Looker semantic layer
The need for precise and reliable data insights has never been greater in the AI era, when data powers intelligent applications and influences business choices. However, misinterpretations and errors may result from the complexity and clear volume of data, as well as the variety of tools and teams. Looker semantic layer-managed trustworthy definitions are necessary. Equipped with distinct business knowledge and standardized references, the semantic layer offers a consistent and business-friendly interpretation of your data, ensuring that your analytical and Artificial Intelligence projects are grounded in reality and capable of producing dependable results.
By serving as a single source of truth for business measurements and dimensions, Looker semantic layer makes sure that your company and its products are using clear, consistent terminology. By eliminating ambiguity through essential signals that translate to business language and user purpose, the semantic layer provides a basis for generative AI technologies to read business logic rather than just raw data, resulting in correct replies. LookML (Looker Modelling Language) facilitates the development of the semantic model that enables your company to specify the logic and structure of your data, abstract complexity, and provide users with an easy way to get the information they require.
In the case of gen AI, a Looker semantic layer is very crucial. Gen AI can provide stunning but fundamentally erroneous and inconsistent findings when applied directly to ungoverned data. Sometimes, especially when creating sophisticated SQL, it miscalculates crucial variables, organises data incorrectly, or misinterprets definitions. Missed income opportunities and a wrong approach may be the outcome.
Reliable business information is essential in every data-driven organisation. Looker semantic layer can cut data mistakes in gen AI natural language queries by up to two thirds, according to our own internal testing. A recent study by Enterprise Strategy Group found that the biggest problem for businesses’ analytics and business intelligence platforms was guaranteeing data consistency and quality. For the whole company and all linked apps, Looker offers a single source of truth, guaranteeing data correctness and supplying reliable business logic.
The foundation of trustworthy Gen AI
A strong Looker semantic layer, which serves as your company’s data intelligence engine and offers a centralized, controlled framework that outlines your key business ideas and contributes to the maintenance of a single, consistent source of truth, is necessary for gen AI to be genuinely trusted.
In order to fulfil the promise of reliable gen AI for BI, the Looker semantic layer is crucial, providing:
- Trust: Ground AI replies in controlled, uniformly specified data to lessen “hallucinations” caused by generative AI.
- Deep business context: Data and AI agents ought to be as familiar with your company as its analysts. By giving those agents knowledge of your KPIs, business language, and connections, you may enable them to correctly read customer enquiries and provide pertinent responses.
- Governance: Protect sensitive data and grant auditable access to it by enforcing your current data security and compliance regulations in the gen AI environment.
- Organizational alignment: Ensure that all users, reports, and AI-driven insights make use of the same concepts and terminology by implementing data consistency throughout your whole organisation.
The semantic layer advantage in the gen AI era
Designed for the cloud, LookML, Looker’s semantic modelling language, provides many essential features for completely incorporating modern AI in BI:
- Centralized definitions: To ensure consistent responses that get everyone on the same page, experts may design metrics, dimensions, and join connections once, and then reuse them across all Looker Agents, conversations, and users.
- Deterministic advanced calculations: Looker removes randomness and produces predictable, repeatable results, making it perfect for intricate mathematical or logistical processes. Furthermore, you may swiftly and easily complete difficult tasks by using dimensionalized measures feature, which aggregates variables so you can conduct operations on them together.
- Software engineering best practices: Looker makes sure that code updates are regularly tested and monitored using version control and continuous integration, which keeps production apps operating without hiccups.
- Time-based analysis: Time-based and duration-based computations are made possible by built-in dimension groups.
- Deeper data drills: By examining a single data point, drill fields enable users to investigate data in more depth. This functionality may be used by data agents to let consumers go further into various data slices.
With the support of a Looker semantic layer, an LLM can focus on its strengths, which include searching through well-defined business objects within LookML (e.g., Orders > Total Revenue), instead of writing SQL code for raw tables with confusing field names (order.sales_sku_price_US). Human-friendly descriptions and information (such as “The sum of transaction amounts or total sales price”) can be included in these items. This is crucial when business customers employ the business language, “show me revenue,” as opposed to the data language, “show me the sum of sales (price), not quantity.” LookML helps an LLM better discover the right fields, filters, and sorts data agents into intelligent ad-hoc analysts by bridging the gap between the data source and the decision-maker’s priorities.
With the help of LookML’s well-organised library catalogue, an AI agent can locate pertinent data and summaries to provide an accurate response to your query. The next step for Looker is to actually retrieve the information from the appropriate location.
Intelligent, reliable, and conversational insights are promised by the combination of AI and BI. Their customers may benefit from these improvements across all surfaces where people interact with their data with Looker semantic layer. To make data engagement as simple and effective as speaking with your most trusted business counsel, Google Cloud will keep adding features to conversational analytics, improving agent intelligence, and broadening support for a wide range of data sources.
Conversational Analytics
Gemini for Google Cloud powers Conversational Analytics, a chat-with-your-data tool. Conversational analytics enables people who are not familiar with business intelligence to ask questions about data in everyday, conversational language, moving beyond static dashboards. Looker (Google Cloud core), Looker (original) instances, and Looker Studio as part of a Looker Studio Pro subscription all offer conversational analytics.
A user may engage with Conversational Analytics in a natural, back-and-forth manner, as seen in the sample dialogue that follows. “Can you plot monthly sales of hot drinks versus smoothies for 2023, and highlight the top selling month for each type of drink?” is the query the user poses in this example. In response, Conversational Analytics creates a line graph showing 2023’s monthly sales of smoothies and hot beverages, emphasising July as the month with the greatest sales in both categories.
Conversational Analytics can understand natural language requests, such as multi-part questions that use common terms like “sales” and “hot drinks,” as demonstrated in this sample conversation. Users don’t need to define filter conditions or specify exact database field names, such as “Total monthly drink sales” or “type of beverage = hot.” Conversational Analytics offers a response that contains text and, where applicable, a chart, together with an explanation of its main conclusions and logic.
Key features
Key elements of conversational analytics include the following:
Use Conversational Analytics in Looker: In a Looker (original) or Looker (Google Cloud core) instance, you may use Conversational Analytics to ask natural language queries about your Looker Explore data.
Use Conversational Analytics in Looker Studio: To query data from supported data sources in natural language, use Looker Studio’s Conversational Analytics feature. need a membership to Looker Studio Pro.
Create and converse with data agents: By giving the AI-powered data querying agent context and instructions unique to your data, you can personalize it and enable Conversational Analytics provide more precise and contextually relevant replies.
Enable advanced analytics with the Code Interpreter: Your natural language queries are converted into Python code by Conversational Analytics’ Code Interpreter, which then runs the code. The Code Interpreter uses Python, which allows for more sophisticated analysis and visualisations than typical SQL-based searches.
Setup and requirements
The following prerequisites must be fulfilled in order to use Conversational Analytics in Looker Studio.
- It is required that you have a Looker Studio Pro membership. Looker users can obtain licenses for Looker Studio Pro for free.
- Gemini has to be enabled in Looker for Looker Studio by an administrator.
The following conditions must be met by both you and your Looker instance in order to use Conversational Analytics within it:
- Gemini must be enabled in Looker for the Looker instance by a Looker administrator.
- You must be given the Gemini role in the Looker instance by a Looker admin. Additionally, the access_data permission for the model you are querying must be contained in a role.
