Business

Conversational AI Analytics Platform

Deren Tavgac
March 30, 2024
5 min read

Traditional business intelligence platforms tout the ability for nontechnical stakeholders to “self-serve” without requiring help from their technical counterparts. In reality, most nontechnical stakeholders lack the technical knowledge and skills (e.g. SQL, Python, etc.) to manage their data analytics and reporting initiatives without involving their data or IT teams.

With modern advancements in AI, enterprises no longer need to struggle with data access and utilization. Imagine a world where anyone (CEO, CFO, CMO) can ask any data-driven business question (as they would to their data or analytics team), and get desired insights and reporting outputs in real-time. Platforms like Redbird enable just that - nontechnical users now have access to truly self-serve analytics.

What is AI-Powered Analytics?

Traditional business intelligence platforms require extensive setup and configuration to enable data pipelines, analytical workflows and reporting outputs for stakeholders. However, there is a new breed of business intelligence platform, the AI-powered analytics platform. Redbird has built this next gen BI platform from the ground up, with AI infused throughout. An AI-native business intelligence platform is significantly more efficient than alternatives. AI functionality drives connectivity, orchestration, deep automation and self-service through natural language across the data analytics and reporting lifecycle. This means more time for business users to run the business vs collecting data from clunky UIs, doing excel jockeying or building repetitive PowerPoint presentations.

The Evolution of Self-Serve Analytics

Visualization tools such as Tableau, Looker and Power BI have positioned their offering as a way to easily democratize analytics for the less technical business user. In reality, these tools operate within a specific component (dashboarding) of the overall analytics value chain. They also often assume that for stakeholders to self-serve the data must already be “clean” (i.e. within a clean SQL-accessible table). Additionally the UIs within these platforms involve a learning curve, which results in friction in the data democratization process.

But what would a truly self-serve analytics platform of the future look like? Enter Redbird, an AI-infused conversational BI platform that operates end-to-end across the full data analytics and reporting value chain. Users can utilize natural language to ask any business question, and Redbird automates the data collection, wrangling, analytics, data science and reporting steps without a single manual SQL query, workflow configuration or dashboard build.

By enabling less technical audiences with a ChatGPT-like interface for asking data analytics and reporting questions, Redbird customers are able to see >100x increases in data-driven decisionmaking across the organization.

Advancements in Generative AI / LLMs

How is this magical conversational BI experience possible? In recent years, there have been significant advancements in generative AI and LLMs, which enable real-time AI-driven chat communication through human-inputted prompts to answer any question. This empowers users to drive efficiencies in workflows through automated document summarization, natural language generation and other applications.

The Custom Enterprise LLM for Data Analytics

ChatGPT, Llama and other LLMs come pre-trained on years of historical data from across the web, but don’t come ready to answer business questions on top of an enterprise’s data. Customizing an LLM to the enterprise involves securely connecting into all of an organization’s data and using that data to fine-tune a custom instance of the LLM. In the absence of this connectivity, LLMs will not be contextually aware of the enterprise’s data and will be unable to answer questions that are posed. Importantly the data must be clean, which is a rarity in most organizations. Garbage in, garbage out, as data practitioners often say. To feed clean data into an LLM and orchestrate components, data engineering organizations must invest significant time and resources. This is where many tools like ChatGPT and Copilot fall short when enterprises begin experimenting with business applications for LLMs.

Another challenge with LLMs is that they’re notoriously bad at doing basic math, data analytics and reporting work. LLMs are great at predicting the best next word from text, but are stumped by basic data, analytics and reporting tasks and tend to hallucinate. In order to empower LLMs with analytical chops, you’d need to build an analytics translation layer that understands your enterprise’s data ontologies, business logic and analytical rules – this undertaking can be daunting for most organizations who lack the resources to invest in R&D to build such a solution (which can cost millions of dollars to develop).

Redbird’s AI-Powered Conversational Analytics Platform

Let’s dig into how Redbird solves these challenges as your AI-native, end-to-end conversational analytics platform.

