Industry Insight
Continuous close. Old ambition, persistent gap.
In 2022, Gartner reported that 55% of finance functions were aiming for a touchless close — by 2022. Four years on, the ambition is unchanged and the arrival keeps moving. This brief asks what is actually holding it in place.
The ambition is not new. What is striking is how stable it has been while the arrival date keeps moving.
In September 2022, Gartner reported that 55% of finance functions were aiming for a touchless close — by 2022. The target was already the present tense when it was published. Four years on, the same functions are still aiming.
The gap, in published figures
What continuous close actually means
The term is used loosely, and three adjacent terms get folded into it. Separating them matters, because they describe different things and are owned by different people.
Terminology
| Term | What it describes | Origin |
|---|---|---|
| Continuous accounting | Distributing close activity across the period rather than concentrating it at period end | Coined and popularised by BlackLine |
| Touchless close | A close requiring no manual intervention | Used by Gartner in its finance surveys |
| Autonomous finance | A broader state in which finance processes self-execute and self-correct | Gartner's umbrella term |
| Continuous close | Work distributed *and* continuously reconciled, so period end is a review rather than a construction | In general use; not owned by one vendor |
The useful distinction is not automation versus manual work. It is when the work happens. A close that is fully automated but still runs entirely in the five days after period end is a fast close, not a continuous one. The change continuous close describes is temporal, not technological.
Why the traditional close resists change
The month-end close is not slow because people are slow. It has a shape, and the shape is load-bearing.
- It is defined by a boundary. The period cut-off creates a hard edge, and much of the work exists to establish what belongs on which side of it.
- Its dependencies are sequential. Sub-ledgers close before the general ledger; reconciliations precede adjustments; adjustments precede reporting. Sequence resists parallelisation.
- Its controls are period-shaped. Sign-offs, certifications and review thresholds are all defined against a period, not against a continuous flow.
- Its evidence is assembled late. Support is gathered when a reviewer asks for it, which means the reviewer's question is what triggers the work.
Automating any single step inside that shape makes the step faster without changing the shape. This is why so many teams report faster tooling and an unchanged close calendar — an observation we treat at length in why month-end close is slow.
Where continuous close breaks
1 · The exception queue does not distribute
Routine transactions distribute across the period easily. Exceptions do not, because resolving one usually requires a person, and people are not continuously available. A continuous process with a discontinuous escalation path re-concentrates at period end anyway.
2 · Controls stay period-shaped
You can reconcile daily and still be unable to certify daily, because the certification framework refers to a period. Until the control design changes, continuous execution sits underneath a periodic sign-off and the calendar does not move.
3 · The reason is not carried with the work
This is the failure we consider most consequential, and it is the subject of our research on AI adoption in finance. If a variance is investigated in week two and its explanation lives in a conversation, then at period end the explanation has to be rebuilt. Distributing the work without distributing the reasoning simply moves the reconstruction earlier.
4 · Upstream data lands late regardless
Bank files, intercompany confirmations, supplier statements and accruals from operational teams arrive on their own schedule. Continuous close is bounded by the least continuous input, and some of those inputs are outside finance's control entirely.
What resolves them — and what only appears to
Interventions
Appears to help
Faster close software applied to the same sequence
More automation of already-routine transactions
Additional dashboards over unchanged data
Headcount added to close week
Actually changes the shape
Redesigning controls so certification is possible outside a period boundary
Making exception resolution continuous, with routing that does not wait for close week
Capturing the reason alongside the entry, at the moment the judgement is made
Moving upstream dependencies earlier by agreement, not by escalation
The left column improves throughput inside the existing shape. Only the right column changes when the work happens.
The adoption gap
Prediction accuracy in this space has an interesting asymmetry. Deployment forecasts tend to land. Outcome forecasts tend to slip. Gartner's 2024 prediction that 90% of finance functions would deploy at least one AI-enabled solution by 2026 looks broadly right. Its 2022 finding that 55% were targeting a touchless close by 2022 plainly did not arrive.
In May 2026 Gartner put this more directly, warning that CFOs must stop mistaking finance AI deployment for value creation. That is the gap stated by the same organisation that measures it: tools get installed, outcomes do not follow.
Three explanations recur in the published commentary, and they are not mutually exclusive.
- No protected time. Finance teams are too busy closing to build the thing that would make closing easier. The work that would break the cycle is scheduled against the cycle itself.
- A planning-to-piloting gap. Around a quarter of finance organisations report being unsure how to make that specific leap — not unconvinced, unsure.
- A skills horizon. Gartner's expectation that 90% of finance talent will need digital skills by 2030 implies the constraint is partly a workforce transition, which moves on a slower clock than software.
What this does not establish
Evidence vs inference
Published and attributable
The touchless-close ambition was widespread by 2022 and has not been broadly realised
Deployment of AI in finance is proceeding; value realisation is lagging
A quarter of finance organisations report a planning-to-piloting gap
Our reading — not proven
That period-shaped controls are a primary constraint rather than a secondary one. We can find no study isolating this.
That un-carried reasoning is what re-concentrates work at period end. Consistent with our other research; untested.
That the four interventions on the right above are the ones that matter. Argued, not measured.
We would rather this were measured. A study of when close activity actually occurs, rather than how long it takes, would settle most of it.
One further caution. Gartner's public press releases are cited here and linked; its subscriber research is not reproduced. Vendor-published benchmarks in this category — and there are many — are directionally interesting and are not independent evidence.
What follows, if the argument holds
- Continuous close is a control-design problem before it is a software problem. Tooling applied to an unchanged control framework produces a faster periodic close.
- Measure when, not how long. Close duration is the wrong metric; the distribution of activity across the period is the one that reveals whether anything has changed.
- Distribute the reasoning, not only the work. Otherwise reconstruction moves earlier rather than disappearing.
Sources
- Gartner — Finance survey shows 55% of functions aiming for a touchless close by 2022 (Sept 2022)
- Gartner — Embedded AI in cloud ERP will drive a 30% faster financial close by 2028 (Feb 2026)
- Gartner — CFOs must stop mistaking finance AI deployment for value creation (May 2026)
- Gartner — 90% of finance functions will deploy at least one AI-enabled solution by 2026 (Sept 2024)
- Gartner — Finance AI adoption remains steady in 2025 (Nov 2025)
- CFO Dive — CFOs' AI adoption slows as challenges mount
- Five standout themes from Gartner Finance Symposium 2026