How we structure our briefings and why
How Iris is different from regular AI updates?
Currently, AI out of the box is not effective at producing briefings. Try it - every week, the briefing will have the same sources and overall message.
Furthermore, finding the right information is only half the job. The other half, which is easy to underestimate, is getting it into their head in the few seconds they're actually willing to give up. This is the gap that Iris fills - efficient, scheduled information that you can ingest at your own convenience.
What we looked at
Compressing information for a reader who won't read for long is an old problem, and it's been solved, partially, by a few different disciplines for different reasons.
Newsrooms solved it for readers with no time and shrinking attention: Axios's Smart Brevity format forces a headline under six words, a single lede sentence, and a "why it matters" line before anything else is allowed onto the page. The military solved it for readers who need to act quickly with BLUF (Bottom Line Up Front). Government agencies solved it for readers who can't be assumed to be experts: the US Plain Writing Act pushed federal agencies toward short sentences, concrete nouns and active voice.
But none of these were built for what we're doing and none of them were usable off the shelf. But together they pointed at the same underlying shape: say the one true important thing first, in the fewest possible words, then let the reader choose how much further to go.
What we tested
We ran a small internal comparison, taking the same set of source articles and generating briefings under three conditions: an unconstrained summary (roughly how most AI tools write today), a Smart-Brevity-only structure, and a combined BLUF plus Smart Brevity structure similar to what now sits inside Iris.
Time to 'comprehend to a point of being able to accurately recall the information' per method (internal test, n=10)
| Method | Avg. words | Time to comprehend (sec) |
|---|---|---|
| Unconstrained AI summary | 142 | 34.2 |
| Smart Brevity only | 71 | 21.8 |
| BLUF + Smart Brevity (Iris) | 58 | 16.4 |
Why the combination, not just one
Smart Brevity is very good at forcing brevity, but on its own it can still bury the actual point inside a well-written paragraph. BLUF is very good at forcing the point to the front, but on its own it can read like a memo. Put together, BLUF decides what goes first, Smart Brevity decides how it's said. That's the rule Iris's writing agent now runs on for every briefing it generates.
Where this leaves us
This isn't a solved problem, it's a moving target. As we get more real reading data from Iris subscribers, we'll keep tightening the format, squeeze out more efficiency and give you some extra time back.
Iris is being developed in London, UK.
This article was written by a human.