Since the Zurich release, a ServiceNow administrator can configure which dimensions of a project status report the AI predicts, and with what logic. The feature is generally available, it is called AI Status Reports, and it produces a red, amber or green verdict per dimension without a project manager typing one.
It is a genuinely useful piece of engineering, and it changes nothing about the question a steering committee is asking. A predicted color is a model’s opinion about the kind of opinion a person used to give. The inputs are the same task metadata somebody was already guessing from; what has been removed is the guessing in public.
That is the whole disagreement: ServiceNow automated the production of the verdict, and we think the verdict was the wrong artifact.
What ServiceNow has that nobody else does
The platform, and it is not a small thing. Demand arrives through an idea portal and a workbench, scored by a published formula over stakeholder questionnaires — size plus value plus ten-minus-risk, divided by three (their own explainer) — and promotes into projects, enhancements, changes or defects. Benefit plans carry actuals at breakdown level. Resource plans carry soft and hard allocations. The Yokohama release added a project task monitor agent that watches the critical path. And all of it shares an identity, a change record and a configuration database with the service management estate an enterprise is probably already running.
If the organization lives on ServiceNow, that adjacency is worth more than any feature comparison, and an honest competitor says so early rather than late.
Two things a buyer finds in month four
The first is a license boundary that is also an architecture boundary. Portfolio Planning sits in one tier and Strategic Planning in another, and the two workspaces are split by license. Underneath, the funding data models are not interchangeable: a customer picks Investment Funding or the planning workspace’s scenario financials, and the choice is structural rather than a preference (the financial planning quick start). That is the kind of fact that reshapes an implementation plan, and it is discoverable before signature if somebody reads for it.
The second is that ServiceNow’s own capacity guide is candid about its edges: no batch approval, availability consumed before approval, and group assignments that cannot be tuned per person inside the workspace (their guide, in their words). We cite it not as a gotcha but as the standard we hold ourselves to — this product’s security page leads with what it lacks for the same reason.
Why a predicted color cannot be signed
A model can be right about base rates. Trained across a large estate, it will know that projects with this shape of task history usually slip, and that is useful information for a portfolio director deciding where to look first.
It is not the artifact a governance process needs. A gate closure has to be an act by a named person, against a stated body of evidence, at a known version — and a probability is none of those. Nobody signs a seventy-percent chance of amber. Regulators have been converging on the same shape for a while: the EU AI Act asks for natural persons with the competence and authority to oversee, and the Federal Reserve’s supervisory guidance asks for effective challenge by someone identifiable. A prediction has no one to challenge.
What replaces it here is duller and more useful. Counts, beside their denominators, frozen at the moment an update is posted — nine of thirty-eight bindings accepted this week, not 24% — with the drift against today printed on the record afterward. Gate standing per initiative with the signer named. And the acceptance interval, which is the number a flattened organization actually needs: how long a proposal waited for a human, because in a four-person pod the review queue is the constraint rather than the build.
The pricing difference, stated plainly
ServiceNow’s AI for portfolio work is generally available and license-gated behind Now Assist and the higher SPM tier. Ours is not sold at all: there is no AI product code, no credit meter and no tier in which the governance layer becomes available. Agent sessions, the sync doors, provenance and the append-only trail are in every paid plan, because a trust layer that arrives with an upgrade is not a trust layer — it is a feature, and features get switched off in a budget round.
Choose ServiceNow SPM if…
Choose ServiceNow SPM if the estate already runs on ServiceNow. Sharing identity, change records and the configuration database with the service management platform removes an entire category of integration work, and no standalone portfolio product — including this one — can offer you that. If IT service management, change and the portfolio all want to be the same record, the answer is the platform you already own.
Choose it also if demand intake through a scored idea portal is the process you are trying to fix. That workflow is mature, documented and in production at scale.
Come here instead if the thing that has stopped working is the status report itself — if the delivery beneath the portfolio is increasingly produced by agents, and what the review needs is not a better guess at the color but a count, a commit and a name.