Quite unlike any earlier technology, the integration of AI in finance is progressing at a breakneck speed. Foundational models are being upgraded faster than ever – and this is especially significant because a new version doesn’t just represent a marginal improvement, but a step change in capability (case in point: Anthropic’s newest model Mythos is so powerful that the company took the decision to defer the release to the general public). The finance and accounting space has witnessed rapid AI advancements:
- The incumbent angle: KPMG announced its plans to remove humans from parts of the audit process as early as summer of 2026, with full implementation planned in 2027; today, all big 4s have their own AI-infused proprietary platforms.
- The insurgent angle: Several startups operating applications have zeroed in on accounting and auditing as verticals ripe for disruption – notably, some startups in the space have already emerged as unicorns with valuations exceeding $1 billion.
What AI in Finance can and cannot do
AI will accomplish parts of the finance and accounting workflow – arguably better than humans. AI can match or outperform human capabilities on both speed and accuracy for predictable, high volume and pattern-based workflows. Some examples of workflows where AI does well today:
- Bank reconciliations or three-way reconciliations
- Invoice processing – the combination of OCR and LLMs can easily address edge cases such as non-standard formats, handwritten invoices, multi-language invoices.
- Anomaly detection – one thing that AI is (and will continue to be) better than humans at is deducing and recognising patterns from large amounts of data.
- Transaction categorization
- Variance commentary
However, parts of the workflow will, undoubtedly, require a human-in-the-loop. These will involve tasks that require judgement or carry accountability for the person executing them or require context that is not explicitly captured in data. Some illustrations of transactions where AI would not do well without a human-in-the-loop:
- Subjective calls on gray areas such as whether to capitalise or expense a given transaction, when to recognise revenue, whether to adopt a conservative or aggressive stance in recognising revenue for a period
- Treatment of unreconciled or unmatched transactions
- Categorisation of a new one-time transaction that hasn’t occurred before
- Application of a new accounting standard
Human-in-the-loop is here to stay
Given the regulations in finance and accounting, even an AI optimist will agree that the human-in-the-loop will not go away anytime soon and is here to stay for the foreseeable future. Financial statements cannot be signed by AI simply because AI cannot assume accountability. As best-in-class finance and accounting applications will continue to be built on top of foundational models with advancing capabilities, we could see a reduction in human-in-the-loop operations over a period of time, but unlikely fully agentic workflows in the near future.
Implications of a bundled solution
The near-term best-in-class solution will then be a bundle – consisting of an AI platform and human-in-the loop operations. These solutions will be built around a tight interface between AI and humans – with intelligence that knows when to route a transaction to AI, when to escalate it to a human and how to seamlessly pass context between the two.
A common mistake is to think of the human-in-the-loop as a band aid on an imperfect AI. It is a strategic part of the bundle and not less important than the AI platform. Every correction that an accountant makes isn’t just fixing an error for the client, it is making the underlying AI better. The humans provide continuous feedback to the AI, and this feedback loop improves the quality of the underlying AI product.
Let us dive deeper into a few important implications if finance and accounting workflows will indeed be bundled.
Know your provider’s AI playbook
Cost will continue to be a differentiator for human-in-the-loop operations. And here is where cost-competitive outsourced service providers such as SBS Global will continue to have an edge. It is likely that incumbent outsourced service providers enter into partnerships with AI applications companies to leverage the product layer of such companies to bring a bundled solution to their customers. As a finance professional, you should be sure to ask your outsourced service provider about their plans to infuse AI into their operations – whether through proprietary builds or partnerships. SBS Global has established strategic partnerships with leading AI application providers to deliver standardized, cost-effective AI solutions, leveraging organizational scale and volume to benefit its customers.
A bias for action, not perfection
Organisations should now look to move from experimentation to production. The need of the hour is to infuse AI into operations; there is no time to lose in waiting for the perfect technical solution to come along. The companies making meaningful strides are the ones that are deploying available products into workflow, learning from them and iterating.
Shorter turnaround cycles
The kind of work done by the human remaining in the loop will look fundamentally different. A junior accountant who spent 6 hours reconciling general ledger accounts will now spend 30 minutes reviewing AI-flagged unreconciled transactions and 2 hours creating cash flow simulations for the CFO ahead of a Board Meeting. Simply put, the repetitive, mechanical work shrinks and the thinking work requiring judgement expands. This reset will have visible follow-on effects for customers. Audit cycles that required clients to put together data months in advance could now compress to weeks. Traditional pricing models will come to be challenged, and outcome-based contracts could emerge as the new standard.
A faster shift to outcome-based contracts
An implicit and often understated shift being ushered in by the new AI world is the shift to outcome-based contracts. There are two forces driving this movement. The first is that solutions are now bundles. The traditional pricing model of “X hours of a senior accountant” isn’t meaningful when the client is buying an outcome produced by a mix of a product and humans. Second, the market is already moving. Forward deployed engineers (FDEs) from leading frontier labs that deploy AI products at enterprises are structuring agreements around outcomes only. While this shift has already been unfolding in the outsourcing space, AI will expedite it.
Outcome-based pricing is a core element of SBS Global’s operating model. As finance and technology services increasingly integrate automation platforms, AI tools and skilled talent, clients are shifting focus from effort-based metrics to the quality, timeliness, and reliability of deliverables. Accordingly, SBS Global is progressively adopting pricing structures aligned to clearly defined outcomes—such as timely month-end close, accurate financial reporting, compliant payroll processing, and successful system implementations. This approach strengthens alignment with client objectives and reflects the evolving delivery landscape, where technology and human expertise combine to drive measurable business outcomes.
How AI will replace Parts of the Accountant’s Work in Finance
Is your accountant being replaced by AI? Parts of the work, yes; the role, no. The more pertinent question for finance leaders is how AI will reshape finance. The winners in this transition won’t be the companies with the most sophisticated AI, nor the ones with the biggest finance teams. The winners will be the ones who pick the right bundle of AI and humans working as a single system and the right partner to deliver it. Start now—deploy what works today, ask hard questions of your providers, and price for outcomes and not hours.