What your firm cannot see across clients

The previous guide described a blind spot on the client’s side. The founder can see the bank balance, a point, but not the sequence of commitments the firm assembles into a forecast. The individual facts may be available to them, while the combined timing remains invisible.

There is a matching blind spot on the firm’s side, and it works the same way. A partner can see every client in the portfolio, and still not see a capability that is missing across all of them. This guide is about that capability, why it is so hard to see, and why the answer is not simply “hire someone”.

The short answer

The controller gap is not necessarily a missing job title. It is a portfolio of small attention tasks with no visible queue: unverified commitments, stale forecasts, missed expectations and client questions. Each fragment gets handled somewhere, but the firm cannot see whether the capability is covered across all its clients.

The need for additional bookkeeping capacity is comparatively legible. Transaction volume rises, the reconciliations slip, someone is visibly underwater, and the hire follows. Because bookkeeping workload can usually be counted, it tends to scale with client count.

The work that sits above bookkeeping and below the partner has no equivalent trigger. Nothing in a small firm announces that it has become unmanageable. It is absorbed by partners in fragments, a question here, a forecast refresh there, a commitment carried in someone’s head, and the firm grows for years without ever hiring for it or deciding who holds it.

Part of the reason is that the role itself is blurry. One practitioner calls it a controller, another a VP of finance, a third an FP&A function, and each of those names carries assumptions the others reject. You cannot hire a role you cannot name, and you cannot see a gap you have no word for. The fix begins with noticing the work, not with a job description.

This guide draws on qualitative design-partner conversations with CPA and fractional CFO practices. We describe what practitioners told us about their own firms.

The bookkeeper trigger is legible

It is worth being precise about why bookkeeping workload can usually be counted, because the contrast is the whole point.

Bookkeeping work is countable. Transactions, accounts, entities, reconciliations. When a firm adds clients, the volume rises in a way that can be measured, and the strain shows up in places everyone can see: the close runs late, the reconciliations lag, the person doing the work says so. The hiring decision is not easy, but it is legible. The firm knows what it is buying, roughly what it costs, and what “enough” looks like.

The practitioners we spoke with could readily describe the moment their firm knew it needed another bookkeeper. They struggled to describe an equivalent moment for the layer above.

The controller signal does not form a visible queue

The work above bookkeeping is not countable in the same way. It is not a volume of transactions. It is a set of things that need to be held open and watched.

Someone has to remember what the client committed to on the last call and check whether it happened. Someone has to notice that the forecast said forty thousand would land on the fifteenth, and it is now the eighteenth and it has not. Someone has to answer the question the client sends on a Tuesday afternoon, quickly enough that the client does not act without asking. None of that produces a backlog you can point at. When it is not done, nothing visibly piles up. The miss just becomes a surprise on the next call.

Because the strain never surfaces as a queue, it never trips a hiring decision. And because the work arrives in fragments, it gets absorbed in fragments. A partner takes the Tuesday question. A senior refreshes the forecast when there is time. The commitment from the last call stays in the partner’s memory, or in the notes, and is checked if someone happens to remember. Each fragment is small enough to absorb, so nobody adds them up.

Each fragment looked reasonable in isolation; no single event called for a staffing decision. That is what a blind spot looks like from inside.

You cannot hire a role you cannot name

There is a second reason the gap stays invisible, and it is the one we found most interesting.

When we started describing this layer as controller work, two advisers pushed back. To them, controller meant the accountant who closes the books and reports on what already happened. Backward-looking, by definition. What we were describing, watching a forecast and catching a miss while it is still actionable, sounded to them like FP&A, or like something a VP of finance would do.

They were not wrong about what those words mean to them. They were describing the problem. The same set of functions is called a controller in one firm, an FP&A analyst in a second, a fractional VP of finance in a third, and each label comes with its own assumptions about seniority, scope and cost.

A firm that wanted to hire for this would have to write a job description for a role the market does not agree on. The role often never becomes explicit. Not because the firm has judged the work unnecessary, but because it has never had a clean name for it, and a gap with no name does not get onto the agenda.

What the work actually is

So set the titles aside and look at the functions. Across the firms we spoke with, the same four kept appearing.

Continuity across clients. On a client call, decisions get made and commitments get stated. Somebody has to carry those forward so that the next conversation starts from what was agreed, not from a reconstruction of it. For one client this is a matter of good notes. Across a portfolio it is a capability, and it becomes increasingly fragile as a firm grows, because it depends on the same partner remembering across dozens of relationships.

Forecasting. Not producing the forecast once, but keeping it alive. Refreshing it as the books change, folding in the timing the client mentioned on the call, and keeping it current enough that it is worth consulting between check-ins. A forecast that is only rebuilt before the meeting is a report about the meeting, not a working instrument.

Backward and forward at the same time. This is where the controller label earns its keep, and where the pushback we received turned out to be useful. The books say what happened. The forecast says what was expected. The work is the comparison between them, held open continuously: the receivable that was supposed to arrive, the payroll that ran higher than planned, the distribution the client took that the forecast did not know about. A bank feed shows what arrived. It cannot show what was supposed to arrive and did not. Catching that on the eighteenth, while the client can still do something about it, is different from reporting it at the end of the month.

