Can everyone have a personal agent?How the boundary of delegation moves
In the past, only the few households with the most property could afford a dedicated steward. A steward had to learn each household's harvest, debts and intentions, and then choose ships, keep the accounts and make adjustments on schedule for it. This labour had to be done afresh for every household, and households with little property could not pay for it. Most people could only buy a standardized arrangement, such as simply holding a proportional share of all the ship shares. Once the scribes arrived, this labour cost almost nothing, the threshold fell, and standardized arrangements could become arrangements made for each household. Yet the cost of checking and bearing responsibility remains, and the poorest households are still shut out; of the agency costs, only monitoring falls clearly; and an agent that charges its principal nothing will sooner or later work for whoever does pay it.
1. Stewards used to belong only to the wealthy
On the island, the efficiency of farming, weaving and shipbuilding rises year after year, and grain and cloth grow ever cheaper. A steward's work, however, has not become any faster: getting to know a household, and choosing and checking on its behalf, takes as much time now as it did a hundred years ago. As wages rise in other trades, nobody would be willing to work as a steward unless stewards' pay rose too; so stewards' pay rises as well, and hiring a steward becomes ever more expensive relative to buying a bolt of cloth. The steward has not got worse; other things have simply become cheaper.
In rerunning the step of delegation, only one thing changes: the scribes make the cognitive labour in a steward's work almost free.
Today's counterpart. One-to-one wealth management services usually set a high minimum investment and mainly serve clients with substantial assets.
Why hiring a steward keeps getting more expensive
The efficiency of weaving rises 3% a year, a steward's efficiency stays the same, and wages on both sides rise together. The vertical axis shows a year of a steward's services measured against bolts of cloth. Tick "the scribes arrive" to see what happens after year 30.
Illustrative: Baumol's (1967) two-sector model. The steward's relative price = (1 + 3%) to the power t; after the scribes arrive, the labour a steward needs falls to 20% of what it was, and the remaining checking and bearing responsibility continue at the old efficiency.
2. The threshold: a fixed cost divided by wealth
A large part of a steward's cost does not depend on how much property a household has. Whether a household has a hundred portions or a thousand, learning its circumstances and keeping and checking its accounts take about the same labour. Divide this fixed cost by the household's property and you get the cost of delegation as a share of its property; the less property, the higher the share. Delegation pays only when this share is lower than the benefit delegation brings. Every household compares three approaches: managing on its own, buying a standardized arrangement, or hiring a personal steward; it chooses whichever has the lowest total cost. Once the cognitive labour within the fixed cost becomes cheaper, the cost of a personal steward falls, and households that could previously only manage on their own or buy a standardized product can now afford a personal arrangement.
Reher and Sokolinski (2024) studied a large US robo-adviser that cut its minimum investment from US$5,000 to US$500. Participation by middle-income investors rose clearly, but participation by the poorest investors did not. The threshold came down, but not far enough for everyone to clear it.
Today's counterpart. Robo-advisers offer portfolios allocated by questionnaire and rebalanced automatically, with relatively low minimum investments and fees.
The threshold: a fixed cost divided by property
A year's cost of delegation as a share of property falls as property grows; delegation pays only where it lies below the benefit delegation brings (dashed line). Drag the price of cognitive labour to see where the threshold moves.
Illustrative: fixed cost = 3 portions × the price of cognitive labour + 0.3 portions for checking and bearing responsibility; the benefit of delegation is 1.3% of property. The horizontal axis is property in portions, on a logarithmic scale.
3. From standardized to personal
Households that cannot afford a steward can only accept arrangements designed for the majority: one portfolio, one default proportion, one path that adjusts with age. These arrangements suit the average household but do not necessarily suit any particular one. If someone in the family has an unstable income, or the household will need a large sum in a few years, the right arrangement differs from the average one. The scribes can work things out for each household separately, so standardized arrangements can become personal ones.
