Fluent

Why do so many business intelligence tools fail to win over business users? It’s about trust. Natural Language Querying, instead of forcing every user to work like a data scientist, lets anyone ask questions in plain English, explore data and find answers. In short, a great way to build confidence and trust. But even the best NLQ solution rollouts can’t succeed without some groundwork. Here's how to bridge the adoption gap, grow trust in new NLQ tooling, and manage the process of change that comes with AI adoption. Link in comments ⬇️ #nlq #data #changemanagement

"Our data is a mess." This is the most common concern we hear from data teams trying to achieve self-service insights. "Rubbish in, Rubbish out" is the challenge that plagues many of them, and the introduction of AI to achieve self-service doesn’t change the quality of underlying data. Our team keeps track of the messiest data setups we see. Rarely are they that bad, and with a bit of wrangling, we're able to get Fluent working well for them. That said, now and then you have to admit defeat and file some data environments under ‘lost cause’…

Are traditional dashboards & reports just too reliable to beat? We’re digging into the effectiveness of emerging alternatives to traditional BI - and what needs must be met before business intelligence can evolve. ➡️ https://hubs.li/Q02WBzzT0 #bi #ai #dataculture

Is AI going to replace data teams? Not quite. There’s a lot of buzz (and many misconceptions) about AI’s potential role within data teams—and it’s easy to get caught up in the hype. In our latest post, we tackle four common myths about adopting AI:
 1️⃣ AI replaces data teams

 2️⃣ Implementing AI is plug-and-play
 3️⃣ AI runs itself 4️⃣ AI solves all data problems Time to put these myths to bed. ➡️ https://hubs.li/Q02W6XxL0 #data #ai #dataculture

LLMs aren’t infallible. When they encounter frequent hallucinations or inaccuracies, ‘trust issues’ can develop. They can stem from little more than one unsatisfactory interaction with a senior stakeholder. You wouldn’t expect a new junior analyst to instantly know every nuance of a business on their first day. It takes training to build context and familiarity. The same principle applies to an LLM. In this article, we explore how a data team can translate their most frequently used queries into a metrics layer that takes their LLM from 0 to 1 and eventually to its peak performance within an organisation. ➡️ https://hubs.li/Q02TXplX0 #data #metrics #ai

We speak to a lot of data people about how AI helps unpick the problem of generating company-wide, self-service data insight. One of their biggest worries is that their data simply isn’t mature enough for that kind of enhancement. Questioning your readiness for AI tooling is understandable, but a message to ‘data doubters’: Your organisation and your data are more ready than you think. Find out why ➡️ https://hubs.li/Q02SjWpq0 🎙️ P.S. If the challenge above sounds familiar, sign up for our free webinar 'Introducing AI to Your Data Stack' on October 28th! https://hubs.li/Q02SjX430 #datamaturuty #data #sql

We are live from Kensington Olympia for the next two days at Big Data London! Come say hello at Y746 and be sure to grab a logo sticker before they’re all gone… #bigdatalondon #bigdataldn #thefutureofinsight

Eliminating the burden of ad-hoc data queries is something text-to-SQL is great at. But, there's another benefit which I refer to as amplification. By channelling the knowledge, expertise and voice of the data team through an accessible, self-serve solution like Fluent, you can transform the role of the data team - and internal data culture of a business - for the better. #texttosql #datatculture #data

Ever since Yale released its first semantic parsing and text-to-SQL test, Spider 1.0, many LLM models have been trained and reviewed on it. The highest accuracy scores on Spider 1.0 are often publicised as proof of a model's efficacy in the real world. With the recent release of Spider 2.0 and a range of other text-to-SQL 'challenges' now available, is there really as much value in these tests as we think? And, if not, how do we actually gauge the effectiveness of a text-to-SQL model? ➡️ https://hubs.li/Q02PvV6c0 #texttosql #ai #llm

In the average business, only 10-20% of available data is structured. But it packs quite a punch. Many data teams are overworked, fielding ad hoc queries from business users. Many are exploring natural language querying solutions to lower the barrier to entry for data querying. It could effectively eliminate the time-suck that basic data questions create. But first, they have a decision to make: Should we build on our structured data warehouse or our larger, unstructured database? A SQL generation solution on structured data is key to true self-serve, business data insight. Here’s why.   https://hubs.li/Q02MGTlS0 #sql #structureddata #texttosql