Innovating with Google Cloud Artificial Intelligence · Topic 3 of 4
Choosing Google Cloud AI Solutions
How to choose a Google Cloud AI solution: speed, effort, differentiation and skills, Gemini Enterprise Agent Platform, Gemini models and the pre-trained APIs.
- 5 min
- Includes a diagram
- Consider speed, effort, differentiation, expertise and flexibility.
- Pre-built and pre-trained options are fastest; custom models give the most differentiation but need the most effort.
- Gemini Enterprise Agent Platform (previously Vertex AI) is the unified platform to access, customize, build and deploy models and agents.
What it is
Organizations can get AI value in several ways: use a pre-built AI application or agent, call a pre-trained API or foundation model, customize a model with their own data, or build their own. Choosing between them is a business decision. Google Cloud's platform for building with AI is Gemini Enterprise Agent Platform. The exam expects you to weigh the options and match common use cases to the right API or model.
How it works
Strategic considerations
| Consideration | Question to ask |
|---|---|
| Implementation speed | How quickly must the solution deliver value? |
| Development effort | How much building, testing and maintenance can the team take on? |
| Business differentiation | Is this capability unique to us, or common to every company? |
| Technical expertise | Do we have ML engineers and data scientists? |
| Choice and flexibility | Do we need to choose among models and change them later? |
The tradeoff runs along one line:
| Option | Speed | Effort | Differentiation | Skills needed |
|---|---|---|---|---|
| Pre-built agents and applications | Fastest | Lowest | Lowest | Little |
| Pre-trained APIs and foundation models | Fast | Low | Low to medium | Developers |
| Customize models with your data | Medium | Medium | Medium to high | Developers, some ML |
| Build custom models | Slowest | Highest | Highest | ML experts and data |
"Common task, no ML team, needs results quickly" points to a pre-built option or API. "Unique competitive advantage with proprietary data" points to customizing or building.
Gemini Enterprise Agent Platform
Gemini Enterprise Agent Platform (previously Vertex AI) is Google Cloud's unified AI development platform. Businesses use it to:
- Access models: Google's Gemini models and partner and open models through Model Garden.
- Prototype and customize: design prompts and tune models with their own data.
- Build and run agents: create agents that use tools and company data, and deploy them.
- Train and deploy custom models, including no-code training with AutoML.
- Manage and govern: monitor models and agents in production.
Typical business use cases: customer service agents, document processing, content generation, search over company data, demand forecasting and personalized recommendations.
Pre-trained APIs and foundation models
Pre-trained APIs are ready-made models called with a simple request. No training is needed.
| Offering | What it does | Example use case |
|---|---|---|
| Gemini (through the Agent Platform API) | Multimodal foundation model: understands and generates text, code, images, audio and video content | Summarize documents, draft replies, answer questions over content |
| Vision API | Analyzes images: labels, objects, text (OCR), faces and explicit content | Tag product photos, read text from images |
| Cloud Translation API | Translates text between languages | Translate product listings for new markets |
| Speech-to-Text API | Converts spoken audio to text | Transcribe and analyze customer calls |
| Text-to-Speech API | Converts text to natural-sounding speech | Voice responses in an app |
| Natural Language API | Analyzes text for sentiment, entities and categories | Measure sentiment in reviews |
A foundation model is a large model trained on broad data that can handle many tasks and be adapted. Gemini is Google's family of foundation models; it is multimodal, meaning it works across text, images, audio and video.
A worked scenario
A travel company wants three things: translate its website into ten languages next month, transcribe call center recordings, and build a booking assistant that can change reservations. With no ML team and a short deadline, it uses the Cloud Translation API and Speech-to-Text API for the first two, and builds the assistant as an agent on Gemini Enterprise Agent Platform using Gemini and its own booking API as a tool.
How the exam asks about this
- "Extract text and labels from images." Vision API.
- "Convert recorded calls into text." Speech-to-Text API.
- "Translate content into other languages." Cloud Translation API.
- "Generate summaries and answer questions across text and images." Gemini.
- "Fastest way to add AI with no ML expertise." A pre-built agent, application or pre-trained API.
- "One platform to access models, tune them and build agents." Gemini Enterprise Agent Platform.
What to remember for the exam
- Consider speed, effort, differentiation, expertise and flexibility.
- Pre-built and pre-trained options are fastest; custom models give the most differentiation but need the most effort.
- Gemini Enterprise Agent Platform (previously Vertex AI) is the unified platform to access, customize, build and deploy models and agents.
- Model Garden offers Google, partner and open models.
- Vision, Translation, Speech-to-Text and Natural Language are pre-trained APIs; Gemini is a multimodal foundation model.
Building a custom model for a common task such as translation wastes time and money. Use the pre-trained API.
Common mistakes
Choosing to build because it sounds more advanced. The exam rewards matching effort to the need.
Confusing Vision API with Speech-to-Text. Vision handles images; Speech-to-Text handles audio.
Forgetting that foundation models can be adapted. You can customize Gemini with your data rather than train from scratch.
Check yourself
Which API converts recorded customer calls into text?
Speech-to-Text API.
A company with no ML team needs AI results within weeks for a common task. Which approach fits?
A pre-built AI application or agent, or a pre-trained API.
What is Gemini Enterprise Agent Platform?
Google Cloud's unified platform to access models, customize them, build and deploy agents and custom models.
Topic 3 of 4 in Innovating with Google Cloud Artificial Intelligence
Test yourself on Innovating with Google Cloud Artificial Intelligence
The full Google Cloud Digital Leader question bank has 512 questions, with timed practice exams and an explanation for every answer.
This study guide is independent and is not affiliated with, endorsed by, or sponsored by Google Cloud. Vendor names, certification names and exam codes are trademarks of their respective owners and are used for identification only.