Straight-through processing (STP) is underwriting that moves a submission from intake to a decision with no manual touch. In commercial P&C, auto-clear the clean risks and route exceptions to a human.
Straight-through processing (STP) is underwriting that moves a submission from intake to a bound decision with no manual touch. In commercial property and casualty insurance, STP works well for clean, simple risks. It is the wrong goal for complex ones. The profitable move is to auto-clear the easy submissions and route every exception to a human. Exceptions are where the loss ratio is won or lost.
How does straight-through processing work in underwriting?
STP works by chaining four automated steps so a submission never waits in a queue. First, the system ingests the submission from an email inbox or portal, the stage known as submission intake. Second, it extracts the data from each document, such as an ACORD form or a loss run. Third, it checks the risk against coded appetite rules. Fourth, it prices and clears the account, or routes it onward. Each step hands structured data to the next. A clean risk moves start to finish in minutes, not days.
What is a good STP rate, and why is 100% the wrong target?
A good STP rate clears the simple risks and stops there. The STP rate is the share of submissions decided with no human touch. Chasing 100% forces the system to auto-decide risks it should never own. That is how adverse selection enters the book. Commercial casualty severity keeps rising. US liability claims rose 57% over the prior decade (Swiss Re Institute, 2024). One misjudged complex risk can cost more than a hundred clean auto-clears ever saved. So the metric that matters is not the STP rate alone. It is the STP rate on the risks that actually deserve it.
Where does straight-through processing break down?
STP breaks down on the submissions that are incomplete, unusual, or high in severity. Real commercial submissions arrive messy. A loss run may be a scanned PDF with inconsistent columns. A schedule of values may hold hundreds of locations. An account may sit right at the edge of appetite. These are the risks a machine should flag, not force. This is the core reason AI alone cannot close the intake gap. Pibit.AI pairs AI extraction with a managed human-in-the-loop review team. The AI handles volume. The human handles judgment and edge cases. That pairing is what holds accuracy at 99.9% contractual field-level accuracy (Pibit.AI internal data).
| Dimension | Straight-through processing | AI plus human-in-the-loop |
|---|---|---|
| Best fit | Clean, simple, in-appetite risks | Complex, incomplete, or high-severity risks |
| Human touch | None | Review on every exception |
| Speed | Minutes per submission | Fast on volume, hours on the hard cases |
| Loss-ratio risk | High if the auto-clear band is too wide | Contained by human judgment on edge cases |
| Auditability | Weak unless every rule is logged | Strong; AI extraction and human edits both tracked |
How do carriers raise their STP rate safely?
Carriers raise STP safely by widening the auto-clear band only as accuracy and audit trails prove out. Start narrow. Auto-clear the cleanest risks with the tightest appetite match. Keep a human on everything else. Log what the AI extracted and what a person changed. That audit trail matters, because regulators now expect AI-assisted decisions to be tested and explainable (NAIC Model Bulletin on the Use of AI Systems by Insurers, 2023). Widen the band as the data earns trust, an approach that fits an agentic underwriting architecture. One workers' compensation MGA using Pibit.AI processed 1,053 loss runs and 626 submissions in a single month with zero errors (Pibit.AI internal data). The speed came from automating the clean work, not from removing the human.
Frequently asked questions
What is a straight-through processing rate?
A straight-through processing rate is the share of submissions decided with no manual touch. In commercial P&C, a healthy rate reflects how many risks are genuinely simple, not how aggressively the system auto-decides.
Is straight-through processing the same as underwriting automation?
No. Automation speeds up individual steps such as data extraction and appetite checks. Straight-through processing chains those steps so a submission needs no human at all. Automation helps every submission; full STP suits only the simplest ones.
Can AI make commercial underwriting fully straight-through?
Not for complex risks. AI clears clean, well-structured submissions quickly. Complex commercial accounts still need a human for exceptions, edge cases, and defensible judgment. The reliable model is AI plus managed human-in-the-loop review.
