What do you do when your business has mountains of data — images, documents, customer calls, sensor readings — but no data scientists to make sense of it? That's the exact problem the Oracle Cloud Infrastructure (OCI) AI services solve: they give you ready-made intelligence without requiring you to build or train any machine learning models yourself. For the 1Z0-1127 exam, you need to know which service does which job, so you can match a real-world use case to the correct tool.
Jump to a section
A simple way to picture Overview of OCI AI Services
Have you ever walked into a restaurant and been overwhelmed by a menu that lists every single ingredient, spice, and cooking technique used in every dish?
That's what raw data feels like for a computer. Now instead imagine a place where you don't need to know that the chef uses sous-vide or that the sauce has five secret herbs. You just point at a picture of a creamy mushroom risotto, and the dish appears perfectly cooked in front of you. That's what Oracle's AI services do for your data. Each AI service is like a specialist chef. One chef only works with images, another only with text, another only with spoken words. You don't need to understand how they season or stir. You just hand them your raw ingredient (your data), tell them what kind of dish you want (analyse this image, translate this speech, detect this anomaly), and they deliver the finished product (the insight). The restaurant is Oracle Cloud Infrastructure, and the menu is the suite of pre-built AI services. You pick the one that matches your need, and you never have to worry about building the kitchen from scratch.
Oracle Cloud Infrastructure (OCI) is a set of cloud computing services that runs on servers in data centres around the world. Think of it as a giant, secure computer you can rent over the internet. On top of that OCI platform, Oracle offers several pre-built AI services. Each service is like a specialised machine that performs one specific type of intelligent analysis. They are all 'managed' — meaning Oracle takes care of running and updating the underlying technology, so you just send your data and get results back.
The six AI services covered in exam objective 1.2 are:
AI Vision: This service analyses images and videos. You give it a picture, and it can tell you what objects are in it, whether there are faces, what text is visible, or even where things are located (like 'a dog is in the top-left corner'). Use cases include checking quality on a production line, moderating content on social media, or counting people in a store.
AI Language: This service understands written human language. It can extract key phrases ('the product was defective'), identify sentiment (positive, negative, neutral), classify text into categories, and even recognise named entities like people, places, and dates. A common use is analysing customer feedback or support tickets to spot recurring problems.
AI Speech: This service converts spoken language into written text (speech-to-text) and can also turn text into spoken words (text-to-speech). It is optimised for real-time conversations, making it ideal for transcribing customer service calls for quality monitoring, or enabling voice commands in an app.
Document Understanding: This service extracts information from scanned documents, PDFs, and images of documents. It goes beyond simple text extraction. It can identify key-value pairs (like 'Invoice Number: INV-123'), table data, and field labels. Businesses use it to automate invoice processing, insurance claims, and mortgage applications.
Anomaly Detection: This service looks at time-series data — data points collected over time, like temperature readings every minute from a machine. It learns what 'normal' looks like and then alerts you when something unusual happens. For example, it can detect when a machine's vibration pattern shifts, signalling an impending failure, or when online shopping traffic suddenly spikes (which could be a malicious attack).
AI Forecasting: This service predicts future values based on historical data. You give it past sales numbers, and it predicts next month's demand. You feed it weather data for the last decade, and it forecasts tomorrow's temperature. It uses statistical models that handle trends and seasonal patterns (like higher sales in December).
Why do these services exist? Before cloud AI, if a company wanted to automatically read text from images, they would need to hire AI experts, collect thousands of labelled images, train a custom model for weeks, and maintain it constantly. With OCI AI services, you just call an API (a standard request over the internet) and get results in seconds. It replaces the need for an entire team of data scientists for common tasks.
Each service can be used via a user interface (the OCI Console), command-line tools (CLI), or programmatically through software development kits (SDKs) in common languages like Python or Java. They integrate with other OCI services like Object Storage (to store files) and Functions (to trigger actions when results appear).
Identify the data type
First determine whether your input is an image, audio, text, document (PDF/scanned), or time-series numbers. Each OCI AI service accepts only specific input types. This is the first filter in selecting the right service.
Define the required output
What do you want to get back? If you need a description of objects in the image, use AI Vision. If you need key-value pairs from a form, use Document Understanding. If you need a forecast number, use AI Forecasting. Matching output to service is core to the exam.
Check if the data has a time dimension
If your data includes timestamps and you want to predict what comes next, use AI Forecasting. If you want to detect rare events that deviate from normal patterns, use Anomaly Detection. Static data (no time) cannot use either service.
Chain services if needed
Often a single use case requires multiple services. For example, an audio recording of a customer call: first use AI Speech to transcribe to text, then pass that text to AI Language for sentiment analysis. The exam asks about these integrations.
