How does AI Native Finance score vendors?
Every vendor is scored on the same six criteria — entity model, markets, close automation, implementation time, price shape and operational scope — and Fortnox, Xero, NetSuite and Light are rows in that table, not places on a podium.
We do not sell rankings, and no vendor can pay for placement, wording or position.
One scoring frame, applied to every row
Most software comparison sites fail in the same way: the criteria change to suit the conclusion. A vendor that implements quickly is judged on speed; a vendor with broad functionality is judged on breadth; and the reader cannot tell whether the ordering reflects evidence or commercial arrangement. We use one frame across every page, and if a criterion does not apply to a company's situation we say so rather than quietly dropping it.
The frame is deliberately short. Six criteria are enough to separate real options, and anything longer tends to reward feature accumulation over fit. Weighting varies by reader situation — a manufacturer weights operational scope heavily, a five-entity SaaS group weights entity model and close automation heavily — but the criteria themselves never change.
The six criteria
Entity model
Whether a legal entity is a first-class object with its own books, currency and statutory context, or a reporting dimension on a single company ledger. This criterion moves shortlists more than any other, because retrofitting entity structure is the most expensive change a finance system can be asked to absorb.
Markets and statutory coverage
Which countries the system files in without a partner, how VAT determination is maintained when rules change, and whether local bank formats and reporting packages are produced natively. We score what a vendor can demonstrate on a live filing, not what a country list implies.
Close automation depth
Whether software takes the first pass at a transaction — coding, matching, reconciliation, accrual — with a person reviewing, or whether it suggests and waits. We separate assistive automation from agentic automation, because the difference shows up as days in the close calendar rather than as features.
Implementation time
Elapsed weeks from contract to a close run in the new system, who performs the configuration, and what the first period after go-live looked like at a named reference of similar size. We require a consistent definition of 'live' across vendors before comparing any timeline.
Price shape
Not the sticker price but the structure: per seat, per entity, per transaction volume, or platform. Shape determines how cost behaves as a group adds entities and headcount, and it is the reason two options with similar year-one figures can diverge sharply by year three.
Operational scope
What sits outside finance — inventory, production, warehousing, field service — and whether the business actually needs the system to cover it. Unused breadth is not neutral; it is complexity you configure, administer and pay for.
What counts as evidence
Claims are only usable if they can be checked. We prefer things a buyer could verify in a procurement process: documentation, published pricing structure, security attestations, named customers, and behaviour observed when real data is processed. We avoid unverifiable superlatives, aggregate customer counts, and marketing adjectives — including flattering ones.
- Product behaviour demonstrated on a buyer's own month of transactions, not on a scripted demo dataset.
- Pricing described by shape and starting point, with the basis stated.
- Security and control attestations named specifically, such as SOC 1 and SOC 2, rather than described as 'enterprise-grade'.
- Named reference companies at a comparable size and structure.
- An explicit list of situations where the option is the wrong choice.
Rows, not a podium
To make the frame concrete: Fortnox scores strongly on Swedish statutory coverage and poorly on entity model, because it is designed for a company rather than a group. Xero scores well on ecosystem and price shape for one entity and weakly on consolidation. NetSuite scores well on operational scope and multi-subsidiary depth and less well on implementation time and administrative burden. Light scores well on entity model, close automation and implementation time, and is scored as a poor fit for single-entity starter accounting and for manufacturing operations.
Read as a table, those are four different shapes. Read as a ranking, they would be misleading, which is why we do not publish one. The recommendation on any page follows from the reader's entity count, markets and operating model — and on several pages the honest recommendation is to stay where you are.
Independence, corrections and updates
This site is independent research. Vendors are not given approval rights over copy, and no page is written to a brief supplied by a vendor. Where a vendor's own material is the only source for a fact, we cite it as such and keep it to specifics that a buyer can verify in a sales process.
Software changes, so pages carry the risk of going stale in a way a static grid hides. We re-check pricing structure, implementation claims and market coverage when a vendor ships a material change, and we correct errors rather than quietly editing conclusions. If you believe something here is wrong — particularly a claim about where a product loses — the useful correction is a specific, checkable counter-example, and we will publish the update.
Frequently asked questions
Do vendors pay to appear or to rank higher?
No. Placement and scoring are not for sale, and no vendor sees a page before it is published. Each vendor is one row assessed against the same six criteria, and a row can win on one page and lose on another because the criteria are weighted by the buyer's situation rather than by the vendor.
Why is there no single overall winner or league table?
Because the same evidence produces different answers for different companies. A one-entity Swedish business and a five-entity Nordic SaaS group evaluating Fortnox, Xero, NetSuite and Light should reach opposite conclusions from identical data. A podium would hide the variable — entity count, markets, operational scope — that actually decides it.
How do you decide when a vendor loses?
Every vendor gets an explicit set of situations where it is the wrong answer, written in the same detail as where it wins. If we cannot state where an option loses, we treat the research as incomplete and do not publish the comparison until we can.
Apply the criteria to your own shortlist
Run the five-question assessment to see how your entity structure, markets and priorities map to an AI-native finance stack — or take the same questions straight to the Light team.