Hiring · August 3, 2026 · 9 min read

How to hire an accountant: skills to test, not claims

How to hire an accountant in the AI era: what the role really owns, the skills that separate a real one from a bookkeeping operator, and how to test them.

← Part of The five pillars of hiring: what assessments measure

This guide is for founders, finance leaders, and hiring managers working out how to hire an accountant — and it starts from an uncomfortable truth: the résumé is worse than useless here. Accounting attracts tidy CVs. Everyone lists the same software, the same certifications, the same "managed the full month-end close." On paper a bookkeeping-software operator and a genuine accountant look identical. The difference only shows up under pressure — when a number is confidently wrong, a deadline is real, and someone has to decide whether the figure that just landed in the board pack can be trusted. That judgement is the whole job now, and it is exactly what a CV cannot show you. Hire the wrong one and the failure is quiet: a misstated accrual, a reconciliation that was never really done, an error that surfaces months later in an audit or a fundraise. The stakes are high precisely because the mistakes are invisible until they are expensive.

What does an accountant actually do?

An accountant turns raw transactions into numbers a business can trust and act on — reconciling accounts, applying the reporting standard, owning the close, and safeguarding the controls that keep the figures honest. That is the capsule; here is the week. It is not data entry, and it has not been for years.

• Owns the month-end close — reconciles accounts, posts accruals and prepayments, and produces management figures that leadership can actually rely on, on a fixed deadline that does not move.
• Investigates the exceptions — the balance that does not tie out, the expense in the wrong period, the vendor payment that appears twice — and works out the cause rather than forcing the number.
• Applies the reporting standard correctly — revenue recognition, capitalisation, provisions — and knows when a judgement call needs documenting rather than guessing.
• Maintains controls — approval limits, segregation of duties, an audit trail — so errors and fraud are caught before they compound.
• Reviews the machine's output — AI now categorises transactions and drafts reconciliations, so much of the day is checking, correcting, and handling what the automation got wrong or flagged as uncertain.
• Explains the numbers to non-finance colleagues — why margin moved, what a variance means, whether a cost is one-off or structural — in plain language a founder can act on.

Do you actually need to hire an accountant yet?

The honest answer is often not yet. A full-time accountant is justified when your reporting, transaction volume, or external scrutiny outgrows what a bookkeeper plus software can safely carry — around fundraising, fast growth, or a real audit and compliance burden. Below that threshold, hiring one is expensive idle capacity. Here is who should hold off.

• Early-stage companies with low transaction volume and no audit obligation — a competent bookkeeper plus a fractional or outsourced accountant usually covers you, and costs far less.
• Businesses where the real need is a controller or finance leader — someone to build the function and own strategy — not a hands-on accountant to run the close. Name the gap before you hire for it.
• Teams reaching for a full-time hire to fix a messy ledger — clean the mess first, or you will simply pay a skilled accountant to do a bookkeeper's remediation for months.

Hiring too early buys capacity you cannot keep busy; hiring too late buys errors that reach the board before anyone qualified catches them. The cost of a bad hire in finance is unusually steep because the damage is silent and compounding — a wrong number that everyone downstream trusts. Get the timing right, then get the assessment right.

What skills separate a strong accountant from a plausible-sounding one?

The impostor in this field is specific and common: the bookkeeping-software operator who can navigate the ledger fluently but cannot tell you what the entries mean. They will describe the close in perfect sequence and never mention a single thing that went wrong in it — because they were operating a process, not owning it. The real accountant is defined by judgement over mechanics. Here is what actually separates the two.

• Understanding, not operating — can they explain why an entry sits where it does, not just where the software puts it? The operator recites steps; the accountant reasons about the accounting.
• Exception-handling — when a reconciliation will not tie, do they investigate the cause or plug the difference to a suspense account and move on? This single instinct separates the two.
• Controls awareness — do they think about who approves what, where fraud could enter, and what the audit trail proves? A strong accountant designs safety into the process, not just outputs.
• Accuracy under a real deadline — the close does not wait. Can they stay precise when the pressure is on, or does quality collapse the moment the clock does?
• Communication — can they translate a variance into a decision for someone who has never read a balance sheet? Numbers nobody understands are numbers nobody uses.

