On this page

In October 1958, an IBM engineer named Hans Peter Luhn published a six-page paper with a title that would outlive almost everything else written about computers that year: A Business Intelligence System. It had no charts, no dashboards and no databases in the sense we mean today. It was about something harder: getting the right piece of information to the right person, before they knew to ask for it.
IBM Journal of Research and Development · Vol. 2, No. 4 · October 1958 · pp. 314–319
A Business Intelligence System
H. P. Luhn
“the ability to apprehend the interrelationships of presented facts in such a way as to guide action towards a desired goal.”
01Who Luhn was
Luhn was born in Germany in 1896, trained in the textile trade, and moved to the United States, where he joined IBM in 1941. He was an inventor before he was a computer scientist, with dozens of patents to his name. Today his name survives mostly in the Luhn check digit: the simple sum that tells a payment form you mistyped your card number. But inside IBM in the 1950s, his work was on text: how a machine could read documents and decide what mattered in them.
Earlier in 1958 he had shown that a computer could pick the most telling sentences from an article by counting which words it used often and where they clustered: one of the first automatic summaries. The business intelligence paper took that idea and asked what an organisation could do with it.
02What he meant by intelligence
Luhn did not mean cleverness. He borrowed a dictionary definition and made it the paper's spine: intelligence is "the ability to apprehend the interrelationships of presented facts in such a way as to guide action towards a desired goal."
Three ideas sit inside that sentence. Facts on their own are not enough; what matters is how they relate. The point of seeing those relationships is action, not a report for its own sake. And the action has a goal, so the system has to know what each person is trying to do.
By "business" he meant any organisation at all: a company, a laboratory, a government office, an army. Anything with people who need to act on information.
03The system he drew
Luhn's design, in plain words, worked like this:
- Every new document is read by the machine and turned into a short abstract and a set of key words, so nobody has to read the whole thing to know what it is about.
- Every person, or every part of the organisation, has a profile of what they act on. Luhn called these the "action points".
- The machine matches the two. Each document goes to the people whose profile it fits, and only to them. Nobody gets everything; nobody misses what is theirs.
- The recipients' replies tune the system. When someone asks for the full document, or ignores the abstract, their profile shifts a little.
Luhn's system, 1958
- A new documentarrives: a report, a letter, a paper.
- The machine reads itand writes a short abstract and its key words.
- Matched to action pointsthe profiles of what each person or team acts on.
- Sent only where it fitsNobody gets everything; nobody misses what is theirs.
↺ The replies tune the profiles: asking for the full document, or ignoring the abstract, shifts them a little.
He called the matching step "selective dissemination". The name stuck in library science for decades. The rest of the paper is the plumbing of 1958: punched cards, magnetic tape, and printers that send the abstracts out.
04The part that still matters
Read the paper today and the striking thing is what it is not about. It is not about storing data, or querying it, or drawing it. Luhn took for granted that an organisation would have more information than anyone could read. His problem was attention: which of the many facts each person should see, today, so they can act.
The decades that followed built the half of business intelligence that Luhn skipped. Databases, then data warehouses, then dashboards made it cheap to store everything and draw anything. By 1989, when the analyst Howard Dresner proposed "business intelligence" as an umbrella term for fact-based support for decisions, the phrase had come to mean the reporting stack: the tools you use to go and look.
Who goes to whom
- 1958Luhn's paper: information comes to you, sorted and cut short
- 1970sDatabases: facts stored, to be queried
- 1989“Business intelligence” becomes the name for the reporting tools
- 1990sData warehouses: everything in one place
- 2000sDashboards: you go and look
- 2026Othor AI: a Morning Brief at 6 AM with what changed
Information comes to youYou go to the information
What got lost on the way was Luhn's direction of travel. In his system, information came to you, already sorted, already cut short, because the machine knew what you were trying to do. In most BI tools since, you go to the information, and you have to know what to ask.
05Where Othor AI picks up the thread
We did not set out to rebuild a 1958 paper, but reading it now, the overlap is plain.
- The abstract is the Morning Brief: every night Othor AI tests every number and every breakdown, and at 6 AM you get the few things that changed and why they matter, not the whole pile.
- The action points are the people a brief goes to and the question each finding ends with: one question, for a named person, about something they can act on.
- The interrelationships are what the checks look for: a number that moved because its mix shifted underneath it, a region that leads another, a change hidden inside a steady total.
Two things Luhn could not have had. His machine summarised documents; it never checked whether a claim in them was true. In Othor AI, every number in the brief is computed by code and can show how it was worked out, and the AI only writes the words around it. And his feedback loop needed people to keep asking for documents; ours starts with the owner keeping the business model true, so the facts are joined up correctly before anything is reported.
Sixty-eight years on, the hard part of business intelligence is still the one Luhn named first: not having the facts, but seeing how they relate, in time to act.