Vertex AI Review – Unified AI Platform on Google Cloud

Vertex AI Review – Unified AI Platform on Google Cloud

In today’s fast-paced business environment, AI-powered search solutions are becoming essential for staying competitive and efficient. This blog includes a first-impression review of Vertex AI, Google’s unified AI platform for developers on Google Cloud.

Solutions such as generative answers and chat bots, which are powered by natural language processing, generate results that are tailored to the user’s intent and context, even if their query is not explicitly stated. These solutions enable organizations to streamline their operations, enhance customer service, and support informed decision-making, etc.

Many technical and business teams are asking themselves: how can I achieve this? In this post we’ll explore one of the many options to achieve it: leveraging Vertex AI Search & Conversation by Google.

Vertex AI

Vertex AI is a managed platform from Google Cloud to build and deploy machine learning models. Or as Google refers to it: “One AI Platform, every ML tool you need”. Vertex AI can be used to solve a wide range of machine learning problems, including predictive analysis, image classification, object detection, NLP, recommendation systems, and chatbot development

Since August 2023, Vertex AI Search and Conversation has been now publicly available. This article will explore the features and benefits of Vertex AI Search and Conversation in more detail.

Vertex AI Search and Conversation

Formerly known as Generative AI App Builder, Vector AI Search and Conversation is a suite of two products that enable developers to build intelligent applications that combine enterprise search and conversational AI capabilities. It was formerly known as

  • Vertex AI Search: A fast and accurate vector search solution for text, images, and audio.
  • Vertex AI Conversation: A platform for building natural-sounding and human-like chatbots and voicebots.

Vertex AI Search and Conversation can be used to build a variety of intelligent applications, such as:

  • Enterprise search portals: Search for information across the enterprise, including documents, code, and other data sources.
  • Customer service chatbots: Answer customer questions and resolve issues quickly and efficiently.
  • Virtual assistants: Provide employees with access to information and resources they need to do their jobs.
  • Product recommendation systems: Recommend products to users based on their past purchases and browsing behavior.

Vertex AI Search and Conversation is still under development, but it is already being used by businesses of all sizes to build innovative and intelligent applications.

No code/low code

One of its big selling points is that applications can be spun up without having to write code. At the time of writing, Google Cloud console allows to create three types of applications: Search, Chat and Recommendation. Creating an application has three steps: Select the app type, configure it, and select the data source. That’s it.

Review of Google Vertex AI - screenshot 1

Each application type offers configuration properties. For example, enabling Enterprise Search Edition ( which includes extractive answers, image search, website search), Advanced LLM features (such as search summarization and search with follow-ups), generic or personalized recommendations, etc.

Data can be obtained directly from various sources, such as websites, APIs, or Google storage options such as BigQuery and Cloud Storage. This is just the beginning though, Google has other interesting integration possibilities such as Jira, Salesforce and Confluence, but this are in preview mode and not generally available yet.

Demo: Vertex AI in Action

Using Pureinsights’ web site, we’ve created a chat agent to answer questions using only information found within www.pureinsights.com. The agent was configured using Google’s console:

Review of Google Vertex AI - screenshot 2

The website URL option was selected as the data source to feed the agent with information to answer questions:

Review of Google Vertex AI - screenshot 3

Once those two steps had been completed, and the website was fully indexed (this took a few minutes to complete), then the agent was ready to be tested using Dialogflow CX (chatbot platform to design agents using a state machine approach).

Review of Google Vertex AI - screenshot 4

Let’s ask some questions:

Does Pureinsights offer a processing platform?

Review of Google Vertex AI - screenshot 5

The response is generated using information sourced from https://pureinsights.com/technology/ . It’s important to note that this answer is not a direct replication of the content found on the mentioned page, but rather a synthesized summary based on the information available

Which sales rep can I contact?

Review of Google Vertex AI - screenshot 6

In the above query, rather than a long generative answer, a link to the ‘Contact Us’ page is provided when the question “Which sales rep can I contact?” is presented.

What is SearchOps?

When asking “What is SearchOps?” this is the answer, the platform links to https://pureinsights.com/blog/2021/why-you-need-enterprise-searchops/.

Review of Google Vertex AI - screenshot 7

What’s the difference between Google Bard and ChatGPT?

For this question, no direct answer is presented, however a relevant link that covers this topic is returned: https://pureinsights.com/blog/2023/what-is-chatgpt-ai-and-search-perspectives

Review of Google Vertex AI - screenshot 8

How many blog posts has Matt Willsmore authored?

For the question at hand, the chatbot doesn’t present a specific answer. There is no specific content that answers that question, even though the system could infer it. Fo this, Generators come into play. These allow to plug in large language models (LLMs) with custom prompts to allow more complex conversation flows.

Review of Google Vertex AI - screenshot 9

Conclusions from our Vertex AI Review

Vertex AI Search & Conversation provides interesting options for swiftly deploying conversational bots and search features from diverse sources. Achieving an out-of-the-box experience with acceptable results is straightforward.  It is worth noting that there are pre-built conversation flows useful for many uses cases different than the one presented in this article, for example: Credit card enrollment, credit card statement, account balance,  etc.

However, a trade-off becomes evident when balancing simplicity and customization. While Google offers a degree of configuration, it appears to lack advanced settings necessary for fine-tuning the search and chatbot for complex use cases that demand transparency and explainability.

I hope you found this first-impression demo and review of Vertex AI helpful. If you have any questions about deploying AI in search applications, or about deploying Google’s Vertex AI, please feel free to CONTACT US.

      – Ricardo

Related Resources


Stay up to date with our latest insights!