Offering controller services without hiring a controller

Controller work rarely arrives at a firm as a request for controller work. It arrives as a question. Can we afford the second location. Will cash hold through the slow quarter. Should we take the bigger contract if it pays in ninety days. The client is asking for someone to know the business well enough, and recently enough, to answer.

Most firms can answer on the day. What they cannot do is be in a position to answer every day, for every client who is starting to ask. That is the controller role, and it is the part a firm struggles to staff.

This piece is about that staffing problem. What the role actually involves, why the obvious hire is hard to justify early, and three ways firms deliver the work without making it. For how to scope and price the service once you can deliver it, see packaging and pricing controller services.

The hire comes before the book that would pay for it

The straightforward answer is to hire a controller. The problem is sequence.

A controller is a full salary. The US Bureau of Labor Statistics puts the median annual wage for financial managers, the category that includes controllers, at $166,570 as of May 2025. Recruiting takes time as well: Personiv’s Spring 2025 CFO Pulse puts the time to fill an accounting role at more than 60 days, and the Controllers Council’s 2026 talent study found 61 percent of finance leaders reporting shortages, up from 46 percent.

The firm is committing that salary to clients who each need a fraction of one person. A hire makes sense once there is a book of controller clients to carry it. But the book does not form until the firm can deliver the work, and it cannot deliver the work without the capacity. So the service stays informal. The partner answers the big questions when they come up, and the work between those questions does not happen.

That is a structural problem, not a commitment problem. The firm is not choosing to be reactive. The staffing model gives it no other option until the book is large enough.

What the controller role actually is

Before choosing how to staff the role, it helps to be precise about what it contains. It is easy to treat “controller” as one job. It is really two kinds of work, and the second kind has two directions. Where the line sits against bookkeeping is covered in what a controller does that a bookkeeper does not.

Judgement

Deciding what a number means for this business. Telling the owner the second location should wait. Choosing which of three problems to raise first. Knowing when a variance matters and when it is noise.

This has to stay with a person the client trusts, and in most firms it already does. The partner or senior manager who owns the relationship supplies it. Judgement is rarely what is missing.

Attention

Attention is the standing work that makes judgement possible. It is what the controller does between the conversations, so that when the question arrives there is something current to judge.

It looks in two directions.

Backward: what happened and what is happening. Checking what the books say against what the bank did. Noticing that a customer who always pays in 30 days is now at 55. Seeing that payroll moved, a vendor changed terms, or margin on one line slipped. Following up on what the owner said they would do last month and whether they did it.

Forward: what comes next. Carrying all of that into a forecast. Keeping the forecast current as things change, not only when the next meeting is booked. And grading it: when the quarter lands, checking what the forecast said against what actually happened, so the next one is better and the firm knows how far to trust it. We cover the grading step in learning from forecasts.

Why watching alone is not the role

It is tempting to solve only the backward half, because it is the easier half to picture. Set up monitoring, get an alert when cash drops below a line, and call that controller coverage.

But an alert answers a different question from the one the client asked. The alert says cash fell. The client wants to know whether they can make payroll in six weeks and still hire in the spring. That answer needs the forward half: a forecast that already reflects what the alert just showed, and a track record that tells the firm whether the forecast is good enough to stand behind.

The reverse is also true. A forecast built fresh before each meeting, with nobody watching in between, is a snapshot. It is accurate on the day it is built and silently wrong from then on. The controller role is the connection between the two halves: what the backward view notices feeds the forward view, and what the forward view expected gets checked against what the backward view later sees.

That connection is what clients are paying for when they ask for a controller, even if they would not describe it that way. It is also what makes the role hard to staff, because attention scales with clients and calendar days, not with how good the people are.

Three ways to staff it without the hire

Firms that deliver controller work before they can justify the hire tend to use one of three models, or a mix. Each is a reasonable choice. Each breaks somewhere specific.

1. Stretch an existing senior

The most common starting point. A senior manager or the partner takes on controller scope for a handful of clients alongside their current work.

What it gets right. Judgement stays inside the firm, with someone who already knows the client. No new cost, no recruiting delay, and the firm learns what the service involves before committing to it.

Where it breaks. Attention. The senior’s existing work has deadlines; the controller work mostly does not. So the watching happens in bursts, usually in the day or two before a client call, and the forecast gets rebuilt rather than maintained. The hours run out quickly. As an illustration, say a senior can free 40 hours a month from existing work, and each controller client needs about 12 hours a month. That covers roughly three clients, before any time for coordination, review or the client who has a bad month. Consulting Success reports 10 to 15 hours a month per client as a common level for fractional engagements generally; it is a reference point for this kind of work, not a measured figure for controllers inside a firm. We work through the arithmetic in more detail in how many clients a fractional controller can carry.

This model works for a pilot. It does not grow past a few clients without eroding the senior’s other work or the quality of the attention.

2. Outsource the production

The firm subcontracts the work to an outsourced accounting provider or an independent fractional controller, often under the firm’s own brand. The firm reviews the output and owns the client conversation.

What it gets right. Capacity without a salary, and it scales up and down with demand. For the backward half, especially reconciliation and reporting, it can work well, because that work is well defined and easy to review.

Where it breaks. The handoff. The forward half depends on what was said in the client conversation: the contract that is about to be signed, the hire the owner mentioned, the customer who is wobbling. A capable outsourced controller can forecast well, but only from what reaches them. When the provider receives only the accounting records, someone at the firm has to carry decisions and plans across the boundary, or the forecast reflects the books and not the plans. Bringing the provider into client conversations solves that, but it costs coordination time on both sides, and that time grows with every client added.

There is also a relationship question worth asking early: who the client talks to when something looks wrong, and whether the provider ever deals with the client directly.

