Every company devotes considerable resources to producing knowledge: proposals, contracts, technical specifications, meeting minutes, procedures, price lists, customer correspondence. Almost none devotes comparable resources to making it findable. The result is a paradox that anyone who has ever worked in an organization recognizes at first glance: the information exists, someone has already written it down, but finding it costs more than redoing it.
This Study starts from a simple question, one the Italian debate has surprisingly never answered with a number: how much does the time people devote to searching for internal information cost the companies of our country every year? Not new information to be acquired on the market, but information the company already owns, sitting in archives, shared folders, inboxes and, more and more often, in the memory of individual people.
The answer that emerges from the analysis is a figure in the order of 66 billion euros a year for the private sector alone: about three points of gross domestic product, more than double what the entire country invests in research and development. A figure built, as the methodological note explains, on deliberately prudent criteria: the time estimate adopted is the lowest available in the literature, operational staff are assumed by convention to cost zero, and the indirect costs of knowledge that can’t be found, from duplicated work to errors, up to the know-how that walks out the door with every resignation, are left out entirely.
The topic is not new in absolute terms. The moment, in our view, is. The spread of artificial intelligence in companies is making the question more urgent, not less: ever more powerful tools in the hands of individuals coexist with organizations that have not yet built a shared memory. The concrete risk is that individual productivity grows while organizational productivity stands still, widening a gap Italy cannot afford.
// 01-10
The ten key messages of the Study
Italian private companies spend about 66 billion euros a year on time devoted to searching for information they already own, a figure equal to about three points of GDP and more than double the national spend on research and development. The estimate is built to understate.
Every white-collar worker devotes 19% of working time to searching for internal information: 42 days a year, eight and a half weeks, roughly twice the annual leave provided by most collective agreements. The figure, from IDC and the McKinsey Global Institute, is the most conservative in the available literature.
The phenomenon is not shrinking with digitalization: it is getting worse. The applications used on average by a knowledge worker went from 6 to 11 between 2019 and 2022, and nearly one digital worker in two reports struggling to find the information their job requires (Gartner, 2023). More containers, not more access.
The highest incidence is in the most knowledge-intensive sectors. In banking, the search for information absorbs 18.4% of labor cost, 12,452 euros per employee per year; insurance, engineering, IT and consulting follow. For a 100-person consulting firm the annual bill approaches 870 thousand euros; for a bank it exceeds 1.2 million.
The ranking by search cost does not match the salary ranking. Manufacturing, which sits mid-table for pay, slides towards the bottom for search cost, because the phenomenon concerns the white-collar component, not the shop floor. It is this distinction that makes the comparison between sectors informative rather than redundant.
The measured cost is only the visible part. Outside the perimeter remain duplicated work, errors from information not found, delayed decisions and, above all, the knowledge never documented that leaves the company with every termination. In a country of family businesses and a progressively aging workforce, this last item is patrimonial in nature before it is operational.
The problem is organizational before it is technological. Adopting individual-use artificial intelligence tools, in the absence of a shared company memory, moves the problem instead of solving it: each person becomes the manual bridge between systems that don’t talk to each other, with a context cost that grows at every handoff.
Even a partial recovery is worth points of margin. The literature estimates that making internal knowledge searchable can cut search time by up to 35%. Prudently assuming a recovery of one third, a 100-person consulting firm would free up capacity worth about 290 thousand euros a year; a bank of the same size, over 400 thousand.
For small and medium enterprises the stakes are higher than their size suggests. Lacking dedicated knowledge-management functions, SMEs concentrate critical know-how in a few people: the relevant question is not how much time is lost searching, but what stays in the company when those people are no longer there.
What is not measured is not governed. We propose treating search time as a cost item in its own right: measure it, assign it an organizational owner, attack it with information-architecture choices and not with yet another tool. From policymakers, we ask that knowledge infrastructure be recognized among the intangible investments eligible for incentives.
