A new home for our applied research

Brightbeam Applied Research: What makes an expert expert? A brewery interior alongside the headline and brightbeam.works address.

One of the most widely used tools in statistics was worked out by a Guinness brewer checking malt.

William Gosset needed to judge whether the malt was any good. But the statistical methods available to him assumed more samples than his work allowed.

Guinness sent him to study with Karl Pearson. And in 1908, Gosset published the method he needed, under the name ‘Student’. His work grew into the t-test, now used in everything from drug trials to A/B tests.

The question came from the work.

And your organisation will have its own brewer’s problems.

Your work isn’t quite like anyone else’s. Your experts have learned which exceptions matter, why things are done a certain way, and when the usual answer is wrong. Some of that is in your documents. A lot of it isn’t.

The AI labs can build models that know a great deal about your industry. But they don’t have that knowledge, so they can’t solve your specific problem. Someone still has to understand how your organisation actually works.

And now imagine 10,000 agents working alongside thousands of people.

How do you make what one expert knows useful across the business? How does what one team learns help the next team make a better decision, so that your organisation’s intelligence compounds into value?

And how do you govern all the decisions being made in your name? Who decides what an agent may learn, where it can act for itself, and when it needs to ask?

These are some of the questions behind our applied research.

And our new research site brings together our papers, open-source tools – CHAP, Metis and Brevet – and the questions we’re working through.

Explore our applied research at brightbeam.works. (opens in a new tab)

Related

More from Brightbeam

So tell us what you’re working on.

Get in touch