Almost every article about founder delegation is written by someone who wants you to delegate to them. Virtual assistant agencies, fractional operators, recruitment firms, coaching practices, they all publish the same piece, and it always concludes that you are the bottleneck and you should let go of more. The advice is not wrong, exactly. It is just being given by an interested party, which is why it never contains the second half.
The second half is that some handovers are not a transfer of work. They are a transfer of signal, and the signal does not survive the trip.
That distinction is the whole of this post. When you delegate invoicing, someone else does the invoicing and you still know what was invoiced. Nothing is lost. When you delegate the first conversation with a prospective customer, someone else has the conversation, and what reaches you afterwards is a summary written by a person with an incentive to sound competent. The work moved. The information did not move with it. It was compressed, filtered, and rounded off, and a year later you are making decisions on a picture of your own business assembled entirely from other people's summaries.
I have delegated three of the five things below at different points, and each time the cost showed up somewhere between four and twelve months later, in a form that did not obviously trace back to the handover. That delay is the reason this is hard to learn. If losing the signal hurt immediately, nobody would need the list.
What makes a function undelegatable
A function should stay partly yours when the by-product of doing it is worth more than the work itself, and the by-product cannot be written down.
Most tasks fail this test. Bookkeeping produces a ledger, and the ledger is the point, you can read it later and know everything the bookkeeper knew. Scheduling produces a calendar. Payroll produces payslips. In each case the artefact carries the full information content of the work, so handing over the work costs you nothing but the hours.
The five below are different. Each one produces an artefact that is a genuine but lossy record, and a founder reading only the artefact ends up confidently wrong. The pattern to look for:
- The useful part is tone, hesitation or wording, things that do not survive being typed into a summary field.
- The person reporting has a stake in how it reads. Not dishonesty; just the ordinary human tendency to present your own work in its best light.
- You would change a decision if you had heard it directly. This is the sharpest test. If hearing it first-hand would not change what you do, delegate it and stop worrying.
- The absence of the signal is silent. You do not get an alert saying "you no longer know why customers say no."
That last property is what makes this list worth having in advance. Every other business problem announces itself. These do not.
1. The first sales conversation
Keep taking a portion of first calls yourself, permanently, because the reason a prospect does not buy is almost never the reason your salesperson reports.
This is first on the list because it degrades fastest and is disguised the longest. A salesperson's summary of a lost deal is honest and useless: price, timing, went with someone else. Those are the categories the CRM offers. What actually happened is that four minutes in, the prospect said something slightly confused about what you do, your salesperson smoothly corrected it, and the deal was already gone by then. The confusion is the finding. It never reaches the CRM because it was resolved in the moment by a competent person doing their job well.
Multiply that across a quarter and you get a positioning problem you cannot see. Your own description of your business drifts away from how the market hears it, and the only people who witness the gap are the people paid to close it, for whom the gap is a routine objection rather than news.
What happens the year you hand it over completely: your messaging stops updating. The website, the deck and the pitch keep saying what was true two years ago. Win rates soften slowly enough that it reads as market conditions. You start hearing "the market is tougher now" from people who are not lying but are also not in a position to know the difference between a harder market and a stale pitch. When performance channels get expensive at the same time, the diagnosis gets even murkier, that is the situation when performance marketing stops working and brand is the only lever left is written about, and a founder with no direct call exposure has no way to tell the two apart.
The surrogate if you must step back: do not ask for a report. Reports are the failure mode. Instead, listen to or sit in on a fixed small number of calls per month, a genuinely small number, chosen by someone other than the person being observed, and including at least one deal that was lost. The rule that matters is that you do not pick the calls, because you will pick interesting ones and the salesperson will offer flattering ones. Rotate whose calls you hear.
2. The pricing decision
Never fully delegate the decision to change a price, because pricing is the one lever where everyone else in the business is structurally biased toward the same direction.
Sales wants a lower number, because a lower number closes. Delivery wants a lower number, because a higher one raises expectations they have to meet. Finance wants a higher number in the abstract and will defer to sales in the specific. Nobody in the room carries the cost of a discount for the full eighteen months it persists, except you.
This is not a claim that founders price better. Often they do not; founders under-price out of anxiety at least as often as employees under-price out of incentive. It is a claim about who absorbs the consequence, and pricing authority should sit with whoever absorbs it. If you delegate the number, delegate it with a floor and an exception process, not as a blanket authority.
There is a second-order effect that is easy to miss. Once discounting authority is distributed, every account's price becomes a record of who negotiated it and when, rather than of what the work costs and is worth. Two years on, your rate card is fiction and your actual realised rates are a scatter plot nobody can defend to a client who compares notes with another. Untangling that is a much harder exercise than the original decision, the sequence for it is in how to raise your prices without losing the accounts you actually want to keep, and the reason it is hard is almost always that the pricing decision was distributed years earlier without anyone deciding to distribute it.
