Revolutionize the internal search engine with Google Agentspace
As a step towards revolutionizing the world with AI, Google has made another stride by launching Google Agentspace for Enterprise, which brings AI-powered search to enterprises and MNCs. This blog is entirely about the newly launched Google Agentspace.
This blog is a collaboration of Vahan Yeghiazaryan (Google Workspace Deployment Engineer) and Rohan Singh (Senior Cloud Infrastructure Engineer and Google Cloud Champion Innovator) at SADA, An Insight Company.
Vahan and I got the opportunity to explore and use Google Agentspace with early access, which is still ongoing, and we have planned to write a few technical blogs(but easily explained) about it. This one is the first installment. So let’s started
What is Google Agentspace???
We all are aware of how powerful and paramount Google Search is, but it is public. What if an enterprise has its own internal search engine or intranet search with the same power as Google Search with additional reinforcement of Gemini? That is something Google Agentspace can do. However, Google Agentspace is more than just an enterprise search engine. Rather than just a quality search, an agentspace acts as a central source of enterprise truth for your entire organization, with expert agents to automate your business functions, enterprise-personalized GenAI with multi-language support, and multi-modal AI assistance, ensuring the security and compliance of the organization.
Google Agentspace includes:
- Notebook LM Enterprise
- Google Agentspace Enterprise
- Google Agentspace Enterprise Plus
How does Google Agentspace benefit enterprises?
Just like you, we also had the same question: “How does having an intranet AI search engine help enterprises?” So for this, we conversed with people from different departments at our company, and based on that, here are the views. With Agentspace:
The Engineering Team can leverage it to easily locate old scope of works(SOWs), technical documentations, messages, and documents regarding clients and organizations they work for and make reports/audio(NotebookLM) summaries based on that. It can summarize and analyze extensive documentation, like TDDs, Confluence pages, and identify similarities across multiple clients. It can offer feedback (as an additional pair of eyes, not replacing human intelligence) on diagrams, answer related questions, and automate repetitive tasks. Act as a unified internal source of truth for engineers without accessing the internet.
By connecting AgentSpace to GitHub repos, engineers can easily ask questions about the repository and code, identify and resolve bugs, potential bottlenecks, and areas of optimization, scan repos to find any sensitive information, troubleshooting guides, quickly find relevant information, and write a ReadMe(even multiple ReadMEs).
The Solution Architecture Team can leverage support from diverse sources while drafting the SOW for clients, incorporating best practices, efficient solutions, design patterns, and technical specifications. They can easily locate and reference key insights from previously written SOWs. Additionally, they can be assisted in formulating pertinent questions for discovery calls based on available information, facilitating a deeper understanding, and consequently identifying deliverables. Can automate the generation of documentation (not full but at least something to start with or maybe partial document using company-wide template) — design documents, potentially architecture diagrams, and reports; analyze and fetch valuable insights from current and past project data, and incident reports.
The Support Engineers can easily search through their knowledge base articles and SOPs that would allow them to be more efficient and productive while handling the cases. They can search and analyze, as in this example, ServiceNow.
The People Operations Team can use Agentspace to fetch information on current employees, retrieve region-based policy information, check policy history and understand whether it is applicable or not, analyze the latest HR policies and regulatory changes. Check the applicants’ history(like whether he/she has applied before or not), get more value from applicant CV and streamline the onboarding process for new joiners.
The Project Managers Team can leverage Google Agentspace for tasks such as securely accessing customer information, quickly and qualitatively searching customer data, easily accessing project information, budget management like unexpected funds adjustment information, budget adjustments, etc, creating reports based on certain criteria like timelines, project type, funding type, budget, resource type and practice, extracting SOW value, and fetch JIRA boards(current and new project) summary, then even emailing those.
The Sales Team can understand margins and profits, gather information from both single and multiple client deal sheets, and quickly obtain key information from SOWs. It might be helpful in the transition of the accounts (which happens every year in sales) — gives you an opportunity to get an understanding of the customer portfolio, history of relationships, and overall strategy.
The Billing Operations Team can use Google Agentspace to assist with invoicing, client communication regarding invoices and checks, and cloud credit management. Agentspace can be particularly helpful in identifying which customers have been given cloud credits. This can be achieved by exporting SKU information to BigQuery and then using BigQuery as a data source for Agentspace.
App and Data Stores for Agentspace
For an enterprise, Agentspace opens up a wide range of possibilities for building various applications. You start by creating an app and then attaching data store(s) to the application. The app offers versatile accessibility, functioning both internally within an organization and externally, with availability across multiple platforms, including desktop and mobile devices. The app is accessible on various platforms, including websites and mobile applications; wherever it is integrated.
The app types with their use-cases are:
Notes: The company names are purely fictional.
Search & Assistant App
- Enterprise Search & Assistant (Preview): The Pearson Hardman law firm utilizes Agentspace to establish a secure, centralized repository of knowledge for sensitive legal documents and case strategies, allowing lawyers to conduct natural language searches to access confidential information, thus streamlining research and collaboration while upholding strict data security protocols.