Choose between pre-trained and custom models
For common use cases (like detecting faces in images), use the pre-trained model. For domain-specific tasks (like identifying your unique product defects), use custom training available in AI Vision, AI Language, Anomaly Detection, and AI Forecasting. The exam tests this distinction.
Imagine a mid-sized e-commerce company called 'ShopFast' that sells hundreds of thousands of products. They receive thousands of customer return requests every week, each with a photo of the item and a written reason. Manually inspecting each image and reading each note costs them a fortune in employee time.
An IT professional at ShopFast would implement OCI AI services to automate this process. Here is a step-by-step breakdown of what they actually do:
Configure Object Storage: First, they set up a bucket in OCI Object Storage (a place to store files in the cloud). When a customer uploads a photo of their returned item, it automatically goes into this bucket.
Call AI Vision for Image Analysis: They write a small script (maybe in Python) that triggers every time a new image appears in the bucket. The script sends the image to the OCI AI Vision service. The service returns a list of detected objects ('shirt', 'stain', 'tear'). It also performs optical character recognition (OCR) to read any text on labels or tags.
Call AI Language for Text Analysis: The written return reason from the customer ('the shirt arrived with a yellow stain on the collar') gets sent to OCI AI Language. The service extracts key phrases ('yellow stain', 'collar') and classifies the sentiment as negative. It also identifies the entity 'collar' as a product part.
Combine Results: The IT pro writes a simple rule: if both the image analysis detects a 'stain' and the text mentions 'stain', then flag the return for manual review. If only one source mentions it, approve the refund automatically. This reduces manual checks by 80%.
Monitor with Anomaly Detection: They also send the daily count of 'stain' complaints into the OCI Anomaly Detection service. If the number suddenly spikes (say three times the normal rate on a single day), the system sends an alert to the quality team, possibly indicating a bad batch from a supplier.
Forecast Future Returns: Using historical return data, the IT professional sets up AI Forecasting to predict return volumes for the upcoming holiday season. This helps the warehouse plan for extra staff and inventory.
The IT professional does not write a single line of machine learning code. They just configure integrations, write simple business rules, and monitor the outputs. Their main job becomes connecting the right service to the right data and handling exceptions.
They also manage costs. Each API call costs a small amount, so they learn to batch requests and only analyse samples when appropriate. They set up alerts in OCI Monitoring so they get notified if usage spikes, preventing surprise bills.
The 1Z0-1127 exam tests your ability to recognise which AI service to use for a given business scenario. Traps are common. Here is exactly what you need to know.
Question types you will see:
Multiple-choice: 'Which OCI AI service would you use to extract text from a scanned invoice?' The answer is Document Understanding, not AI Vision. Both can read text, but Document Understanding is designed for structured documents.
Match-the-use-case: They give you a description like 'A bank wants to predict customer churn based on monthly account activity' and ask which service to use. The answer is AI Forecasting (with custom model training).
True/False: 'OCI AI Language can convert speech to text.' That is false — AI Speech does that.
Key concepts they love to test:
The difference between AI Vision (general image analysis, including object detection and face detection) and Document Understanding (specialised for text-heavy documents like forms and invoices).
The difference between Anomaly Detection (finds unusual events in time-series data) and AI Forecasting (predicts future values). Both work with time-series data, but Anomaly Detection spots outliers, while Forecasting predicts the next number.
That AI Speech does both speech-to-text *and* text-to-speech.
That AI Language covers sentiment analysis, entity recognition, key phrase extraction, and text classification — but not language translation. (Translation is a different service not on exam objective 1.2.)
The difference between pre-trained models (ready to use, no training needed) and custom models (you train with your own data for a specific domain).
Traps they set:
They give a use case that involves scanning a car licence plate from a video. An unwary candidate might choose AI Vision because it's video. But Document Understanding also does OCR. The correct answer is AI Vision (specifically the object detection module) because it is designed for real-time video streams and can detect multiple objects at once.
They present a scenario using sound (a recorded interview) and ask about 'AI Language' because it involves language. But if the input is audio, you need AI Speech first to transcribe it, then you could optionally feed the text to AI Language. The exam expects you to pick AI Speech.
Anomaly Detection: the exam sets traps by describing a scenario with static data (like a single list of values). Anomaly Detection requires *time-series* data — data with a timestamp and order. If the data has no time dimension, it cannot be used.
They ask about 'real-time' use cases. AI Speech handles real-time audio streams. AI Vision handles real-time video streams. AI Language processes static text and is not real-time for streaming.
What you must memorise:
The exact name of each service and its primary function.
That all six services are 'pretrained' but some (Anomaly Detection, Forecasting, AI Language) also support custom model training with your own data.
That the output of one service can be the input of another (e.g., AI Speech output feeds into AI Language). This is a common exam integration question.