The AI-era skills layer

The job moved up a level, and your assessment has to move with it. AI categorises transactions, drafts reconciliations, and writes first-pass commentary in seconds — usually well, sometimes confidently wrong. The accountant's value is no longer producing the number; it is reviewing it. The skill you are now hiring for is spotting the plausible-but-wrong figure before it reaches a report, knowing which parts of the close to delegate to automation and which demand a human judgement call, and keeping ownership of the controls even when a tool did the work. There is also a discretion layer: an accountant handling payroll, forecasts, or acquisition numbers must guard confidentiality when using AI tools on sensitive material. This is the AI fluency signal that now runs through every knowledge role — the difference between someone who trusts the output and someone who verifies it.

How do you test those skills in a real assessment?

Stop testing what candidates say and start testing what they do. Claims about the close are not verifiable; accuracy under a deadline is. The most predictive step for this role, by a wide margin, is a short, job-shaped work sample — the same logic behind work-sample tests generally, applied to finance. Give every candidate the same realistic exercise on the same terms: reconcile an account that contains a deliberately planted discrepancy, or hand them a batch of AI-categorised transactions and a draft reconciliation and ask them to review it — which entries are miscoded, which number is confidently wrong, what would they refuse to sign off. You are not watching for speed. You are watching whether they investigate the exception or paper over it, whether they question the machine or trust it, and whether they can explain their reasoning cleanly afterwards.

Pair the work sample with focused capability checks. A structured spreadsheet exercise shows whether they can actually build and audit a reconciliation rather than merely open one — see what an Excel test measures. And because tiny discrepancies are where finance errors hide, an attention-to-detail test is a fair, fast filter for the precision the close demands. A numerical reasoning exercise adds a further layer if the role leans analytical, though for most accountant hires the reconciliation work sample plus attention to detail carries the signal. Run these inside a realistic environment — an AI Sandbox where the automation the job actually uses is present — rather than banning AI and pretending the modern close does not involve it. This is the practical face of skills-based hiring: assess the work itself, on equal terms, and let evidence rather than pedigree decide.

What does the interview loop look like?

Once the work sample has done the heavy lifting, the interview verifies judgement — and it has to be structured, or it becomes a comfort test that the polished operator wins. Same questions, same order, same rubric for every candidate, scored independently against defined anchors; structured interviews beat gut-feel conversations on both accuracy and fairness, which matters when everyone in the pipeline sounds equally competent. A workable loop for an accountant looks like this.

• A short screen to confirm the basics and set expectations — held to the same questions for everyone, not a free-form chat.
• The work sample — the reconciliation-and-review exercise above, done before the deep interview so you are discussing real evidence rather than hypotheticals.
• A structured interview with the finance lead — walking a specific month-end close the candidate owned: what broke, what they investigated, what they changed, and how they knew the final numbers were right.
• A cross-functional conversation — a founder or operations lead — testing whether they can explain a variance to someone non-financial and whether they will push back when a number is being pressured into a shape it should not take.

Give every interviewer the same interview scorecard so you capture evidence against each skill rather than a general impression. The single most revealing question is the variance walk: ask them to describe a variance they investigated and what they actually changed as a result. The operator narrates the report. The accountant tells you what did not make sense, how they chased it, and what the business did differently once they explained it. Keep the whole loop tight — strong finance candidates are courted, and a slow, unstructured process both loses them and signals that rigour is not how you operate.

Seniority, and the first 90 days

Think in qualitative bands rather than titles. A junior accountant executes the close under supervision and grows into ownership. A senior one owns it outright, designs the controls, handles the judgement calls the standard leaves open, and is the person the auditors talk to. The AI-era shift raises the floor: because automation now handles more of the mechanical recording, even junior roles are judged sooner on review and exception-handling than on data entry. Hire for the level of judgement the role genuinely requires, and if what you actually need is someone to build the function rather than run it, you may be looking for a controller — a different hire, and worth naming honestly before you advertise. A clear, skills-anchored job description forces that distinction into the open.

In the first 90 days, a strong accountant owns a clean, on-time close without heroics; surfaces at least one control gap or recurring error the previous process missed; builds a trust relationship with the numbers so leadership stops second-guessing the management accounts; and forms a clear, defensible view on where automation helps and where it must not be trusted unsupervised. What you should not see is someone who runs the process flawlessly and never questions a single figure inside it — that is the operator revealing themselves after the offer, which is precisely the outcome a job-shaped assessment exists to prevent.

The software can categorise the transaction and draft the reconciliation. What it cannot do is feel that a number is wrong before anyone can prove it — and then go and prove it. Hire for that instinct, test for that instinct, and you have hired an accountant rather than an operator of one.