Ground Truth
A weekly show of AI's talking about trusted AI, data operations, and the market signals shaping how companies make AI actually work. This is made for our team at Entrecore to easily consume our research. Feel free to enjoy.
Episodes
16 episodes
Ground Truth EP 16: Before Anyone Asks
Our once-a-month episode where we talk about what Entrecore is building. Over the last five episodes, the judgment that should have governed a number kept getting lost: in one person's head, behind a vendor setting, in a dropped row, inside a spre...
Ground Truth EP 15: Nothing Wrong On The Report
For three weeks in 2022, Equifax sent lenders credit scores that were calculated wrong, from a legacy platform the company was already planning to retire. The credit reports themselves were never changed. Anyone who pulled their report that week w...
Ground Truth EP 14: Inside The Limits
In 2012 a formula in a JPMorgan spreadsheet subtracted one rate from another and then divided by the sum of the two instead of the average. It roughly halved the reported risk, so a dangerous position looked like it was sitting inside its limits. ...
Ground Truth EP 13: It Never Throws An Error
In autumn 2020, 15,841 positive COVID tests vanished from England's contact tracing system because a spreadsheet hit its row limit and dropped everything past it without a warning. The dashboard looked complete. This show has spent twelve episodes...
Ground Truth EP 12: They Can't Guarantee It
OpenAI wrote this about its own products: while unlikely, we cannot rule out that de-identified data derived from their usage helped improve our models. On September 11 the four hosts of All-In spent a good part of an hour and a half on exactly th...
Ground Truth EP 11: A Thousand Definitions For One Metric
The VP of Data at Shopify, running a 400-person team, went on a podcast this year and said semantic layer decisions have been her greatest failure as a leader, and that her company still has a thousand definitions for one core metric. Atlan's own ...
Ground Truth EP 10: They Bought A Store. The Problem Was A Process.
The most expensive version of the bet this whole category made: about eight billion dollars for a catalog, to give an AI agent product a governed data foundation. That product reached roughly a billion in recurring revenue, and a partner survey th...
Ground Truth EP 09: Six Kinds of Software, One Name
A vendor published a page this year admitting that searching for the name of its own product category returns six different kinds of software, and that the industry's converging vocabulary is confusing buyers. This week Maya and Theo trace three n...
Ground Truth EP 08: The Number Was Right. Nobody Used It.
An analytics practitioner wrote a line this week that reframes two years of dashboard complaints: nobody believed the numbers, so nobody acted on them, even when the numbers were right. This week Maya and Theo follow the trust thread from a data r...
Ground Truth EP 07: All Five Numbers Were Correct
Five systems, five different numbers for the same campaign, and a consultant who ruled out the obvious answer: all five may be correct, and that is the problem. This week Maya and Theo follow a confession from the firms companies pay to build the ...
Ground Truth EP 06: Decisions Everywhere, Owners Nowhere
95% of data leaders say they cannot trace an AI agent's decision end to end. 72% admit they have approved critical agent decisions without reading the rationale. This week Maya and Theo follow the ownership thread through a finance survey where CF...
Ground Truth EP 05: The Part We Usually Leave Out
Four episodes in, we have described the same hole from four different directions: everyone agrees AI has a context problem, nobody agrees where the fix should live, the business logic was never in a file, and a new tool does not change old behavio...
Ground Truth EP 04: New Tool, Same Behavior
A vendor sat in on 25 customer meetings in three days and wrote down what people actually said. One line stuck: "new tool, same behavior," from an insurance leader whose staff still type in the wrong number to dodge a broken workflow. This week Ma...
Ground Truth EP 03: The Business Logic Was Never In The File. It Was In Someone's Head.
A piece published this week gave the industry's favorite illusion a name: automation theater, impressive dashboards while the real work still happens in spreadsheets and email, because the business logic that actually decides a number "exists o...
Ground Truth EP 02: The Market Started Arguing With Itself
This week the enterprise AI market published two opposite fixes for the same problem, twelve days apart, in the same outlet. One said buy a governed context layer. The other said fix the data engineering. Three vendors shipped context machinery...
Ground Truth EP 01: Everybody Wants Context. Nobody Can Agree Where It Lives.
For two years the whole industry chased bigger models. This year the biggest names all landed on the same sentence: the models are fine, what is missing is context. Then they spent the year fighting over who owns the fix and renaming it every f...