What happens the year you hand it over completely: average realised rate drifts down while your rate card stays put, so the gap between them becomes invisible in reporting. Nobody notices, because no single discount is unreasonable. The compounding shows up in margin, and margin problems get attributed to delivery efficiency, which sends you optimising the wrong side of the equation. If you have never set the number deliberately in the first place, start at how to price your services as a freelancer or consultant rather than at the discount policy.
The surrogate: a floor that cannot be crossed without you, and a monthly list of every deal that landed within ten per cent of the floor. The second half is the part people skip. A floor with no visibility just relocates the drift to just above the floor.
3. The first-time complaint from a customer
Stay on the path of the first serious complaint from any customer, because the first complaint is the only one that arrives before the customer has decided how they feel about you.
By the second complaint, the customer has a theory. They have concluded you are careless, or slow, or that this is just how you are, and everything after that is filtered through it. The first one is different: they are genuinely still asking a question, and the answer determines which story they tell for the rest of the relationship. That is a brand event, not a support ticket, and it is the mechanism behind your brand is whatever your worst touchpoint is.
You do not need to handle it. Handing complaint resolution to someone good at it is usually an upgrade, most founders are worse at this than a trained support person, because founders get defensive about their own work. What you should not do is fall off the notification path entirely. There is a large difference between "I do not resolve these" and "I do not know these happen."
What happens the year you hand it over completely: you lose the leading indicator on churn. Complaints are the earliest thing that moves, well before renewal conversations and long before revenue. A team resolving them competently produces exactly the same clean dashboard as a team with nothing to resolve, and you cannot distinguish the two from the outside. The related failure is that you stop being able to tell a difficult customer from a customer you are genuinely failing, which is precisely the judgement when to fire a client turns on.
The surrogate: be on the notification, not in the workflow. You see that it happened and what it was about; you do not respond unless asked. And whoever answers first should already hold the authority to fix it, the amount they can spend or concede without asking permission needs to be decided before the complaint arrives, because a responder who has to escalate for authorisation produces a delay the customer reads as indifference. This is not a new observation, which is some comfort when you are arguing for it internally: Hart, Heskett and Sasser made the same case in The Profitable Art of Service Recovery in 1990, arguing that service companies cannot prevent every failure but can learn to recover from one — and that recovery depends on front-line people who are trained, trusted and empowered to act the moment the problem surfaces. The part worth stealing is the framing: the authority is not a reward for seniority, it is a piece of infrastructure you install before you need it.
4. The exit interview
Do the exit conversation with departing employees yourself, because the reason someone leaves is the one piece of information about your company that nobody has an incentive to give you accurately except the person walking out.
And even they will not, if the wrong person is in the room. An exit interview run by the manager the person is leaving is not an exit interview; it is a formality with a predetermined answer. It will say "a great opportunity came up." It will almost never say "my manager takes credit for my work" or "I asked for scope twice and nothing changed," and those are the only two categories of answer worth the meeting.
The failure is well documented at much larger companies than yours. Reviewing the practice in Making Exit Interviews Count, Everett Spain and Boris Groysberg found that too few leaders pay the practice any real attention, and that the programmes which do exist typically fail either to improve retention or to produce useful information — owing to poor data quality and the absence of any consensus on how it should be collected. If organisations with dedicated HR functions produce unusable exit data, a form-based process in a twenty-person company will not do better by itself.
This one is the most commonly and most confidently delegated of the five, usually to HR, on the reasonable-sounding grounds that HR is neutral. It is a mistake in a small company for a simple structural reason: in a small company HR reports to someone the departing person may be leaving because of. Neutrality is a property of the reporting line, not of the job title.
What happens the year you hand it over completely: you accumulate a stack of exit forms that all say "career growth," and you conclude you have a compensation problem, because compensation is the thing that shows up when nobody will name a person. You then spend money on the wrong fix. Meanwhile the actual pattern, one manager, one broken process, one promise made at hiring and not kept, keeps operating and keeps costing you people. The cost of that mistake is not really the salary; it is the second-order damage described in startups and team building, where the same departure repeats with a new name each time.
The surrogate: if you genuinely cannot do it, it must at minimum be run by someone outside the departing person's reporting line, and you should read the raw notes rather than a themed summary. Themes are where the specific name gets removed, and the specific name was the finding. A useful discipline: after six months, look back at whether the reason given at exit matches what you now know. The gap between those two is a direct measure of how much your exit process is worth.
5. The weekly cash-in, cash-out review
Look at what actually came in and what actually went out every week, yourself, because the monthly P&L will tell you a business is healthy right up to the week it cannot make payroll.
This is last on the list not because it matters least but because it is the one most founders keep instinctively. Still, it gets delegated more often than you would expect, usually not deliberately, but by a gradual slide from reading the bank to reading a summary of the bank to asking whether we are fine.
