Anthropic, the company behind Claude, just shared how they use AI for their own data.

They say Claude now answers 95% of their internal data questions. And they say those answers are right about 95% of the time.

If you are an aspiring data analyst, your stomach probably just dropped.

But before you panic, take a deep breath.

I read their entire 4,500-word blog post. And what they actually proved is that human data analysts are more valuable right now than ever before.

Let me explain.

Why AI Sucks At Analytics

For years, companies have wanted 'self-serve analytics,' which means people can get data answers without always waiting on the data team. They want people in sales, marketing, product, and ops to ask questions and get quick data answers.

But it has always been hard to get right. People use the wrong tables, definitions get mixed up, and the dashboards are confusing.

So when AI arrived, everyone got excited. Anthropic tried throwing AI at the problem. They fed Claude all of their old SQL files and told it to learn.

It failed.

Claude only got the right answer 21% of the time.

The problem was never the AI's ability to write code. Claude can write SQL just fine.

The problem was context.

So, to fix this, Anthropic had to rely on pure, old-school data rules. They built a 4-step system to guide the AI.

Anthropic's 4-Step Playbook

1. Data foundations:

AI is smart, but it can't read minds.

Before Claude could answer data questions well, Anthropic had to explain what the data actually meant. They wrote clear notes for their tables, rows, and columns. They explained what each table meant, how often it changed, and when Claude should use it.

That is the kind of context a human analyst learns from meetings, Slack threads, old dashboards, and years of company knowledge.

2. One source of truth:

Then they had to make sure Claude knew which data to trust.

In big companies, teams often create different versions of the same table. One team makes a weekly sales table. Another team makes a filtered sales table. Someone else makes a copy for one project and never deletes it.

So Anthropic had to make the trusted source clear. Only one option.

3. Skills:

Then came the part that changed everything: skills.

Skills are reusable instructions that tell Claude how to do a task the right way. Not just "answer this question," but "use this table, follow this process, watch out for this mistake, and show the result this way."

That matters because code was not the real issue. It failed because it did not know the business rules. Humans had to write those rules down.

4. Validation:

Finally, they added checks.

Even if Claude has the right data and the right instructions, someone still has to check the work. So Anthropic built ways for Claude to show its sources, show the table it used, show how fresh the data was, and learn from corrections when it got something wrong.

Why You Aren't Going To Lose Your Job

You might look at those four steps and think, "Wow, this is really cool."

And you are 100% right. It is a very cool system. But it takes a lot of work to set up.

Setting up this system takes hundreds of hours. Keeping it updated is the harder part.

Anthropic admitted that when they stopped updating the AI's 'skills,' the accuracy dropped from 95% to 65% in just a few weeks.

The AI cannot run itself. It needs humans to organize the data, define the rules, write the skills, and check the answers.

Think about this: Anthropic has more money, talent, and AI power than most companies. They built the AI. Yet, they still have nearly 200 data people on staff, and they are actively hiring more data analysts right now.

AI isn't eliminating data jobs. It's changing them.

What You Need To Do Right Now

Just like the field moved from Excel to dashboards, we are now adding AI chatbots to the workflow.

This is not just a threat. It is also a big opportunity.

We are all new to AI, even Senior Data Analysts.

You can starting learning AI Analytics WHILE learning the fundamentals.

That's why I've reworked my bootcamp, The Accelerator, to not only teach you the fundamentals of Excel, SQL, and Tableau…but how to utilize AI in analytics.. Check it out here.

📺 LATEST EPISODE

Or if you enjoy the audio version, you can listen to it here!