- Search for your Website (Intranet): Gringotts Wizarding Bank leverages Agentspace to enhance its intranet search capabilities. Employees can easily find internal policies, procedures, and financial data using natural language, improving efficiency and knowledge sharing.
- Media Search: The Umbrella Corporation Broadcasting Network uses Agentspace to create a searchable database of internal training videos and product demonstrations. Employees can quickly find relevant materials for onboarding, skills development, and troubleshooting.
- Healthcare Search: Massive Dynamic pharmaceutical company implements Agentspace to provide researchers with secure access to confidential clinical trial data, research findings, and intellectual property. This accelerates drug discovery and development while maintaining data security.
- Search for Commerce (Internal Product Catalog): Buy More retail chain uses Agentspace to create an internal product catalog for its employees. Store managers and sales associates can quickly find product information, pricing, and inventory data, improving customer service and sales efficiency.
Custom Search App
- Custom Search Conversation Agents: An AI-powered chatbot for About The Fit’s HR department uses Agentspace to answer employee questions about benefits, policies, and internal procedures. This provides instant support and reduces the workload on HR staff.
Recommendations Apps
- General Purpose Recommendation Engine: Agentspace recommends relevant research papers, patents, and internal documents to scientists and engineers to support innovation and knowledge sharing at Stark Industries.
- Media Recommendation Engine: About The Fit corporate training department uses Agentspace to recommend relevant training videos and courses to employees based on their roles and skill gaps, improving employee development and performance.
- Retail Recommendation Engine (Internal Procurement): Agentspace analyzes data of Stark Industries manufacturing company to recommend suppliers and materials, helping the procurement team optimize costs and streamline the procurement process.
App creation and data ingestion depend on your data type:
- For 3rd-party data, you should use the Google Cloud console, not the API, to create an app and ingest data.
- For other data, you can use either the Google Cloud console or the API.
The data store is an entity that contains data for the apps that can be GCP native offering like Storage bucket, BigQuery, etc, Google Workspace services like Google Drive, Gmail, Calendar, and 3rd-party applications/tools like Slack, Box, JIRA, Adobe, etc.
Data could be structured(BigQuery, JSON, CSV, CRM, etc) or unstructured(HTML, PDF with embedded text, TXT, PPTX, DOCX, etc); Agentspace is smart enough to understand and work with both types of data.
As per the time of writing this blog — data from Confluence, Jira, Salesforce, Sharepoint Online, and Slack (third parties) is considered structured data.
When importing structured data using the Google Cloud console, Agentspace will automatically detect the schema. One can either use the auto-detected schema in the engine or provide a schema using the API. Should you provide a schema and later update it, the new schema must be backward-compatible with the original schema, or the schema update will fail.
Note: If you don’t provide a schema, the auto-detect feature can update your schema when you import new data by incorporating any newly detected fields. If any documents in your imported data contain new fields that are not backward-compatible with your original schema, those documents will fail to import.
Both apps and data stores complement each other and cannot function without each other. The relationship between an app and data source/store is one-to-many, meaning a single app can be connected to multiple data stores which let apps broaden the search parameter and provide better answers. This is called a blended search. Agentspace syncs with the third-party data source at a specified frequency and ingests data after a connection is established.
Agentspace creates a data connector and associates entity data stores to it for specified entities when a third-party data source is connected. The data source you connect to determines the types of entities. Jira Cloud, for instance, has entity types like issues, attachments, comments, and worklogs.
As per the time of writing this blog — Agentspace supports One-time ingestion and Periodic ingestion for GCP BigQuery and GCP Storage data stores(first-party or native data stores).
Agentspace Apps and Data Stores
Security and compliances
A lot of enterprises are concerned about working with AI or AI tools, and as engineers, we comprehend that very well and acknowledge that it is reasonable and justifiable.
Just like Google Cloud's other products, Agentspace is also built on Google Cloud’s secure-by-design infrastructure, which gives enterprises the confidence to deploy AI agents across their enterprises, ensuring data protection and compliance through granular IT controls, including role-based access control (RBAC), VPC Service Controls, and IAM integration.
In the case of Google Agentspace offering both Agentspace Enterprise and NotebookLM Enterprise components have different compliance certifications and security controls. Particularly, Agentspace Enterprise is HIPPA, ISO 27001, ISO 27017, ISO 27018, ISO 27701, and SOC 1, SOC 2, SOC 3 compliant.
Read more about the Google Workspace security whitepaper: How Google Workspace protects your data
Conclusion
Google Agentspace offers a powerful solution preferably for enterprises seeking to revolutionize their internal search and knowledge management. By leveraging AI-powered search, automation, and personalized GenAI, enterprises can enhance productivity, improve decision-making, and streamline various operational processes across different departments. With robust security and compliance measures built on Google Cloud’s infrastructure, Agentspace provides a secure and reliable platform for accessing and analyzing internal data, making it a valuable tool for modern enterprises looking to leverage the power of AI.
We have plans to write more technical and tutorial blogs about Google Agentspace, some in collaboration, whereas some will be solely by Vahan. Follow for updates and connect with Vahan over LinkedIn.
Connect with Vahan Yeghiazaryan on LinkedIn
Read my other cloud blogs:
Read industry professional interviews
