Two debugging features, two very different access stories
The Brazil release brings two additions for debugging flows, and only one of them is something most ServiceNow developers will get their hands on this month. Input Modification ships with Brazil and lands on personal developer instances on September 24. The Troubleshooting Agent requires Otto and Now Assist for Platform, and ServiceNow says it is not yet available on personal developer instances.
That gap matters more than the feature list. A developer who cannot run something cannot vouch for it, and the people asked whether a new capability earns its money are often the same people debugging flows on a personal instance after hours. Both features point at the same target, which a September 2026 Developer Passport post frames as winning back time otherwise lost to manual digging and cutting down escalations to support.
Both sit inside ServiceNow's workflow automation surface, and the detail comes from the Developer Passport post announcing them. That post discloses no licensing cost, so any discussion of what the agent is worth starts without a number.
Changing an input mid-run changes how you reproduce a bug
ServiceNow describes Input Modification as 'an enhancement to the Flow Debugger, breakpoints and step-through shipped in earlier releases, and it lives in Flow Designer's new UI, built on Lux, which you opt into.' The debugger is not new. Altering the values going into a step while the run sits paused is the new part.
Reading what happened is the easy half of flow debugging. The hard part is getting the instance back into the state that produced the failure. A bad value arrives from an integration you do not control, a lookup returns a user who left in March, a date lands a day early because something upstream converted a time zone. Reproducing any of it means manufacturing a record that carries the same defect, and that is where the afternoon goes.
Modifying the input during a paused run removes that setup work. You stop on the step before the failure, put the offending value in by hand, and continue. Testing fifteen variations of a condition turns into fifteen resumes rather than fifteen data builds. For flows that branch on fields other teams own, that is the difference between testing the branch and testing whether you faked the data correctly.
The agent's root cause is a hypothesis, its links are checkable
On the other feature, the post says: 'When a flow breaks or misbehaves, the Troubleshooting Agent investigates on its own, pulls together a flow summary and a root cause, and hands you next steps in plain language, with links to the specific records and metadata it flagged.'
Read that sentence in two halves. A generated flow summary and root cause are conclusions drawn from execution data, carrying about the weight of a capable colleague's first guess before anyone opens a record. Sometimes right, occasionally confident and wrong, and least dependable on the failures that cost you a weekend, the ones with two contributing causes rather than one.
The links to the specific records and metadata it flagged are a different kind of output, because you can open them and check. A record either holds the value the agent says it holds or it does not. Our read is that those links carry most of the value here, and the root cause is best used to decide which record you open first. Cutting escalations is the stated goal, and a wrong diagnosis pasted into a ticket travels fast, so the developer reading the summary stays the control on it.
Otto and Now Assist decide who gets to try this
The Troubleshooting Agent requires Otto and Now Assist for Platform, the same platform capability we wrote about when Otto turned up for Creator. Anyone who has already had the Now Assist conversation with their account team knows roughly where that discussion goes.
The absence from personal developer instances has a practical consequence for consultants. If you are asked whether the Troubleshooting Agent earns its place, you will answer from a demo and a blog post rather than from a debug session of your own. Ask for a sandbox with the entitlements switched on before you put a recommendation in writing, and say plainly which parts come from running it.
For platform owners, the planning question is whether flow debugging alone justifies Now Assist for Platform or whether it rides along with a wider set of reasons. The decision belongs with the same group that owns release and environment practice, because an agent reading execution data touches every flow in the instance.
Input Modification comes with the Lux opt-in attached
Input Modification lives in Flow Designer's new UI, built on Lux, which you opt into. Adopting the feature means adopting the interface, a bigger decision than a debugging improvement usually carries. Lux is ServiceNow's newer front-end framework, and the new Flow Designer will be where many developers first meet it in daily work.
For a team, the opt-in turns into a small change management item. Panels move, keyboard habits stop working, and the person who knew the old layout blind relearns where things live. Do that in a development instance on a quiet week. If your flows sit inside custom applications, the same opt-in reaches the people who build them, so fold it into whatever cadence you already run for App Engine changes.
What to run on September 24
Pick the flow that has embarrassed you most this year, the one whose bug you have reproduced four times by cloning records. Once Input Modification reaches your personal developer instance, opt into the new Flow Designer, set a breakpoint on the step before the failure, and type in the value from the production record that broke. Time how long it takes to reach the failing state.
Compare that number with your last real reproduction. If it halves, you have an argument for the platform owner about opting the team into the new UI sooner rather than later. And when someone forwards the Troubleshooting Agent announcement asking whether to buy it, the honest answer is that you cannot try it on a personal instance, so any opinion stays secondhand until somebody hands you an instance with Otto and Now Assist for Platform switched on.