The supported input types: images (JPEG, PNG) for Vision; audio (WAV, MP3, FLAC) for Speech; plain text or URLs for Language; PDF and image files for Document Understanding; time-series numbers in CSV format for Anomaly Detection and Forecasting.
OCI AI services are pre-built machine learning models you can call via an API without any training — just send data and get results back.
AI Vision analyses general images and videos; Document Understanding specialises in extracting structured data from forms and invoices.
AI Language performs sentiment analysis, entity extraction, and key phrase detection on written text — it does NOT do translation.
AI Speech converts audio to text and text to audio but does not analyse meaning — you chain it with AI Language for full conversation insights.
Anomaly Detection identifies unusual events in time-series data, while AI Forecasting predicts future values — both require timestamped sequential data.
The exam frequently tests your ability to match the correct service to a real-world use case and distinguish between similar services like Vision vs Document Understanding.
These come up on the exam all the time. Here's how to tell them apart.
AI Vision
Analyses general images and videos
Detects objects, faces, scenes, and reads text anywhere
Does not understand document structure like forms or tables
Document Understanding
Specialised for scanned documents, PDFs, forms
Extracts key-value pairs (e.g., Invoice Number) and table data
Understands the layout: which text belongs to which field
Anomaly Detection
Finds unusual data points that deviate from normal patterns
Requires time-series data but output is a flag (anomaly or not)
Used for real-time monitoring and alerting
AI Forecasting
Predicts future numerical values based on historical trends
Requires time-series data but output is a predicted number
Used for planning and budgeting
AI Speech (input audio)
Accepts audio files (WAV, MP3, FLAC)
Output is transcribed text or spoken audio
Does not analyse meaning or sentiment
AI Language (input text)
Accepts text strings or URLs to text
Output includes sentiment, entities, key phrases, classification
Does not handle audio input
Mistake
All OCI AI services require you to train your own machine learning models before you can use them.
Correct
All six services come with pre-trained models that work out of the box. You can use them immediately without any training. Custom training is optional for some services.
Beginners confuse cloud AI services with AI tools like TensorFlow, where you always have to build a model from scratch. Managed AI services hide that complexity.
Mistake
AI Vision and Document Understanding are the same service, just with different names.
Correct
AI Vision is for general image analysis (objects, scenes, faces, OCR for any text). Document Understanding is specialised for structured documents (invoices, forms, tables) and extracts key-value pairs and table data.
Both services read text from images, so people assume they are redundant. But Document Understanding understands *structure* (where fields are on a form), while Vision just reads text wherever it appears.
Mistake
AI Language can translate text from one language to another.
Correct
AI Language does not include translation. It only analyses text in its original language (sentiment, entities, key phrases). Translation is a separate service in OCI (not covered in objective 1.2).
The name 'Language' sounds like it should handle all language tasks, and many competing AI services (like Google Cloud Translation) bundle both. Oracle keeps them separate, so the exam tests this boundary.
Mistake
Anomaly Detection can find any unusual value, even in a single list of numbers without a time order.
Correct
Anomaly Detection works strictly on time-series data. It learns patterns over time (e.g., rising/falling trends, seasonal spikes). A standalone list of numbers with no timestamp cannot be used.
People think 'anomaly' just means 'outlier', which is a static concept. But OCI's service is built for time-based data, like sensor readings or server metrics.
Reveal each answer, then mark whether you got it right. Score 60%+ to unlock the next chapter.
No. You can use all six services through the OCI Console web interface without writing any code. However, the exam expects you to understand that they are typically called via REST APIs, and SDKs exist for Python, Java, and other languages.
AI Vision is for general image and video analysis: object detection, face detection, scene classification, and OCR of any text in an image. Document Understanding is specifically designed for structured documents like invoices and forms, and it extracts key-value pairs and table data.
AI Speech only converts audio to text (or text to audio). To analyse sentiment, you must first transcribe the audio with AI Speech, then feed the resulting text to AI Language, which performs sentiment analysis. They are two separate calls.
Use AI Forecasting. It takes historical time-series data (like daily sales numbers) and predicts future values. Anomaly Detection would tell you if today's sales are unusually high or low, not what future sales will be.
Use Document Understanding. It can extract text (including handwriting) from scanned documents and also identifies the structure of forms, such as which text belongs to which label. AI Vision could read the text, but it would not understand the layout.
Yes, you send your data (images, audio, text) to Oracle's servers for processing. The services are cloud-based. Oracle does not store your data longer than needed for processing, unless you configure storage. Security and data residency are important exam topics.
You've finished Overview of OCI AI Services. Continue through the 1Z0-1127 study guide to build a complete picture of the exam.
Done with this chapter?