Your data team is too slow (and accuracy won't save you)

The irony of my career in data: my love of getting to the bottom of things is also my biggest liability. I feel like I’ve really done my job when I have a complete, accurate answer. Sometimes it’s the right thing to do and there are many stakeholders who appreciate it, but there are a lot more who receive the buttoned up output of a data professional and think, “…that’s not really what I needed, and now it’s too late.”

This is a bad experience for all parties involved, but it’s much worse for the data professional.

When you don’t give someone the data they were envisioning at the speed they expected, there’s a silent cost. They start questioning your ability to quickly understand what they need, which starts them down the path of disengaging. The data professional lives or dies by stakeholder trust, and both accuracy and speed are required to build and maintain it.

Accuracy isn’t the only goal

Data teams have been on the quest for reliability for a long time, envisioning a beautiful future where consistent and accurate data is available to anyone, at any time, via a self-service interface.

The good news is we’ve made a lot of progress over the years! The problem is there’s been disproportionate focus on the accuracy piece, while our ability to quickly get data into the business’ hands has largely been static. It’s 2025, and the data profession is overwhelmingly still focused on delivery through dashboards, which often take time to turn around.

Fortunately, things are finally starting to change.

If you take an honest look at the behavior of our stakeholders, their revealed preference (as evidenced by all the QQs we data people receive in our DMs) is to ask questions about data in natural language.

While natural language is imprecise, it has an advantage over your standard issue BI tool: no one needs to be trained on how to use it. As long as you can reliably translate imprecise natural language into technical requirements that enable accurate responses, it really is the best interface for accessing data. Agentic analytics (like Hex’s Threads capabilities) is the an exciting glimpse of that possible future.

When done right, agents can give your data team new leverage in how you support the business.

Agents can:

• Answer more questions immediately and handle multiple simultaneously

• Keep your schedule clear by eliminating endless follow-up questions and meetings with stakeholders

• Let people start with vague, open-ended questions and iteratively refine their hypothesis and mental models

• Enable stakeholders to actually look at data and shape their understanding based on what they see

The iteration advantage: trading perfection for speed

The biggest blocker allowing the business to access data via natural language has been that the translators are usually us, human data people. The current AI boom is our ticket to the data delivery scalability we’ve always lacked. Speeding up this cycle is crucial for stakeholders to learn what it means to be data-informed (or even data-driven), and being able to enable this process at scale is a data team dream.

I recognize that might sound absurd to say! How can a non-deterministic text generator be the way to deliver reliable, accurate, always-on analytical support? It definitely won’t be out-of-the box. Data teams still have plenty to do to make sure the right people have the right data.

The difference is that now we can have an AI agent as our primary interface, which dramatically accelerates our speed of iteration and learning. Imperfect answers are much less of a problem when the cost of generating them (and throwing them out in favor of a correction) is so low. In this new world, accuracy becomes a matter of iteration rather than deliberation, and with the help of increased speed, you can reorient toward a more useful north star — trust.

For companies and data teams that are embracing agentic analytics, things are changing fast. It’s been just as incredible for us to see it happening inside of Hex as it has been to see it transforming data exploration for our customers. It feels like we’re on the cusp of a new golden era, where data teams can reliably scale our work and become the trusted partners our stakeholders have always wanted us to be.