D'Acunto, Prabhala and Rossi (2019) studied a portfolio optimization tool introduced by a large Indian brokerage. Investors who were under-diversified before using it came to hold more stocks, and their portfolios became less volatile and earned better returns. Investors who were already well diversified came to hold fewer stocks and traded more often. Users' biases, such as the disposition effect, trend chasing and the rank effect, also declined.
Today's counterpart. Target-date funds are tiered by age; questionnaire-based portfolios are tiered by risk level; finer-grained personal allocation used to be found only in private banking.
From one arrangement to one for each household
Each of forty households has its own best share in ship shares (dots), depending on age, whether its income is stable and whether it will need a large sum in a few years. The short bars show the share actually received. Switch among three arrangements to see the average gap.
Best shareShare actually receivedGap
Illustrative: forty households arranged from youngest to oldest. The best share falls with age and moves up or down with each household's income and spending plans; "tiered by age" takes the average within each of four age bands.
4. Which agency cost falls
What a principal pays for delegation can be divided into three components: the monitoring cost of keeping watch on the manager; the bonding cost the manager incurs to make the principal trust them; and the residual loss caused by diverging interests that remains after monitoring and bonding. The scribes affect the three differently. The cost of monitoring falls clearly: a principal can have their own scribe check, transaction by transaction, what the manager did. The cost of bonding hardly changes: a scribe has no property to pledge, and whoever guarantees it is still the person behind it. The residual loss depends on whom the scribe works for.
Today's counterpart. Tools for reconciliation, fee comparison and holdings analysis have greatly lowered the cost of checking on fiduciaries; liability for compensation and licensing obligations still rest with the institutions.
How much each of the three agency costs falls
Following Jensen and Meckling (1976), agency costs consist of monitoring costs, bonding costs and the residual loss. Drag the price of cognitive labour, then switch whom the scribe works for.
Illustrative: the three components start at 40, 25 and 35. Monitoring costs are 90% cognitive labour; bonding costs are not cognitive labour at all. When the scribe works for the principal, the residual loss falls by up to half as monitoring tightens; when it works for the seller, the residual loss rises instead.
5. It works for whoever pays it
An agent that charges its principal nothing must be paid from somewhere else. If its pay comes from those selling ship shares, as a commission on each sale, it has reason to recommend the ship shares with the highest commission rather than those best suited to the principal. Sellers then compete on this basis too: not over whose ship shares are better, but over who pays the recommender a higher commission. The principal pays nothing directly, but pays more through higher fees.
Today's counterpart. Rules published by the UK Financial Services Authority in 2010 provide that from 31 December 2012 investment advisers may no longer receive commissions paid by product providers for recommending their products; advisers set their own charges and agree them with clients.
It works for whoever pays it
The market offers three kinds of ship shares: a low-fee whole-market portfolio, a mid-fee portfolio, and a high-fee portfolio that pays the recommender the largest commission. Switch where the agent's pay comes from to see what a hundred principals are recommended and how much more they pay each year.
Illustrative: the three kinds of ship shares carry annual fees of 0.1%, 0.8% and 1.8%, and pay the recommender commissions of 0, 0.3% and 1.0%; when the principal pays an annual fee, the agent charges 0.3% a year. The shares recommended are set according to the agent's income, for illustration (following the line of analysis in Inderst and Ottaviani, 2012).
6. A village where everyone has a steward: Civilization Simulator, version 14
The simulator unlocks "Personal agent". Forty households can manage on their own, simply hold a proportional share of everything, or hire a personal agent. A personal agent charges a fixed annual fee plus a percentage fee. The cognitive-labour part falls with the price of cognitive labour, while the part for checking and bearing responsibility stays the same. It can tailor arrangements to each household and helps people chase rises and sell into falls less often. Drag the price of cognitive labour to see what each household chooses and what property each has after twenty years; then have the agent paid by sellers instead, and see what happens.
