Recurrence alone does not make revenue software revenue; what matters is what obligation sits behind it.
Why it matters when the plan changes
The distinction decides how a business is valued and how it behaves. Software revenue scales without proportionate effort; services revenue does not, and a capacity commitment for expert work carries obligations closer to a statement of work than to a licence. A company reporting services as software looks like a different business from the one it is, and it will make investment decisions on that misreading before any investor does. Keeping software MRR and ARR distinct from services is what lets anyone see whether effort per comparable delivery is actually falling.
The tension is that the honest version of this business is a mix, three lines rather than one: software access, a capacity commitment and one-off delivery. Expert judgement is architecture rather than overhead where it contributes context a model cannot see, so services are part of the product rather than a transitional cost. Saying that plainly is harder than presenting a pure software story and more durable.
In practice
A recurring annual commitment covers platform access and a capacity level for expert-assisted work. Reported as one recurring line it looks like software. Classified properly it is two obligations with different economics, and the split is what tells anyone whether the delivery cost per piece of work is actually falling.
Evidence
A statement of work defines a bounded piece of service work, which is a different obligation from recurring access.
Statement of work, Wikipedia (2026)A framework agreement sets terms for subsequent orders, which may cover either kind of obligation.
Framework agreement, Wikipedia (2026)
What it cannot tell you
The distinction cannot tell you whether the mix is healthy, only whether it is described honestly. It is silent on whether the services component is priced well or whether effort per delivery is actually falling. A business can classify correctly and still be inefficient, or misclassify and still be commercially sound.
Questions
Because the two carry different obligations. A statement of work, as described in the Wikipedia entry from 2026, defines a bounded piece of work rather than recurring access, and a framework agreement can cover either kind. Reporting services as software misreads which obligation actually sits behind the revenue.
No. Recurrence alone does not make revenue software revenue; what matters is the obligation behind it. A committed annual amount of expert-delivered work recurs and is a services obligation, and classifying it correctly is finance's responsibility rather than a presentational choice.
Not where the expert contributes context a model cannot see. Evidence on combining people and models shows the pairing only outperforms under specific conditions, one of which is the human holding private information. That makes the expert layer architecture rather than overhead.
What they are actually committing to and what they get for it: recurring access to what, capacity for how much defined work, and what implementation covers. A single blended figure makes it hard for a buyer to judge what happens if their usage changes.
The aim is that effort per comparable delivery falls as work is productised, shifting the mix without eliminating the expert layer. A framework agreement, as defined in its 2026 entry, sets terms for subsequent orders and can cover either obligation, which is why tracking the split matters as scope changes over time.