01 The scenario: plenty of information, little access
The obligatory starting point is productivity. In 2024, output per hour worked in Italy stopped at 67.6 dollars, against 83.2 in Germany and 81.6 in France: a gap of about 18% versus Europe’s leading manufacturing economy, reflected in real wages down 4% over the 2015-2024 decade while the OECD average grew by 8% (JobPricing Observatory calculations on OECD data). Italy is one of only four countries in the area where real pay has gone backwards. On this terrain, every point of recoverable productive capacity has a value it would not have elsewhere.
Within this picture, the present Study isolates a cost item that company accounts never see, because it is scattered in fragments of a few minutes across all the days of all the offices: the time spent searching for internal information. The reference measure goes back to a joint study by IDC and the McKinsey Global Institute which remains, years later, the fixed point of the literature: interaction workers, that is managers, professionals and clerical staff whose work requires autonomous judgment and access to knowledge, devote 19% of the working week to searching for information or tracking down the colleagues who hold it. Just under two hours a day.
Anyone objecting that this is a dated snapshot will find in the later literature a rebuttal, not a confirmation. The 2001 IDC survey indicated two and a half hours a day; the 2018 edition attributed to data professionals the loss of half their weekly time across searching, preparation and duplication; the 2023 Gartner survey, run on nearly five thousand workers in organizations with over one hundred employees, finds that 47% struggle to find the information their job requires. Meanwhile, the number of applications used on average by a knowledge worker has nearly doubled, from 6 to 11 in three years. The 19% adopted in this Study is, in other words, the assumption most favorable to companies among those documented.
There is a structural reason why digitalization has not solved the problem it promised to solve. It has multiplied the places where information lives, not the capacity to access it: every new tool is born with its own archive, its own search engine, its own permission logic. The result is that company knowledge has not disappeared: it has dispersed. And the job of connecting the containers has been assigned, by default and without deliberation, to people. It is what seems to us correctly called a context debt: every system that shares no memory with the others offloads onto the organization the cost of reconstructing, every single time, the overall picture.
One final element of scenario deserves mention, because it makes the Italian picture worse than that of its main competitors. The average net annual pay of an Italian worker stops at 24,471 euros, against 26,929 for the European average, 30,832 in France and 31,000 in Germany (Eurostat, 2025 data): a German or French worker takes home over five thousand euros more per year. In a system where labor is paid less and hourly productivity is lower, every hour dissipated searching for what the organization already knows weighs proportionally more. The cost this Study quantifies is not, in other words, a luxury the country can book among its physiological inefficiencies: it is productive capacity subtracted exactly where it is scarcest.
02 The quantification: 66 billion a year, understated
The estimation logic is deliberately elementary, because a measure that aims to enter the public debate must be verifiable by anyone with a calculator. The annual cost of information search per employee is the product of three quantities: the sector’s average total employment cost (gross annual full-time-equivalent pay multiplied by 1.40, to cover employer contributions, severance accruals and ancillary charges), the share of white collars in total headcount, and the 19% of time devoted to searching. Sector pay comes from the JobPricing Observatory (JP Salary Outlook 2026, 2025 data); the employment structure is anchored to the INPS Observatory on private-sector employees, which for 2024 counts 6,490,467 clerical workers, 552,934 middle managers and 141,718 executives: 7,185,119 white collars, 40.5% of private dependent employment.
Two choices deserve to be made explicit because they orient the entire reading of the results. The first: the 19% is applied exclusively to the clerical, middle-management and executive component; operational staff are attributed, by prudential convention, a search time of zero. The second: the cost so calculated is referred to the entire headcount, not to white collars alone. The indicator answers the question that matters to whoever runs a company: how much do I spend on information search for every person on the payroll, whoever they are.
The table below shows the picture for the 17 sectors for which 2025 average pay is available.