Civilization Simulator · Chapter 14: a village where everyone has a steward
Forty households are arranged from least to most property. Each chooses to manage on its own, simply hold a proportional share of everything, or hire a personal agent, picking whichever costs least each year. Drag the price of cognitive labour to see who can afford a personal agent and how many times its starting property each household has after twenty years; then have the agent paid by sellers instead.
Manage on its ownHold a share of everythingPersonal agentAgent paid by sellers
Illustrative model: the fleets return 6% a year on average with 15% volatility. Each year's cost is measured as a percentage of property. Managing on one's own costs 1.8% (mistakes in trading and judgement); one household in five is skilled at it and loses only 0.5%. Holding a share of everything costs 1.5% (a 0.1% fee, plus 1.4% from chasing gains, selling into losses and an ill-fitting arrangement). A personal agent costs the fixed fee ÷ property + a percentage fee + 0.2% in residual mistakes, where the fixed fee is 3 portions × the price of cognitive labour + 0.3 portions for checking and bearing responsibility, and the percentage fee is 0.8% × the price of cognitive labour + 0.1%. When sellers pay, the cost the principal sees is only 0.2%, but the principal actually pays 2% more for the high-fee ship shares it is steered towards.
7. What this step leaves open
Personal agents have become cheap, but several questions remain unresolved:
Guarantees. The scribes have no property to pledge and no accumulated reputation. When one does something wrong, who pays compensation?
Checking. A principal can have their own scribe check on the manager, but who checks the principal's own scribe?
The poorest households. The fixed cost of checking and bearing responsibility remains, and the households with the least property still cannot clear the threshold.
When everyone has an agent. When every household has a scribe reading reports and choosing ship shares for it, how will information advantages change?
What this step leaves open
Four questions still unresolved once personal agents become cheap. Select a question to see the assumption behind it.
Personal agentsThe threshold for delegation falls
8. Implications for products
Robo-advisers and wealth assistants built on large models are both pushing the boundary of delegation. Three implications follow. First, the threshold depends on the part of the fixed cost that cannot be brought down. To make delegation affordable for more people, the cost of checking and bearing responsibility must be lowered too, not only the cost of research and advice. Second, the value of a personal arrangement lies in fitting each household, not in looking cleverer; a product should state clearly what it has done for this household that differs from a standardized arrangement. Third, how an agent is paid determines whose side it is on. An agent that charges users directly, with transparent fees, deserves more trust than one that appears to be free.
Chapter summary: how the boundary of delegation moves
Four conclusions, each tied to one mechanism in this chapter.
ThresholdA fixed cost divided by property
As cognitive labour gets cheaper, the threshold falls; checking and bearing responsibility set its floor.
PersonalFrom one arrangement to one for each household
Tailoring that once belonged only to the wealthy can reach many more households.
Agency costsMonitoring falls; bonding does not
Of the three agency costs, only monitoring falls clearly with the price of cognitive labour.
AllegianceIt works for whoever pays it
An agent that appears free may leave the principal paying more.
References
Baumol, W. J. (1967). Macroeconomics of unbalanced growth: The anatomy of urban crisis. American Economic Review, 57(3), 415–426.
Coase, R. H. (1937). The nature of the firm. Economica, 4(16), 386–405.
D'Acunto, F., Prabhala, N., & Rossi, A. G. (2019). The promises and pitfalls of robo-advising. Review of Financial Studies, 32(5), 1983–2020.
Financial Services Authority (2010). Distribution of Retail Investments: Delivering the RDR. Policy Statement PS10/6.
Inderst, R., & Ottaviani, M. (2012). Competition through commissions and kickbacks. American Economic Review, 102(2), 780–809.
Jensen, M. C., & Meckling, W. H. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3(4), 305–360.
Nobel Prize Outreach. The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 1991. NobelPrize.org.
Reher, M., & Sokolinski, S. (2024). Robo advisors and access to wealth management. Journal of Financial Economics, 155, 103829.