From the inbox to a decision in minutes, not days.
The time goes on intake, not on underwriting. New business, renewals, endorsements and coverholder bordereaux arrive the same way and go through the same intake. SnapLine reads the email, the slip, the schedule and the scan, and returns one structured record; your underwriter decides what happens to it. Open any field and you see the document it was read from. It was built inside Probitas, a Lloyd’s syndicate, and it runs in production today at a US wholesale MGA.
- Covering email
- MRC slip.pdf
- schedule_of_values_v3.xlsx
- schedule_of_values_FINAL.xlsx
- survey_scan.pdf
- IMG_4471.jpeg
- named_insuredRavensholme Estates LimitedTraced to source
- total_insured_valueGBP 24,850,000Two sources disagree. Both are on the record.
- locations14Traced to source
- construction_classMasonry, non-combustibleTraced to source
- year_builtNot in submissionSent to outreach, not inferred.
What this is actually fixing
Four things go wrong in submission intake, and each costs money in a different way.
Reading a pack, finding the fields that matter and keying them into three systems is work that scales with volume and nothing else. It is also the work people are doing at seven in the evening in renewal season.
A transposed figure in a declared value does not announce itself. It prices, it binds, and it is discovered at the claim. Cross-document checks raise the disagreement while someone can still do something about it.
Brokers place with the market that answers. When intake is a queue, the quote is late for reasons that have nothing to do with appetite or capacity.
Exposure management, rating and the data warehouse inherit whatever intake produced. Structured at the front means those systems stop being fed by hand, and the numbers in them start agreeing with the documents.
Together these set two things: what it costs your operation to handle a submission, and how often a decision rests on a figure that is wrong. Those are the two numbers SnapLine is meant to move. We are not asking you to take either on trust — both are measurable on your own submissions before you commit to anything.
Renewals and endorsements, not just new business
Most of the documents an underwriting operation handles are not first submissions. They are the same risk coming back a year later, a change part-way through, or a return from a coverholder.
The submission pack as it arrives: email, slip, schedule, loss runs, whatever else the broker attached.
The expiring policy read against the renewal submission, so what changed since last year is the first thing visible rather than something to be found.
A mid-term change read and applied against the policy as it currently stands, instead of against the version someone remembers.
Coverholder returns ingested and checked against the binder, with what does not reconcile raised as a list rather than discovered at the quarterly review.
What an underwriter sees
The add-in works inside the underwriter's own inbox, so a submission doesn't have to be forwarded into another system before anything can read it.
The standard form in each market, and they defeat general-purpose document AI in different ways. An MRC slip is a sectioned instrument where meaning sits in position rather than in labels. An ACORD is standardised, but arrives as a scan, in whichever version the agent uses, and often with the detail that matters added by hand.
An occupancy description carries forward from last year because nobody had a reason to change it, and the tenant has since left. Atlas surfaces vacancy across a book rather than one submission at a time, so the question can be asked before the location is rated as tenanted.
Requires Atlas, which is licensed separately.
An Atlas field arrives with the source it came from, an indication of quality, and the date the data applies to — per field, not per record. That turns "do I trust enrichment" into "do I trust this value", which an underwriter can actually answer.
Requires Atlas, which is licensed separately.
Provenance
How the record gets checked
Extraction is half the problem. The other half is the question any underwriter asks of a number they didn’t key themselves: where did that come from, and is it right? A record nobody trusts gets checked by hand, and then the hours never came out in the first place.
Hover or select a field in the structured record and SnapLine shows you the region of the source document it was read from. You confirm it, correct it, or send it back to the broker. The record carries that decision with it, and the next person to open the file can see who decided what.
Risk code PD
Inception 01.11.2026
- Insured
- Ravensholme Estates Limited
- Interest
- Commercial property portfolio, 14 locations
- Construction
- Masonry, non-combustible
- Year built
- not stated
| Loc | Address | Occupancy | Building | Contents |
|---|---|---|---|---|
| 001 | Ravensholme Works, Leeds | Warehouse | 4,120,000 | 980,000 |
| 002 | Carlisle Road, Bradford | Light industrial | 2,340,000 | 610,000 |
| 003 | Eastgate Retail, Wakefield | Retail | 3,880,000 | 1,240,000 |
| … 11 further locations | ||||
| Total insured value | cited24,850,000 | |||
Illustrative. Broker, insured and all figures are fictional.
From attachment to quote-ready record
- 01Arrives
An email lands in the underwriting inbox with a stack of attachments, several of which are versions of the same schedule and one of which is a photograph of a roof.
- 02Triaged
SnapLine sorts it where it sits.
- 03Extracted
Fields are read across the whole bundle, including the scans and the spreadsheet tabs nobody opens, and disagreements between documents are preserved rather than resolved quietly in the background.
- 04Validated
An underwriter checks what matters and corrects what doesn't look right.
- 05Quote-ready
The structured record goes downstream with its provenance and every correction attached, so the audit trail exists without anyone having to assemble one afterwards.
Where SnapLine came from, and where it runs
SnapLine was built inside Probitas, a Lloyd’s syndicate, against the submissions that syndicate actually receives. That is why it treats an MRC slip as a sectioned instrument rather than as prose, and why a schedule that disagrees with the slip reaches an underwriter as a disagreement instead of a single confident number.
It is in production today at a US wholesale MGA, under a commercial contract, on live submissions. We will not name them here. Neither Probitas nor Lloyd’s endorses SnapLine, and no syndicate is running it in production yet — that is covered on the London Market page.
Independently audited
Lloyd's Lab Cohort 15 FinalistA selection. Not an accreditation, not a deployment, and not an endorsement by any syndicate.
What an engagement looks like
Three shapes, and which one fits depends on the volume, the lines, and what sits downstream of intake. Scope is agreed in writing before anything starts.
Atlas property intelligence is licensed separately and is not part of any of the three.
A demo, then a fixed-scope Proof of Value measured on your own live submissions.
- A demo with five sample submissions already loaded
- Up to three of your own submissions, uploaded and processed in the session
- A Proof of Value run across twenty of your live submissions
- Scope agreed in writing before anything starts
- A written report at the end, yours to circulate internally
SnapLine in the underwriting workflow for a defined team and a defined set of lines.
- The Outlook add-in in your underwriters' own inboxes
- Submission queue and submission detail views
- Open any field and see the document it was read from
- Outreach for missing information, back to the broker
- Quote-ready structured output
More than one underwriting team, more than one entity, or a group-wide rollout.
- Everything in Deployment, across multiple teams or legal entities
- Security review support, with SOC 2 Type 2, ISO/IEC 27001 and ISO/IEC 42001 documentation
Bring three of your own submissions
A demo on our sample data tells you what the software does. A demo on your own slips tells you whether it survives contact with the way your brokers actually send things. The second is the one worth your half hour.
Bring three submissions you already know the answers to, including one that gave you trouble. That is the half hour that tells you something.