  • Intelligent building blocks across the analytical lifecycle - Redbird’s analytics operating system consists of AI-powered building blocks throughout the data analytics lifecycle - across data collection (RPA, API-based integrations, web scraping for 3rd party data + data warehouse and data storage for 1st party data), data wrangling (cleaning, harmonization, advanced calculations), advanced analytics (data science, AI-powered optimization), and reporting (dashboarding, Powerpoint / Excel outputs). Most importantly these components are all interconnected through its orchestration engine which enables end to end workflows to be constructed and run with a few clicks, no coding required. These building blocks become crucial in Redbird’s ability to deliver robust generative outputs in response to any natural language question that is posed within the platform. For example, Redbird can produce a dashboard, PowerPoint, Excel Output, AI-powered data connector or data science model in response to natural language prompts.
  • Customizing LLMs to the enterprise’s data - Redbird comes with an admin layer that enables teams to load up their data ontologies (dataset descriptions, field definitions, relationships), business logic (calculation definitions for analytical processes) and reporting blueprints (in PowerPoint, Excel or dashboard formats). This is an important part of ensuring accuracy of conversational insights and ensuring that the central data team has governance control over data-driven answers that are being shared throughout the organization. AI-powered scanning technology also allows for these admin settings to be loaded up automagically as a first pass, with transparency into what the AI has generated and the ability for data team members to override any configurations.
  • Proprietary domain specific language for data analytics tasks - as noted, LLMs struggle out of the box doing basic math and data analytics tasks. Redbird has built its own proprietary domain specific language that enables seamless translation between natural language and data analytics tasks. This is accomplished through a proprietary advanced AI agent approach.
  • Continuous user feedback for personalized 1:1 BI - users can provide feedback to any prompt response, and this feedback helps drive personalization on future answers that the platform provides. 
  • Adoption across different user mediums - importantly, new users do not have to learn a new system as the Redbird solution also embeds into existing communication channels such as email, Slack, Microsoft Teams, etc.

Competitive differentiation

How does this differ from existing tools in the market? Here’s a quick breakdown:

ChatGPT - ChatGPT is trained on 3 years of historical data from across the web, and is not contextually aware of the enterprise’s data. Hooking ChatGPT up to an enterprise’s data requires an army of data engineers who are scrambling to ensure data pipelines are up to date. ChatGPT also out of the box cannot do data, analytics and reporting tasks, so will be unable to answer most analytics and reporting questions

Copilot - Copilot is a Microsoft product that works within the Microsoft stack but cannot be easily leveraged for non-Microsoft data sources or output formats. A marketer looking to leverage Copilot to analyze campaign performance data from various marketing SaaS platforms for example would not be able to. Copilot also can only produce basic recommendations on how to construct business workflows or more simple text- / image- based Microsoft suite generative outputs but struggles with more complex data-driven tasks.

Legacy BI dashboarding tools - Tools like Power BI, Tableau or Looker are beginning to release very simple generative AI capabilities, which cannot perform advanced analytics and reporting tasks, frequently hallucinate, and often fall short of customers’ expectations. 

Conclusion

In the fast-evolving landscape of analytics and business intelligence, traditional solutions have fallen short in their ability to deliver truly self-service analytics. Most organizations struggle with a hub and spoke divide between technical builders and nontechnical consumers. Most business stakeholders want more access to data, and don’t want to have to bother their technical team counterparts to answer every data-driven business question that emerges; however, they often lack the technical expertise to build data pipelines or construct dashboards. As a result, technical team members have to build a dashboard for most requests that arise (which comes at a cost), and dashboards often sit unused as they don’t exactly meet the specific business stakeholder need or require labor-intensive changes in response to follow on questions that emerge.

Advancements in AI have enabled Redbird to release the AI-powered conversational analytics platform, which solves for these pain points. Redbird helps teams produce insights and reporting outputs 100x faster via an intuitive conversational interface that anyone can learn.