Subscribe
Sign in
Datrics AI Analyst Builder — Your custom GenAI solution for analytics and reporting
AI Analyst Builder enables teams to create custom AI analysts without coding. These analysts answer data questions via a chat interface like ChatGPT. Tailored to specific business processes and data, it ensures accurate results and avoids hallucination.
55
Replies
@anton_vaisburd Huge congrats on your launch! Wishing you a journey filled with success and growth. What are your next steps post-launch?
@kjosephabraham thanks! We’re going to add more data connector like google ads soon + work with the interested existing and new companies to make sure that we are delivering the state of the art tool. In addition, more features are coming, like integration options, charts customization based on the preset style, more setting for flexibility etc.
Hey, this AI Analyst Builder looks super cool! Breaking down data silos and letting anyone chat with their business data is really impressive. But I'm curious - how do you ensure users don't accidentally input incorrect data or rules? What safeguards are in place to catch potential human errors before the AI produces faulty insights? That could have major consequences for important business decisions. Do you have any built-in validation or error-checking to help prevent those kinds of mistakes?
Hey, @p_val ! Happy that you love the product.
1. Governance Tools: We utilize typical governance tools such as access controls and versioning to prevent incorrect settings for the AI analyst.
2. Automatic Checks: Answers are automatically checked for accuracy. Critical logic is developed using deterministic scenarios, such as code or no-code pipelines that could be created on Datrics Platform. These approach prevents hallucinations.
3. Feedback Mechanism: Users can provide feedback on the answers given by the agent, with an automatic semi-assessment of this feedback.
4. Safety Guardrails: We enforce safety input and output guardrails to ensure reliable and secure operations.
Accuracy and hallucinations will be the real problem. Did you tackle it anyhow?
Congratulations on the launch!
@dmytro_krasun, you’re right, accuracy and hallucinations are real problems.
However, the main idea is not to rely entirely on generating answers using LLMs but to leverage “saved” logic.
For example, enterprises have their own methods for calculating business metrics, such as LTV or EBITDA.
These formulas can be saved as text scenarios and used by our Agent.
For more complex cases involving different data sources, this logic can be created using our no-code tool. It might involve getting values from Excel spreadsheets and databases, summing them up, and obtaining the result.
In such cases, the LLM will recognize which “saved logic” to use and will use it to answer questions. Because the logic is, in fact, a deterministic scenario, there won't be any hallucinations.
Hey Anton, this is super impressive! 🙌 I love how AI Analyst Builder lets teams dive deep into their data without any coding hassle. It’s really cool that you’re breaking down data silos and empowering business users to get insights directly. The practical applications for sales, finance, and marketing sound game-changing! Can't wait to see how businesses will utilize this. Also, offering a free trial is such a smart move. Excited to give it a spin!
Datrics AI Analyst Builder is a game-changer for data analytics teams! 🚀 The no-code approach to creating custom AI analysts is perfect for empowering non-technical team members to derive insights quickly and efficiently. The flexibility and ease of use make it a must-have tool for any organization looking to leverage AI for data-driven decision-making. Kudos to the team for this innovative solution!
Wow, I really like how easy it is to use and connect. Well done guys!
Question, in the future, would Datrics be able to create charts and graphical representations of the data and report? Datrics could then directly be used as an analytics plus reporting tool!
@faisal_ishaq1 Hello, Faisal; thank you for your question. Yes, Datrics can create graphical representations of data out of the box—just ask it! Automatic report generation is also coming soon.
Hey Anton,
I'm curious, how does the AI ensure accuracy and prevent hallucinations in responses?
Could you share any success stories from early adopters?
Congrats on the launch!
@kyrylosilin thx for the question!
I believe accuracy/trust is the main issue with a typical text-to-sql tools.
Off-the-shelf products just simply not good enough for the majority of business cases, as companies need to incorporate their own vocabulary and business logic (e.g. formulas).
Our solution for customization: give analysts/product teams the possibility to create their own processes - and then basically users function-call them from chat interface. We have a no-code data platform under the hood to make it happen!
For example, financial analyst creates a process and formula to calculate LTV of the customer - and when Financial Manager asks about LTV in particular country, AI Analyst triggers the process that explicitly created by analyst before.
Some of the examples of early adopters: finance teams (CFO copilot), HR tech companies to match vacancies and candidates, Revenue Operations / Sales (even we use it in a separate vertical product named Dealwave.ai), Product teams for B2B2B solutions to give conversational interface to their customers.
Totally see the value of such a product.
The demo needs a personal touch though. It was rushed and the AI voiceover was lifeless.
Congratulations, Anton! I'm really impressed with the way it breaks down data silos and empowers teams to get real-time answers without coding. All the best to your team
Hey @anton_vaisburd
Datrics AI Analyst Builder is a tool that helps you create smart AI helpers without writing any code. These AI helpers can answer your data questions just like you talk to a friend. They are super smart because they know your business inside and out, so they give you correct answers every time. No more guessing or wrong information!
Congrats on the launch! How do you ensure accuracy of responses? In my experience data is often difficult to analyze without a good understanding of the whole pipeline.
@jack_michalak hey Jack - that’s actually what we solve using our product. So in addition to intelligence layer on top of LLM (Agent logic) that decomposes the task, we give the users possibility to add their own deterministic flows. It could be done in a descriptive way - just telling what steps should be done for particular scenario, or in a way of a low-code interface (basically you easily create an API that is triggered from user interface)
@anton_vaisburd Ah, so you rely on people keeping additional information updated in your tool? I think that makes sense, it's just an additional step that's easy to forget.
Datrics - AI Data Analyst