Answering questions. Clients ask things between calls. Can I take this distribution. Can we hire in October. What happens if this customer pays late again. Some are factual questions that current evidence can answer; others require partner judgement. The factual ones are often questions about what the forecast already says, asked by someone who does not read the forecast. When the firm cannot answer them quickly, the client decides on their own, which is exactly the client-side blind spot the previous guide described.

Read together, these four functions are one job. Either the bookkeeper or the partner may perform it, but someone must own it explicitly. In a mid-sized company it would have a name and a desk. In a small firm serving many small clients, it has neither, so it is done by whoever is nearest when a fragment of it appears.

When the bookkeeper already holds it

One thing needs saying plainly, because the argument above could be read as a claim that every firm has a hole. It does not.

In some of the firms we spoke with, the bookkeeper already does this work. They have been with the client for years, they know what was agreed, they notice when a receivable is late, and they answer the Tuesday question because they are the one the client trusts. Nobody assigned them the controller function. They grew into it.

That is a real path, and it works. It is also fragile in a specific way: it lives in a person, it does not transfer when that person leaves or when the portfolio doubles, and it is invisible in the firm’s own picture of how it is staffed. A firm in this position does not have a gap. It has a capability it has not named, which is the same blind spot in a more comfortable form.

Why this is a blind spot and not a failure

We want to be careful here. The previous guide argued that the client’s blind spot is structural, not carelessness. The same is true on the firm’s side, and for a stronger reason.

The bookkeeping gap is visible because bookkeeping is legible. The controller gap is invisible because the role is contested, the work arrives in fragments, and the fragments are individually easy to absorb. A firm that has grown from eight clients to thirty without ever hiring for this layer has not neglected anything. It has responded correctly to every signal it received. The problem is that those signals arrive as isolated client events rather than one visible measure of workload.

Seeing it requires stepping back from any individual client and asking a question the work itself never prompts: across all of them, who is holding the expectations open, and what happens to the ones nobody is holding.

Why it matters economically

The response we hear most is a reasonable one: we are doing fine. Clients are happy. The calls go well. Why staff a function nobody has asked for.

An unassigned attention layer creates work the firm cannot see clearly or price accurately. It appears as partner interruptions, last-minute reconstruction before calls, and recurring review performed between billable commitments. The engagement may look profitable while depending on attention that is never included in its delivery model.

Naming the function makes the economics visible. The firm can decide what belongs in the engagement, who owns it, when it escalates to the partner, and whether the work should be priced, delegated or systematised. The value is not merely additional client contact. It is more consistent delivery across the portfolio without relying on one person to remember every open thread.

What to do about it

The recommendation is not “hire a controller”. For many of the practices we spoke with, the economics of a full-time hire do not work, and the title would not survive the first conversation about what it means anyway.

The recommendation is to notice that the work exists, name it in whatever words your firm uses, and decide deliberately who holds it. That might be a bookkeeper who has already grown into it, in which case the decision is to recognise and protect that. It might be a partner, in which case the decision is to acknowledge the limit on how many clients one person can hold open in their head. It might be a shared function, or a system, or a combination.

What matters is that the decision is made rather than defaulted. The blind spot is not that the work is undone. It is that nobody chose who does it.

Where AgentLink fits

AgentLink is built for the attention half of this layer. It holds the expectations open: the commitments from the last call, the timing in the forecast, the question the client asked on Tuesday. It watches actual activity against what was expected and surfaces the miss while it is still actionable, under the firm’s brand and within boundaries the firm sets. When the evidence is incomplete or the question requires consequential judgement, it says so and routes it to the firm.

The judgement stays with the partner. That is not a limitation we are working around. It is the design. Software can sustain repeated observation across a growing portfolio without requiring one person to remember every thread. It cannot be accountable for advice, and it should not pretend to be.

Frequently asked questions

Is this not just what a good bookkeeper does?

In some firms, yes, and where that is true it should be recognised as a real capability rather than a happy accident. The point of this guide is that a firm may never have asked the question, and a capability that lives unnamed in one person is not one the firm can rely on as it grows.

Why does the word controller matter? Call it whatever you like.

We agree, and we have no attachment to the word. We use it because the fractional controller market exists and gives firms a reference point for scope and cost. The argument here is about the functions, not the title.

Our clients are happy. What are we actually missing?

Possibly nothing today. The gap tends to show up as surprises on calls, questions the client acted on without asking, and a partner who is the only person who knows what was agreed with each client. If none of those describe your firm, the work is being held somewhere, and it is worth knowing where.

Is this a hiring recommendation?

No. It is a recommendation to decide, deliberately, who holds this work. A hire is one answer among several.

About the author

Nilanjan Raychaudhuri is the founder of AgentLink (Tublian LLC, Columbus, Ohio), which builds controller-layer software sold through CPA and fractional CFO firms. He has spent the past year interviewing practitioners at fractional CPA and CFO practices about how advisory work is actually delivered. Team page

The practitioner observations in this guide come from design-partner conversations and are used with permission where attributed. Firm staffing and workflow are as practitioners described them; we did not audit any firm directly. Corrections to nilanjan@agentlink.finance.