Consulting & contracting
Your records are pictures. We turn them into data you can act on.
Loss runs that arrived as scans. ACORD forms photographed on a phone. Inspection reports, leases and prior applications sitting in a shared drive as PDFs. All of it is information your business already owns and can't use, because nothing can read it. We build the pipeline that reads it, the database it lands in, and the tools your team uses to work with it.
A typical engagement
A wholesale broker has eight years of loss runs, inspection reports and carrier correspondence in a shared drive — around 40,000 files, most of them scans or photos. Questions like which of our accounts had a water loss in the last three yearsor what did we tell this carrier about that rooftake a person half a day of opening files. The answer is in there; it just isn't data.
What we would build:
1Extraction from the files as they are
Scans, phone photos, PDFs with and without text layers, spreadsheets. The same document-reading engine that fills a Schedule of Values in SOV AI, pointed at your paper, with a schema designed around the questions you actually ask. Every extracted value keeps a pointer back to the page it came from.
2A database you own
Structured tables in Postgres — accounts, locations, claims, documents — rather than a search box over a pile of files. Joinable, reportable, exportable to Excel or wherever your analysis already lives.
3An API key for your systems
Your agency management system, your rater, your reporting stack: whatever needs the data gets it through an authenticated API, not a copy-paste. Keys are per-integration and revocable.
4A login and a UI built for your work
Not a generic dashboard. Screens for the two or three things your team does every day with this information — a renewal view, a loss-history check, a document finder — with a login for each person and permissions where you need them.
The result is that the half-day question becomes a query, and the decisions that depend on it — which accounts to remarket, which carrier to approach, where the book is concentrated — get made with the information you already had.
What you get
- A data model designed with you, around the questions you need answered — not a generic template.
- The extraction pipeline, run once over the backlog and then on every new document as it arrives.
- A Postgres database you own, with every extracted value traceable to its source page.
- An authenticated API with per-integration keys, and documentation your developers can work from.
- A login-gated web interface built for your team's actual workflow.
- Hosted and maintained by us, or handed over with the source — your choice.
How an engagement runs
Discovery
A short call, then a look at a sample of your documents — a few dozen files, under NDA. We come back with a proposed schema, a sample extraction so you can judge the accuracy on your own paper, and a fixed-scope quote.
Build
Typically a few weeks. You see the extraction running on real files early, so the data model gets corrected before it hardens, not after.
Handover
Your team is trained on the interface, your developers get the API documentation, and the backlog is loaded. From there it runs on every new document.
Ongoing
Once the interface is established, we can keep maintaining the platform for you: hosting, monitoring, extraction accuracy as your documents change, and new screens as your questions change. Or hand it over entirely — it is your system either way.
Pricing is a fixed scope agreed after discovery, so the quote reflects your documents and your questions rather than a rate card.
Who this is for
Brokers, MGAs, wholesalers, TPAs and risk managers — anyone in commercial P&C whose institutional knowledge lives in ACORD forms, loss runs, inspection reports and flood determinations that were never captured as data. If your immediate need is a clean Schedule of Values for a submission, that is what SOV AI does off the shelf — try it on your own file without an account. Consulting is for everything else in the drive.
Questions we get
- How accurate is extraction from scans and photos?
- It depends on the documents, which is why discovery includes a sample extraction on your own files before anything is quoted. Every value keeps a link to its source page, so a person can check any figure in one click, and low-confidence extractions are flagged rather than silently accepted.
- Where does our data live?
- In a database you own — Postgres, hosted by us or by you. Documents are stored privately and served through signed, expiring links. Our security practices are described on the security page, and we sign a data processing agreement.
- Can it connect to our agency management system?
- Through the API, yes. We have built against spreadsheet-based workflows, carrier templates and REST integrations; discovery is where we confirm what yours needs.
- Do you only work with insurance businesses?
- Insurance is where the team's domain knowledge is, and where we can design a sensible data model quickly. The extraction engine itself is not insurance-specific — if your problem is documents that should be data, write to us.
See also our security practices and the P&C glossary.
Find out more
The fastest way to find out whether this fits is to tell us what you have. A sentence or two about the documents and what you wish you could do with them is enough to start.
Email us a line about what you have and what you want to be able to do with it — or use the form.