The 17 sectors ranked by annual cost of information search
#
Sector
Cost per employee (€/yr)
Company of 100 (€/yr)
% of labor cost
01
Banking and financial services
12,452
1,245,179
18.4%
02
Insurance
10,308
1,030,790
18.4%
03
Engineering
9,956
995,586
16.2%
04
IT services, consulting and software
9,808
980,837
17.7%
05
Telecommunications
9,477
947,676
16.2%
06
Legal, tax and management consulting
8,688
868,833
18.1%
07
Media, communications and publishing
8,196
819,612
16.2%
08
Pharmaceuticals and biotech
5,985
598,455
10.5%
09
Oil & Gas
5,872
587,160
10.5%
10
Energy, utilities and environmental services
5,611
561,138
11.4%
11
Aeronautics
5,094
509,417
9.5%
12
Chemicals
4,426
442,555
8.6%
13
Electronic equipment and automation
4,319
431,937
8.6%
14
Machine tools, plant and metal
3,201
320,096
6.7%
15
Personal services
2,608
260,801
6.7%
16
Hotels and restaurants
1,508
150,790
3.8%
17
Agriculture
850
85,035
2.3%
Source: calculations on JobPricing Observatory, INPS and IDC-McKinsey Global Institute data.
Three findings stand out clearly. The first: banking wears the black jersey, at 12,452 euros per employee per year and 18.4% of labor cost absorbed by the search for information. Put differently: in a one-hundred-person bank, it is as if eighteen people worked full time doing nothing but searching. Insurance and consulting follow at a short distance, both with an incidence above 18%.
The second finding is the reshuffle versus the salary ranking. Engineering, third for pay, stays third; but legal and management consulting, fourteenth for gross pay, climbs to sixth place for search cost, while pharmaceuticals slides from the top of the pay table to the middle of the one proposed here. The explanation lies in the employment structure: where production is physical, the phenomenon touches a minority of the workforce; where the product is knowledge, it touches almost everyone. This is why a ranking of search cost says something the salary ranking does not.
The third finding is the national scale. Multiplying the 7.19 million private-sector white collars by the unit search cost calculated on average clerical pay (9,213 euros a year) yields an estimate of 66.2 billion euros per year: about 3% of GDP, more than double the country’s total spend on research and development. The direction of the bias is worth restating: having applied clerical pay to middle managers and executives too, the true value is higher than the one shown.
Knowledge that can’t be found is knowledge that doesn’t exist.
03 The drivers: why knowledge gets lost
If the phenomenon were a residue of technological backwardness, its trajectory would be downward. The data say the opposite, and the reasons are at least four.
Application fragmentation. Eleven applications per knowledge worker mean eleven archives, eleven search engines, eleven permission systems. Each tool was adopted for a good local reason; none was adopted with the searchability of the whole information estate in mind. The right information almost always exists: the problem is that nobody knows which container it sits in, and the search becomes a pilgrimage across systems that ends, not rarely, with a question to the colleague who happens to remember.
Tacit knowledge and turnover. A significant share of company know-how has never been written down: it lives in the experience of those who were there. In a productive fabric where the average age of the workforce is rising and retirements accumulate, every termination is a small uncounted patrimonial loss. Italian family businesses know the extreme version of the problem well: the rarely confessed fear of what will happen when the person who knows how it’s done is no longer there.
Individual-use artificial intelligence. The wave of generative tools of recent years has raised the productivity of individuals, but in most organizations it has arrived as personal equipment, not as a system. Tools that share no memory with each other, nor with company archives, add containers to the existing fragmentation and once again hand people the job of carrying context from one system to another. The paradox deserves attention: the very technology that could cut the cost of search is, in many implementations, feeding it.
Company size. Large organizations have, at least on paper, knowledge-management functions. Italy’s fabric of small and medium enterprises has nothing of the kind, nor could it afford one in the traditional forms: knowledge management coincides with the memory of individuals and the goodwill of whoever documents. What follows is a fragility specific to our productive system, all the more serious because the competitive advantage of SMEs lies in applied know-how that is hard to replicate, but also hard to transmit.