The distinction that makes this non-optional is the one in cash flow vs profit: profit is an opinion assembled monthly under accounting conventions, and cash is a fact you can check today. A services business can be comfortably profitable on paper and still be a few late payments from a genuine problem, because revenue is recognised when invoiced and salaries are paid when due, and those two calendars do not align. In India there is an extra wrinkle worth knowing precisely, which is that GST and TDS both hit your cash before the client does.
What happens the year you hand it over completely: you lose the ability to feel a slowdown early. Collections stretching from thirty days to fifty is invisible in a monthly profit statement and obvious in a weekly cash view. The founders who get caught out are rarely the ones who did not understand their numbers; they are the ones who understood their numbers monthly, in arrears, in a business where the thing that changed was weekly.
The surrogate: there is not really an acceptable one, and that is why it is on the list. The nearest thing is a fixed weekly view of cash in, cash out, and money owed to you sorted by age, something you read yourself, in the same format every week, taking a few minutes. Not a meeting. Not a deck. The format has to stay identical week to week or you lose the only thing that makes it useful, which is that changes stand out.
The five, side by side
| Function | The signal you lose | How it shows up | Acceptable surrogate |
|---|---|---|---|
| First sales conversation | Why people actually say no, in their words | Messaging goes stale; softening win rates blamed on "the market" | Sit in on a few calls a month, chosen by someone else, including losses |
| Pricing decision | What the market will genuinely bear | Realised rates drift below the rate card; margin blamed on delivery | Hard floor, plus a monthly list of deals landing near it |
| First-time complaint | Earliest churn indicator; whether you are failing them | Clean dashboards, surprising non-renewals | Be on the notification, not in the workflow |
| Exit interview | The specific manager, promise or process that is costing you people | Every exit form says "career growth"; you overspend on pay | Someone outside the reporting line; you read the raw notes |
| Weekly cash review | Collections slowing before it reaches the P&L | Comfortable monthly profit, sudden payroll pressure | None good, a fixed weekly view you read yourself |
What you should delegate hard, and early
Delegate anything whose written output contains everything the work contained, and delegate it sooner than feels comfortable.
That covers most of what founders actually cling to: bookkeeping, scheduling, compliance filing, procurement, expense processing, first-line support triage, reporting production, most of recruitment logistics. In every one of those, the artefact is the information. There is no residue that only lives in the doing.
Two clarifications, because this list gets misread in both directions:
"Never fully delegate" is not "do it yourself." Four of the five above are compatible with someone else doing essentially all the work. What is reserved is exposure, not execution. The failure mode of taking this post too literally is a founder who insists on personally running functions they are worse at than the person they hired, which costs you the hire.
Access is not the same as involvement. A related and much cheaper mistake is losing administrative control of the systems these signals live in, advertising accounts, analytics, the domain, the payment processor. That is a separate annual exercise and worth doing on its own schedule: who owns the ad account covers it.
How to tell whether you have already lost one
Ask yourself five questions once a quarter. Each one has a right answer, and the wrong answer is always some version of "I would have to ask someone."
- Why did the last deal you lost actually get lost? If the answer is a CRM category rather than a sentence somebody said, you have lost signal one.
- What is the lowest price anyone in your business agreed to in the last ninety days, and who approved it? Not knowing means the pricing floor is theoretical.
- What was the last thing a customer complained about for the first time? If nothing comes to mind, either you have an unusually good quarter or you are off the path.
- Why did the last person who left, leave, in their words, not the form's? "Career growth" is the sound of the answer not reaching you.
- What is your cash position today and how does it compare to four weeks ago? If you need to ask, you are reading your business monthly in a world that moves weekly.
Getting one wrong is normal and fixable. Getting three or more wrong is the actual condition this post is about, and it does not usually feel like a crisis while it is happening. It feels like being appropriately senior.
The trap on the way back
If you find you have lost one, do not take the whole function back, take back the exposure and leave the work where it is.
This is where founders do real damage. The discovery that you no longer know why deals are lost feels like evidence that the handover was a mistake, and the instinct is to reclaim the function entirely. What that actually communicates to the person doing the job is that they failed, when in most cases they were never asked to preserve the signal in the first place. You lose a good hire over a design flaw that was yours.
The correct repair is narrow and specific: name the signal, put a mechanism in place that carries it to you, and tell the person plainly why. "I want to hear two calls a month, including one we lost, because our positioning drifts and you are not the right person to catch that" is a sentence that costs nothing. "I am going to start taking the first calls again" costs you a salesperson.
Do this in the sequence above and delegation stops being a question of how much you can let go of. It becomes a question of which information you have arranged to keep receiving, which is the version of the problem that actually has an answer.
There is a short-horizon version of the same exercise that is worth running even if you never restructure anything, and it has a natural deadline attached: the handover document that decides whether you can take a week off. Deciding what a covering person may settle without you — and, just as importantly, the two or three things that simply wait — forces the same list into writing, and the week itself is the most honest audit of your operational dependency you will ever get.