3. Supply the attention with software, keep the judgement with your people

The firm keeps judgement with the partner or senior who owns the client, and uses software to carry the standing attention across the book.

What it gets right. Attention is the part that scales with clients and days, and it is the part software is suited to. It does not get squeezed by a filing deadline, and it covers every client every day rather than whoever has a call this week. The person with judgement starts each conversation from a current picture instead of building one.

Where it breaks. Coverage of the role. Most tools cover one direction. Monitoring and alerting tools cover the backward half. Forecasting and planning tools cover the forward half. Few connect them, and fewer grade their own forecasts against what happened. A firm that buys only the backward half has bought an alerting system and still cannot answer the question the client asked. A firm that buys only the forward half has a forecast nobody is watching.

There is also a data limitation that no tool escapes. At any moment, some of a client’s books will be unreconciled, some transactions uncategorised, and some commitments, a signed contract or an agreed payment plan, not yet recorded anywhere the software can see. Software connected to the bank sees money move before the accounting catches up. A tool worth trusting keeps those apart: it distinguishes what the bank shows from what has been reconciled, and it surfaces what it is unsure of rather than presenting one confident number. A firm evaluating tools should ask how each one handles that gap, because it is where a forecast quietly goes wrong.

Software also cannot supply judgement, and should not be sold to a client as if it does. It supplies attention. The judgement still has to come from someone at the firm, which is also why the client stays the firm’s client.

Comparing the three

Many firms combine them. A senior owns judgement for a handful of clients, an outsourced team handles reconciliation, and software carries the watching and the forecast in between. The useful question is not which model is best, but who is carrying each part of the role for each client, and whether any part is carried by nobody.

What stays with the firm, whichever model you use

Three things do not move, regardless of how the attention is supplied.

The client relationship. The client should be talking to the firm, seeing the firm’s name, and hearing the firm’s view. If any part of the delivery model puts a third party in front of the client, that should be a deliberate choice.

The judgement. What the numbers mean, what to recommend, and what to raise first. This is the work clients value most and the work that justifies the fee.

The pricing. How controller scope is packaged, tiered and priced is a firm decision. Our guide to packaging and pricing controller services covers it, and what a firm should charge for controller work covers the pricing logic in more detail.

How to start without the hire

A practical sequence that avoids committing to a model before the firm knows what the work involves.

Pick three clients who are already asking forward questions. The ones who ask whether they can afford something, not only what happened last month. They are already buying controller work informally.

Write down who carries each part of the role for each of them today. Backward attention, forward attention, judgement. Be honest about the parts that happen only in the week before a call, and the parts nobody carries.

Agree a bounded pilot. Set the scope, the length (a quarter is a natural unit) and the commercial terms with each client before starting. The point is to prove delivery before scaling it, not to deliver extra work for free. Staff it with a senior, a provider, software or a mix, and test whether the firm can answer the client’s forward question between calls without assembling it from scratch.

Grade the forecast as results arrive. Each time a payment lands, a month closes or an expected event happens or does not, compare it with what the forecast said. At quarter end, review the accumulated record. That record is what lets the firm stand behind the next answer, and it is the evidence that the service is real rather than a relabelled monthly package.

Use the pilot to refine the recurring package. By the end, the firm knows what the work takes, who does it and what it costs to deliver, which is what the standing price needs.

The goal of the pilot is not a new service line on the website. It is knowing that every part of the role has an owner for those three clients. Once that is true, the question of whether and when to hire becomes a question about volume, not about whether the firm can offer the service at all.

Where AgentLink fits

AgentLink is an AI controller for firm clients, built for the third model. It carries the standing attention across a firm’s client book in both directions: looking back at what the books and bank show and what changed, and looking ahead with a forecast that is kept current and graded against what actually lands. The firm’s people keep the judgement and the client conversation. AgentLink is sold only through CPA and fractional CFO firms, and the firm keeps the client, the pricing and the judgement.

Frequently asked questions

Can a CPA firm offer controller services without hiring a controller?

Yes, and many start that way. The work can be delivered by stretching an existing senior, outsourcing production to a provider, or using software for the standing attention while firm staff supply judgement. Each works at a different scale and breaks in a different place, so most firms combine them.

What does a controller do that monitoring software does not?

Monitoring covers the backward half of the role: noticing what changed. A controller also carries that forward into a forecast, keeps it current, grades it against what happens, and applies judgement about what it means. Monitoring alone tells the firm something moved; it does not answer whether the client can afford what they are planning.

Is outsourcing controller work to a provider a good option for a small firm?

It can be, especially for reconciliation and reporting, which are well defined and easy to review. Forward-looking work is also possible with a capable provider, but it depends on what the client said. If the provider sees only the accounting records, someone at the firm has to pass plans and decisions across; if the provider joins client conversations, that costs coordination time. Either way, budget for the handoff.

How many controller clients can one person carry?

It depends on how many hours the person actually has and how much each client needs. A dedicated person can carry many more than a senior fitting the work around existing responsibilities. As an illustration, a senior with 40 free hours a month and clients needing about 12 hours each can cover roughly three, before coordination and overhead. The limit is attention, not skill.

When should a firm actually hire a controller?

When there is a book of controller clients large enough to carry the salary, and the firm has run the service long enough to know what it takes. Delivering the work first, through one of the models above, is how a firm gets to that point without committing a salary to demand it has not yet proven.

Our guide on fractional controller client capacity walks through the numbers behind that last answer.

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

Corrections: if we have described something wrongly, email nilanjan@agentlink.finance and we will fix it and log the change. This piece draws on AgentLink’s conversations with design-partner firms; AgentLink sells the software described in the final section.

Update log

• 2026-09-23: First published.