04 What it means for companies
Reading by size class restores the concreteness that percentages hide. For a 50-employee professional-services firm, the annual cost of information search sits between 430 and 500 thousand euros; at 100 employees it ranges from the 320 thousand euros of industrial machinery to the 870 thousand of consulting, exceeding 1.2 million in banking; at 250 employees an IT firm approaches 2.5 million. These are orders of magnitude that, in the leading sectors, compare with operating margins, not with overheads.
Not all of this time is compressible, nor would it be desirable that it were: a share of exploration is part of intellectual work, and no information architecture will eliminate it. The estimate of the addressable share available in the literature nevertheless indicates ample margins: up to 35% of search time, when internal knowledge becomes genuinely searchable. Assuming a recovery of just one third, the consulting firm of the example frees capacity worth about 290 thousand euros a year; the bank, over 400 thousand. Capacity, mind, not savings: people don’t cost less, they go back to doing the job they were hired for. In a country where productivity is the binding constraint, it is the reallocation that counts.
For SMEs the reasoning changes nature. Below a certain size threshold, search time remains a proportionally significant cost, but the dominant risk is another: the concentration of critical knowledge. When the complex quotation is something one person knows how to do, the negotiation with the historic supplier is something one person knows, the machine calibration is in one person’s head, the company holds an intangible asset that is neither on the balance sheet nor insured. The right question, for the owner, is not how much time is lost searching: it is what stays in the company the evening that person hands back the keys.
05 What recovery is worth: three scenarios
Having established the cost of the phenomenon, the next question is how much of it is realistically recoverable. The literature offers a ceiling: up to 35% of search time disappears when internal knowledge becomes genuinely searchable and answers arrive with their sources attached. Below that ceiling, the outcome depends on how you intervene. We built three scenarios, from the most cautious to the documented limit, applying them to two typical 100-employee companies and to the entire private productive system.
The conservative scenario (15% recovery) corresponds to targeted interventions: archive reorganization, naming conventions, a search engine spanning the existing repositories. The intermediate scenario (25%) presupposes a system with shared memory, able to answer people’s questions by drawing on the document estate and citing sources. The full scenario (35%) adds the systematic capitalization of knowledge as a process: what the organization learns is retained, not merely found again.
What recovery is worth, by scenario
Recovery scenario
Consulting, 100 employees (€/yr)
Bank, 100 employees (€/yr)
Italy, private sector (€bn/yr)
Conservative · 15%
130,000
187,000
9.9
Intermediate · 25%
217,000
311,000
16.6
Full · 35% (literature ceiling)
304,000
436,000
23.2
Calculations on model values; national scenario applied to the €66.2 billion estimate.
Two warnings prevent naive readings of these numbers. The first is that this is freed capacity, not cash savings: personnel cost stays unchanged, what changes is the share of time returned to the work people were hired for. In system terms, the intermediate scenario is equivalent to putting back into the economy the productive capacity of over 350 thousand full-time white collars, without hiring a single one. The second warning is that no scenario materializes through technology alone: the discriminating variable, in the experiences observed, is the organizational set-up that accompanies it. And that is what the proposals of the next chapter are dedicated to.
06 The proposals
The indications that follow descend from the analysis and are addressed first of all to whoever runs a company; the last one to policymakers.
First: measure. No company would manage labor cost without knowing it; the cost of information search has been managed that way forever. Even a simple internal survey, run on a sample over a typical week, is enough to build the company baseline and replace impressions with a number. Experience teaches that the number, once seen, does not let itself be ignored again.
Second: assign ownership. The findability of knowledge is today nobody’s job, and nobody’s jobs don’t improve. No dedicated function is needed: what is needed is that someone on the management committee answers for the question “how long does a new person take to find what they need?” with the same naturalness with which the colleague next to them answers for revenue.
Third: capitalize knowledge as a process, not as a favor. As long as documenting remains an act of goodwill subtracted from the real work, it will not happen. The organizations that have moved the needle have made documentation an output of the process: closing a project includes its write-up, the offer won and the offer lost both feed the archive, a person’s departure includes the structured transfer of what they know.
