the scaled 700 with a percentage; see the detailed explanation in our ITS-AI passing score guide. Second, Certiport does not publish pass rates, so any specific figure you encounter is unofficial; our pass rate article explains what can and cannot be said. For a full line-by-line fee view, including the voucher channel differences, read the ITS-AI certification cost breakdown.
On requirements, there is no degree prerequisite beyond knowledge of the objective domains. The program suggests roughly 150 hours of preparation as a readiness profile, not a verified application requirement. The full picture is in ITS-AI requirements, and scheduling specifics sit in ITS-AI exam dates. For CATC testing, bring photo and signature identification, expect an NDA, and note that candidates under 18 need parental consent and may not have personal paper or communication devices.
Sequencing Your Preparation by Domain
The domains are sequential in logic, so schedule them that way. Here is one ordering tied directly to the six domains, using the roughly 150-hour readiness profile as a planning anchor (adjust to your background):
Domains 1 and 2
- Practice scoping problems and judging feasibility
- Work through data cleaning and transformation scenarios
Domain 3
- Feature selection, creation and NLP representation
- Link each technique back to the data issues from Week 1
Domain 4
- Algorithm families, tuning, overfitting
- Metric selection and explainability drills
Domains 5 and 6
- Integration, CI/CD and security basics
- Drift, monitoring and decommissioning scenarios, then full mixed practice
Schedule Domain 4 in the middle because it depends on both preceding data domains, and leave Domains 5 and 6 together at the end because deployment and monitoring concepts reinforce each other. Finish with mixed-domain practice using the ITS-AI practice tests so you get used to questions that cross domain boundaries. For a condensed last-pass review, the ITS-AI cheat sheet is useful the night before.
Who Benefits From This Credential
ITS-AI is an entry-level knowledge assessment, so it fits students, career changers, IT support staff moving toward data work, and early-career developers who need a verifiable baseline in AI concepts. Employers use credentials like this as a screening signal alongside projects and experience rather than as a substitute for them. For role-level detail, see ITS-AI jobs, and for the value question, whether ITS-AI is worth it weighs cost against likely benefit. Earnings figures vary widely by region and role, so treat any single number skeptically; the salary guide discusses this qualitatively.
Key Takeaway
Because the six domains mirror a real project lifecycle, studying them in order while building a small end-to-end project (problem, data, features, model, deployment, monitoring) is the fastest way to turn objective names into exam-ready understanding. Test yourself regularly on the main practice site.
Frequently Asked Questions
The current English objectives list six domains, from AI Problem Definition through Maintaining and Monitoring AI in Production. An older copyright-2025 outline with five domains is still linked on a general Pearson resources page, but it is legacy and should not be used as complete current coverage.
No. Pearson does not publish percentage weights for these six domains. Any weighting you see on a third-party site, including practice-bank item counts, is not an official exam weight, so prepare across all six.
Certiport's exam content update archive lists the English update as June 30, 2026, following an April 29, 2026 announcement. Other languages roll out on different dates, so check the archive for yours.
The MindhubPro listing estimates 45-50 questions with a 50-minute maximum. That is a variable range rather than a fixed count, and the 50 minutes covers timed exam time only, not administration or tutorials.
A scaled score of 700 on a 1-1000 scale. It is not a raw 70 percent, and Certiport does not publish pass rates, so the number of questions you must answer correctly is not something you can calculate in advance.