Send Alternatives: Choosing an Underwriting Workbench or Building Your Own

Why specialty underwriters look past the workbench

Underwriting workbenches exist because a real problem existed. Specialty submissions arrive as email attachments, broker slips, spreadsheets of schedules and PDFs that vary by broker. Before any workbench, underwriters spent a large slice of the week doing clerical work: opening attachments, retyping schedules, checking appetite, chasing missing information, and keeping a personal spreadsheet of what was quoted and what was still open. A workbench takes that away, and the relief is real.

The reason teams then start looking at alternatives is usually one of three things. The first is pricing logic. Your actuaries maintain models that are genuinely proprietary, and getting them into a configured pricing screen either flattens them or turns into a permanent integration project. The second is the broker experience. In specialty distribution the quote turnaround and the way a broker interacts with you is a competitive weapon, and a vendor front end limits how sharply you can wield it. The third is roadmap dependency. You need a capability in eight weeks, the vendor has it planned for a future release, and there is nothing you can do to accelerate it.

There is a fourth, less discussed reason: coverage of the rest of the lifecycle. A workbench is deliberately a workbench. Binding, document production, policy issuance, endorsements, billing and bordereaux either live elsewhere or come as adjacent modules, and stitching that chain together is work regardless of who supplies which link.

What Send genuinely gets right

Be fair about the category and the product. A modern underwriting workbench is built for the way specialty business actually arrives, which is more than most legacy systems can claim. Submission ingestion that reads what brokers send, appetite and triage rules that stop underwriters looking at risks they will never write, a data capture model that handles schedules rather than single risks, and a record of every quote and decline in one place are all genuinely useful and genuinely hard to build well.

The second real strength is the delivery model. It is a configured platform rather than a multi year core implementation, so you can be underwriting on it in a timeframe that a business case survives. For a new MGA or a syndicate launching a line, that speed is worth a great deal, and it is exactly what a custom build cannot match at the start.

The third is that the vendor absorbs the maintenance of a moving target. Submission formats change, brokers change, and someone has to keep the ingestion working. Paying for that is rational when it is not the thing that makes you money.

Where a workbench stops helping

Configuration ceilings arrive first. Every configured platform models risks, products and pricing in a particular way, and your unusual product is unusual precisely because it does not sit in that shape. Facultative reinsurance structures, layered and shared placements, complex schedules with per location sublimits, or a pricing model that consumes third party data at quote time all tend to test the edges. You can usually get there, but the route runs through workarounds, and workarounds are where operational risk accumulates.

Data ownership is second. Your submission and decline data is a strategic asset. It tells you what your brokers are showing you, what you are turning away, and where your appetite is mispriced. When it sits in a vendor model, extracting it into your own analytics stack is possible but rarely as clean as querying your own database, and the analysis that matters most is often cross cutting.

Third is per seat and per volume economics. Workbench pricing generally scales with underwriters or transaction volume. Growing a team of eight to a team of forty changes the annual number materially, while the cost of running software you own is dominated by hosting and engineering rather than headcount.

Fourth is the integration burden. The workbench has to talk to your policy system, your document generator, your accounting, your capacity providers and your data vendors. Those joins are yours to build and maintain whichever platform you pick, and they are a bigger part of total effort than most buyers assume.

The realistic options

Other workbench and specialty platforms are the first option, and there are credible ones. INSTANDA focuses on rapid product configuration, Novidea comes at it from the distribution and broker side, Socotra and EIS bring cloud native policy cores with API access, and Duck Creek, Guidewire, Majesco and Sapiens sit at the heavier core suite end. Whitespace and similar market platforms address placement rather than internal underwriting. Choosing between them is largely a question of where your centre of gravity is: submission handling, product configuration, distribution, or the policy record.

The second option is staying and narrowing scope. A common and underrated move is to keep the workbench for what it is good at, ingestion and triage, and build only the pricing and decision layer alongside it, called through an interface. That preserves the speed you bought and puts your intellectual property in your own codebase.

The third is a full custom workbench. This is the right answer when your submissions are highly structured within your niche, your pricing is proprietary, your broker relationships justify a bespoke front end, and you have the internal product ownership to run it. It is the wrong answer when you are still discovering your underwriting process, because building a system to encode a process you have not settled is the most reliable way to waste six months.

When staying is the right decision

Stay if you are early, still shaping appetite and process, and speed matters more than fit. Stay if your pricing is conventional enough to express in configuration without contortions. Stay if your underwriting team is small, because per seat costs are modest at small scale and the cost of owning software is not. And stay if the workbench is working and your real complaint is about a neighbouring system, because replacing a functioning link in the chain to fix a different link is a common and expensive error.

When custom pays back

Build when the decision is the product. If your edge is a pricing model, a data enrichment pipeline, or a triage approach that lets you quote faster than the market, that logic belongs in code you control and can change on a Tuesday. Build when you need the broker facing experience to be distinctly yours, since submission and quote interaction is one of the few places a specialty underwriter can be visibly better than the next one. Build when the volume of underwriters or transactions has pushed platform cost into a range where a fixed build plus hosting is clearly cheaper over three years. And build when your data needs to be first class, queried directly and joined to claims and exposure without an export step.

Migration reality

The good news is that a workbench migration is less brutal than a policy system migration, because most of the regulated record lives downstream. The work concentrates in three places. First, submission history and quote records: export them, including declines, because your decline data is an underwriting asset and losing it costs you the ability to analyse appetite over time. Second, integrations: every connection you built to policy, documents, data vendors and accounting has to be rebuilt and tested, and this is usually the largest line in the plan. Third, underwriter habit. Run the new system alongside the old for a renewal cycle in one class of business, get the underwriters who will complain loudest into that pilot, and fix the friction before you move the rest.

Cost bands

Workbench pricing is quoted, generally scaling with underwriters or transaction volume, with implementation and configuration on top in year one. On the build side, from Digital Heroes delivery experience, a custom underwriting workbench covering submission intake, triage, data capture, pricing and quote records runs $120k to $280k across four to seven months. A full specialty platform, adding binding, document production, endorsement handling and bordereaux, runs $300k to $550k. Ongoing engineering of ten to twenty percent of build cost a year is realistic, since integrations and products change constantly.

The verdict

Send and its peers solve the clerical half of specialty underwriting properly, and if that is your bottleneck the sensible answer is to keep paying for it. The question worth asking is narrower than replace or keep: which part of your underwriting is genuinely yours? Ingestion and triage are commodity work best rented. Pricing logic, decision rules and the broker experience are not commodity work, and every year they sit inside someone else's configuration model is a year you cannot change them at the speed your market rewards. Split accordingly, and only consider a full custom platform once your process is stable enough to be worth encoding.