Fourth: choose systems with memory, not tools with a license. The criterion for adopting technology must be flipped: not what the tool does for the individual user, but what it leaves to the organization. Systems that share a common memory, cite internal sources and return knowledge in the context where it is needed reduce the context debt; individual tools, however brilliant, increase it. It is the difference between signing a star player and building a team.
Fifth: recognize knowledge infrastructure among incentivable investments. The industrial policies of recent years have decisively supported tangible capital goods and, more recently, the digital and energy transitions. The information estate of companies, which this Study estimates erodes the equivalent of three points of GDP every year, still has no citizenship among the assets deserving support. Extending the existing instruments to investments in organization and knowledge searchability, with priority for SMEs, would have a contained cost and a measurable productivity return.
07 Methodological note and limits of the Study
The formula adopted is: annual search cost per employee = sector average employment cost × white-collar share × 19%. Employment cost equals the sector’s average gross annual full-time-equivalent pay (JobPricing Observatory, JP Salary Outlook 2026, 2025 data) multiplied by 1.40. FTE values were used, rather than administrative-source average pay, because the latter incorporates part-time and shorter-than-year contracts, distorting the cross-sector comparison in seasonal industries. The share of time devoted to searching comes from IDC and the McKinsey Global Institute (The Social Economy, 2012) and refers to interaction workers; the later corroborations (IDC 2001 and 2018; Gartner 2023) indicate equal or higher values.
Sector white-collar shares are the authors’ estimates, bound to two verifiable anchors: the 2024 INPS national composition (40.5% of clerical staff, middle managers and executives over private dependent employment) and the contractual structure of the extreme sectors, starting with banking and insurance, where the blue-collar grade is virtually absent by construction of the collective agreements. Sensitivity analysis on the model shows that a variation of ten percentage points, up or down, applied to each sector share alters neither the order of magnitude of the results nor the structure of the ranking; the same holds for the employment-cost coefficient in the 1.30-1.50 range and for the time share in the 10-25% range. Even halving the time assumption versus the literature figure, the cost for a 100-employee consulting firm would remain above 450 thousand euros a year.
Three limits should be declared plainly. The first: the 19% share is an international average, not a survey run on Italian companies; no comparable national measurements exist to date, and it is also to fill this gap that the Study is published. The second: the perimeter covers the 17 sectors for which 2025 average pay is available, and leaves out relevant industries such as retail, logistics and construction; being sectors with a low-to-medium clerical share, their inclusion would not change the top of the ranking. The third: the national estimate applies the average clerical pay to all white collars, ignoring the premia of middle managers and executives; by this route too, the value shown is a floor, not a ceiling.
The data and the full calculation model, with open formulas, are available on request.
Concluding remarks
Italian companies do not have a problem of knowledge scarcity: they have a problem of access to the knowledge they already own. The bill for this problem, however invisible to the accounts, is in the order of three points of GDP a year, and the technological trajectory under way can shrink it or widen it depending on how organizations choose to adopt it. The good news contained in this Study is that this is a cost that can be attacked with organizational decisions more than with extraordinary investments. The bad news is that nobody will attack it on companies’ behalf. The estate of know-how accumulated over decades of work deserves the same guardianship reserved for plant and cash: what a company knows, and manages to find again, is by now part of its assets. The rest is cost.
References
IDC, McKinsey Global Institute, The Social Economy: Unlocking Value and Productivity through Social Technologies, 2012.
IDC, The High Cost of Not Finding Information, 2001; IDC, The State of Data Science and Analytics, 2018.
Gartner, Survey on Digital Workers, May 2023 (4,861 full-time workers, organizations with over 100 employees).
INPS, Osservatorio sui lavoratori dipendenti del settore privato non agricolo, Statistiche in breve, November 2025 (2024 data).
JobPricing Observatory calculations on OECD data, productivity and real wages, 2015-2024.
// And your company?
How much does knowledge that can’t be found cost you?
ekory builds your company brain: one shared memory that connects the tools you already use and answers with sources attached. A working assistant on your own materials in about a day.