As organizations use AI more to explore their data and get rapid insights, we’re helping clients pause before acting on the results. A confident answer can miss the context, definitions, or controls that make it trustworthy.
This month’s feature shares three questions to ask before you trust an AI insight.
We're also catching you up on Databricks. New capabilities keep coming, and lately the pace has been especially hard to keep up with. We’re looking at two important additions that make it easier to build more with AI on the platform.
And from dbt Summit, we’re sharing our take on a bigger industry shift toward open, flexible data stacks.
Below you'll find:
3 Questions to Ask Before Trusting Your AI Output
Tech Insights: Databricks Apps and Lakebase, our take on dbt Summit
Events: Databricks Data + AI World Tour (Dallas, Chicago, NYC), All Things Open (Raleigh)
3 Questions Before Trusting an AI Generated Insight
Across the industry, the AI conversation is shifting from building better models to figuring out what it takes to produce answers we can trust.
Before you act on an AI answer, ask these three questions:
#1 - Does it understand the business context?
Your AI needs the whole story.
Context is everything. Data platforms are adding ways to capture business meaning and govern how AI uses it, but teams still need to define that meaning and review the answers that follow.
#2 - Is it using the right definitions?
AI can't understand what your business hasn't defined.
“Revenue” might include refunds in one business unit and exclude them in another. This is where semantics matter: shared business definitions give AI a way to interpret terms as your organization uses them, rather than choosing whichever definition it finds first.
#3 - Can you trace and review the answer?
If you can’t check the work, don’t trust the confidence.
Check which data the agent used, whether it was permitted to use it, and how it arrived at the result. If you can’t examine the sources and logic, its confidence isn’t a reason to trust it.
Tech Updates
Databricks keeps adding new capabilities at a rapid pace, while dbt Summit offered a glimpse of where the broader data stack is headed.
Databricks introduces governed app building
Apps on Databricks let you build custom experiences around data and AI
A new beta brings together Genie App Builder, App Spaces, and Serverless Micro Apps. Users can build data and AI apps with natural language, while admins set the permissions and resources those apps can use. The apps also scale to zero when idle, reducing the cost of running apps that aren’t used continuously.
A8 Principal Consultant Ed Peason shared his first hand impresson of Apps:
"We’ve been building apps ourselves, and we’re seeing how quickly teams can deliver a customized experience for users to interact with data governed in Databricks."
Databricks Lakebase gives Apps a database
Lakebase brings a managed Postgres database into Databricks, giving those apps a place to handle transactional data
When asked which new Databricks capability excites him most, A8 Senior Consultant Michael Kollman answered,
"Lakebase, for sure. Connecting transactional data to the rest of the Databricks platform could be huge for building apps that bring analytics and AI into everyday workflows.”
Many applications need a place to save changing information, like customer updates or the status of a request. Lakebase gives Databricks Apps a database for that day-to-day activity. Because it’s part of Databricks, teams can use that information in their reporting and AI work without building as many connections between separate systems.
From dbt Summit: more flexibility in the data stack
Open Data Infrastructure was the big theme at dbt Summit.
The idea "open data infrastructure": define data pipelines and business context in dbt, then choose the tools and compute that fit each job.
Two product announcements that support this direction:
Lake Compute lets teams run selected dbt models on a different compute engine while others continue to run in their warehouse.
dbt Charts brings charts and dashboards into the dbt project, so teams can develop and review them alongside the data models behind them.
For us, it reinforces a familiar principle: build the stack around the business need, with room to choose the right technology as those needs evolve.
Events
Looking for a good excuse to step away from your day-to-day and learn something new? Here are a few events that we think are worth your time.
In this talk, A8 Managing Director of Data Strategy Christina Salmi will talk about how to align Apache Iceberg with broader data strategy and governance objectives.
