The letter was marked “Prime Minister only.” It was dated 21 October 1941 and signed by four men from Hut 6 and Hut 8 at Bletchley Park, where Britain was breaking Germany’s Enigma ciphers. One of them was Alan Turing.
You might expect codebreakers to write about mathematics, or machines. They said they had plenty of machines. Their trouble was a shortage of clerks and typists. Without them, the work “is being held up, and in some cases is not being done at all.”
The details were small and exact. Hut 8’s Hollerith section had stopped working night shifts, so “the finding of the naval keys is being delayed at least twelve hours every day.” Hut 6 was receiving Middle East radio traffic it couldn’t fully decode for lack of trained typists. “Yet all that we need to put matters right is about twenty trained typists.”
And they were going over their managers’ heads to say so. “For months we have done everything that we possibly can through the normal channels,” they wrote, “and that we despair of any early improvement without your intervention.”
Delivering the goods
The giveaway is a phrase in the letter. The third shortage was the WRNS promised in July to test the bombes’ output; by late October nothing had been done. They pressed it less than the first two, “because it has not actually delayed us in delivering the goods.” That was their measure: they ranked the shortages by whether they held up delivery. The Hollerith night shift did. Without it, the naval keys were found at least twelve hours later every day, and the people waiting on what the messages said got it later too.
Effort is what you put in. Contribution is what the people relying on you get out.
Peter Drucker gave the idea its own chapter, “What Can I Contribute?”, in The Effective Executive in 1967. The effective executive, he wrote, “looks up from his work and outward toward goals.” He asks “What can I contribute” to the performance and results of the institution he serves. The question pulls attention “away from his own specialty, his own narrow skills, his own department” and toward “the performance of the whole.” Someone who focuses on effort and stresses his authority over the people below him, Drucker added, “is a subordinate no matter how exalted his title and rank.”
That’s why the question matters most for people whose work goes to someone else. A knowledge worker’s ideas and information, Drucker wrote elsewhere in the book, are useless in themselves until another person takes them up as input. The codebreakers’ output was decrypts. Their contribution came when those reached someone who could act on them, and the bottleneck they found was a shortage of clerks and typists, not of cryptanalysts.
The question has limits. It doesn’t license everyone to go around their managers. Inside Bletchley Park, The National Archives notes, the letter caused tensions when it came to light. The four took care to write that they were not criticizing Commander Travis, who ran the Enigma teams. And it helped that Churchill had visited in September and, as they put it, they believed he regarded the work as important.
Thirty tabs, one list
Imagine you’re an analyst who spends every Friday building a sales report with thirty tabs. It’s careful work, and it takes most of the day. One week you ask the sales director what she does with it on Monday. She opens one tab, the list of accounts whose orders slipped, and calls those customers.
So you move that list to the top and send it by Friday noon, with a line for each account saying why it slipped. The other tabs go into an appendix, once a month. Before cutting them for good, you ask who else opens them, because one might be the finance team’s. You worked less on Friday, and the director made her calls before lunch on Monday.
More code, slower delivery
The same gap shows up when software teams adopt AI. Google’s DORA research program surveys software professionals each year; its 2024 report drew on nearly 3,000 respondents. Its statistical models estimated that a 25 percent increase in AI adoption was associated with about 2.1 percent higher productivity, as individuals rated their own, and with better ratings of documentation and code quality.
The same models estimated that the same increase went with about 1.5 percent lower delivery throughput and about 7.2 percent lower delivery stability. “Contrary to our expectations,” the authors wrote. Their hypothesis was that AI lets people produce more code at once, so changes grow larger, and DORA’s earlier research had found that larger changes are slower and more prone to instability. The authors’ own heading says AI is hurting delivery performance. Still, these are associations in survey data, not a controlled experiment.
What the report does show is that individuals’ self-rated productivity and a measure of what the team delivers can move in opposite directions. If you want to know what an AI assistant contributes to your own work, count at the far end. For a few weeks, note each AI-assisted piece you hand on, whether the person who received it used it, and how much rework they asked for. Then put that beside how much you produced. A few weeks of notes are a lead, not proof, since the tasks you give the assistant may differ from the ones you keep.
Action this day
Churchill’s reply was dated the next day, a handwritten note to his chief military assistant with a red “Action This Day” label: “Make sure they have all they want on extreme priority and report to me that this has been done.”
On 18 November the report came back. The extra staff was arriving fast, though not all of it yet. And at Bletchley Park, the four signatories had picked up a nickname: the “wicked uncles.”