Jesse Jacoby
Every company has a rule like this one. Purchase orders above a certain amount need a second signature from a regional finance lead, even though the regions were merged four years ago and the person who signs now sits in a different function. Nobody likes the step. Nobody can say what it was protecting against. And nobody will remove it, because the one thing worse than a pointless control is removing a control that turns out to have had a point.
That’s what organizational forgetting looks like from the inside. It happens slowly, as a growing pile of decisions, rules, workarounds and system settings whose reasons have left the building, until the organization can neither defend them nor safely change them.
The usual response treats this as a documentation chore, something to fix with a wiki push or a new knowledge tool. The argument here goes further. Memory is infrastructure. It has to be designed, owned, funded and maintained the way a network or a facility is, and the part most organizations fail to keep is the reasoning behind their decisions rather than the decisions themselves.
The Cost Shows Up as Time
The everyday price is easy to underestimate because it’s spread thin across every team. Atlassian’s State of Teams research, based on a survey of 12,000 knowledge workers and 200 executives, found that leaders and teams spend 25 percent of their time just searching for answers.1 Flip and Workplace Intelligence surveyed 1,500 frontline managers and employees in the United States, the United Kingdom and Germany in early 2025, and 72 percent of managers in retail and manufacturing said they doubt their companies can hold on to the expertise leaving with retiring workers.2
Tenure is working against memory too. Average job tenure has fallen from 4.6 years to 3.9 years over the past decade, according to Bureau of Labor Statistics figures cited by Deloitte, while baby boomers still average more than eight.3 Every one of those shorter stints is a partial wipe of what someone knew about how the work actually gets done. Most of the time the organization absorbs it. People rediscover the answer, slowly, at the cost of rework and a few avoidable mistakes. Sometimes it can’t absorb it at all.
When the Recipe Leaves With the People
The extreme version of this problem has a code name. Fogbank is a classified material used in US nuclear warheads. When the government set out to refurbish those warheads roughly two decades after production stopped, it found that very few records of the manufacturing process had been kept and that most of the people with the expertise to make the material had retired or left.4 Getting it back took years of reverse engineering and a great deal of money before the material was finally produced again in 2008.4
The mechanism is one most companies run on a smaller scale. The process was performed rarely, by a few people, and the reasons behind each step lived in their experience rather than in anything the organization owned.
Few businesses make anything that exotic. Plenty run a year-end close, a pricing exception process, a regulatory filing or a disaster recovery test that gets exercised once a year by the same two people. That’s the Fogbank pattern in ordinary clothes: low frequency, high consequence, expertise held by a handful of individuals, and no written record of why each step exists.
The Retirement Math Is Already Running
Demographics make this more pressing. More than 30 million Americans will turn 65 in the next four years, and the Peak Boomers impact study put the potential lost output from that retirement wave at $6.9 trillion to $9.6 trillion, figures Deloitte cited in a June 2026 analysis of the problem.3 The same analysis cites APQC research finding that 92 percent of organizations don’t consistently capture knowledge from employees who are about to retire.3 In APQC’s own 2025 report on the retirement wave, 41 percent of organizations said they rarely or never try to collect know-how from retiring employees, and the barrier cited most often was simply day-to-day deadlines.5
Those numbers describe a handoff that mostly isn’t happening. The exit interview captures why someone is leaving. It almost never captures what they knew that nobody else does.
Decisions Survive. Reasons Don’t.
Most organizations are reasonably good at recording what they decided. Policies get published and systems get configured. What disappears is the reasoning: the alternatives that were considered and rejected, the risk a control was meant to catch, the customer complaint that produced a particular exception, the regulator’s letter that started it all.
Without the reasoning, every inherited rule becomes a fence in a field. Engineers know the principle as Chesterton’s fence. Don’t take a fence down until you know why someone put it up. It’s good advice, and in most companies it’s impossible to follow, because nobody wrote down why. So the organization keeps every fence, adds new ones on top, and calls the result complexity.
This matters more now than it did five years ago. Automation and AI tools can only be built on the process as it’s written down, and they inherit the rules without the reasons. A control nobody can explain gets automated as faithfully as one that still matters, and once it’s buried inside a workflow engine it’s even harder to question.
Building Memory That Lasts
Treating memory as infrastructure means giving it what any infrastructure gets: an owner, a budget and a maintenance schedule. Four practices cover most of the ground.
Record the why at the moment of decision. Software teams use a lightweight format called an architecture decision record, a single page that captures the decision, the context, the options considered and the expected consequences. The same format works for a pricing policy or an approval threshold. It takes fifteen minutes while the decision is fresh and is close to impossible to reconstruct three years later.
Find the low-frequency, high-consequence work. Ask each function to list the processes that run once a quarter or less, matter a great deal when they do, and depend on one or two people. That list is the organization’s Fogbank exposure. Each item on it needs a second person who has actually done the work, not just read the procedure.
Make handoffs part of every move. Transfers, promotions, retirements and role eliminations should each trigger a structured handoff with a named recipient and the reasoning behind every open decision still in force. A thirty-day overlap for critical roles is cheap next to a year of rediscovery.
Give every standing rule an owner and a review date. Controls, approval steps and exceptions should each carry a named owner. When the review comes due, the owner either restates the reason or retires the rule. Anything nobody is willing to own is a candidate for removal.
Memory Is an Operating Decision
Organizations spend heavily on systems that store data and comparatively little on keeping the reasons that make the data meaningful. The result is a company that remembers everything it did and almost nothing about why, and that pays for the gap in rework and, every so often, in a rediscovery that costs millions.
The fix doesn’t require a new platform. It requires deciding that memory is part of the operating model, with named owners and a maintenance schedule like any other piece of infrastructure the business depends on. A good place to start is the five oldest approval rules in the business. Ask who can explain each one. Any rule without an answer has already shown where the infrastructure is missing.
References
- Atlassian, “State of Teams 2025: How Top Teams Work Smarter,” 2025. Survey of 12,000 knowledge workers and 200 executives.
- Flip and Workplace Intelligence, “Frontline Generational Skills Gap Study,” March 17, 2025. Survey of 1,500 frontline managers and employees in the U.S., U.K., and Germany, fielded January 2025.
- Eyal Cahana and Evan Siegel, “The $9 Trillion Knowledge Exodus: How Organizations Can Turn Baby Boomer Retirements Into a Competitive Advantage,” Deloitte Insights, June 18, 2026.
- Claire Asher, “This Is the Strangest Weapon the US Military Doesn’t Want You to Know About,” BBC Science Focus, December 13, 2025.
- APQC, “The Great Retirement: Knowledge Loss, AI and the Workforce Shift,” August 2025, as summarized in Tektome, “APQC’s ‘Great Retirement’ Findings: What Teams Can Do About Knowledge Loss,” October 10, 2025.
