Insurance Broker Software: The Problems Costing You Renewals, and What a Custom Build Actually Fixes

Why broker software makes or breaks a high volume agency

Your agency management system is a ledger, not an operating system. Applied Epic, Vertafore AMS360, HawkSoft, EZLynx, Nowcerts: they hold the policy record, take the Ivans download, run direct bill and agency bill accounting, and spit out ACORD forms. None of them run the actual work of a brokerage. The work runs in Outlook rules, a shared drive folder called Marketing 2026, an Excel tab your commercial lines manager named "Q3 final v4", and the memory of one account manager who has been there eleven years.

The scene repeats in almost every commercial agency we have rebuilt this for. It is 60 days out on a general contractor renewal, $185,000 in premium, roughly $27,000 of commission. Your marketer builds the submission: ACORD 125, 126, 140, five years of loss runs, a supplemental app, a vehicle schedule. It goes to eight markets. Three by email to underwriter inboxes, two through carrier portals, three through a wholesaler. Then it disappears. Two carriers decline in week one, buried in a thread nobody opened. One asks for updated loss runs on a Friday afternoon and the request sits until Tuesday. One never replies at all, and nobody notices until day 12 before expiration, when the producer asks where the quote is and the honest answer is that nobody knows. You bind with the incumbent at a 9 percent increase because there is no time left to do anything else.

Now multiply that across a 12,000 policy book, 14 account managers, six producers, three offices. One bad renewal is not the leak. The leak is that nobody in the building can answer a basic question in under a day: which submissions have gone quiet past four days, which renewals are at risk, whose book is slipping. If your agency does $9 million in commission, four points of avoidable retention slippage is $360,000 a year, and it does not appear on a single standard report in Epic.

Submissions vanish between your marketer and eight carriers

The pre-bind marketing process is the largest unmanaged part of a brokerage. Your AMS models a policy. It does not model a market. Epic gives you an activity, which is a task hanging off a policy record, not an object with a carrier, an underwriter, a wholesaler, a status, a subjectivity list and a clock. So you cannot ask "show me every market silent for more than four business days across all open submissions," because the field does not exist and never will. Comparative raters cover the easy end: EZLynx and PL Rating handle personal lines, Tarmika, Bold Penguin and Appulate cover a slice of small commercial appetite. Your $185,000 contractor is not in that slice. It is in email.

A custom build gives every market its own record. One submission, one row per market: carrier, underwriter name and email, wholesaler, appetite match, date sent, last touch, status (submitted, info requested, quoted, declined, bound, lost to incumbent), open subjectivities, quoted premium, commission, and a coded decline reason. A board view shows the marketing pipeline the way a sales team sees deals. An aging rule flags any market past four business days with no response and pings the marketer, not the producer. The coded decline reasons build an appetite map over time, so the next Texas roofing account skips the six carriers that declined the last six roofers instead of burning nine days finding out again.

There is one genuinely good use of a model in this workflow, and it is not writing anything. Point a shared submissions mailbox at a parser that reads inbound carrier email plus attachments, classifies each one as quote, decline, subjectivity, or information request, pulls premium, effective date, carrier form numbers and subjectivities out of the quote letter PDF, and updates the market row without a human opening it. Of everything we build in this category, that one flow pays for more of the build than anything else we ship.

Renewals live in one person's head and a spreadsheet

Epic and AMS360 fire 90, 60 and 30 day renewal activities. Every account manager gets 40 activities a day, so activities get cleared, not worked. Worse, a $900 personal auto renewal and a $185,000 contractor look identical in the queue. In Q3 you might have 340 policies renewing, 22 of them commercial over $25,000 premium. Ask your CL manager which of those 22 got a pre-renewal exposure call and watch the pause.

The off-the-shelf systems cannot fix this because their trigger is a date, and the thing you need is triage. A custom renewal workbench keys off expiration but tiers by commission, complexity and loss ratio: 120 days out for the top tier, 60 for the middle, fully automated touch for small personal lines. Each renewal carries the context in one screen: date the loss run was last pulled, whether the exposure update went out, the market strategy (remarket or roll), the carrier's rate action, and the client's last three service tickets. Layer a retention risk score on top, built from data you already own: a claim in the last 12 months, a premium increase over 10 percent, a logged service complaint, no producer contact in 180 days, late payments. Keep it as a visible scorecard, not a black box, so your CL manager can argue with it and fix the weights.

AI helps here in two specific places. It drafts the exposure update email per client from last year's application data, so the account manager edits instead of writes. And after hours, a client portal answers the two questions that generate half your inbound calls, "when does my policy renew" and "can I get a certificate," at 11pm on a Sunday.

Every dec page, loss run and quote letter gets retyped by a human

Ivans download works where the carrier agrees to send data, which is standard personal lines and some standard commercial. Specialty, E&S, program business and wholesaler paper do not download. So an account manager opens a 14 page dec page PDF and types the schedule, limits and endorsement numbers into Epic by hand. In the agencies we have worked in, that is 30 to 45 minutes per policy, every time, forever. Applied and Vertafore cannot solve it because their product depends on carrier cooperation that does not exist. Indio and Broker Buddha automate the application going out, not the carrier document coming back.

What a custom build does: a document pipeline on a watched mailbox and drop folder. Classify the file (dec page, endorsement, loss run, quote letter, binder, premium audit, cancellation notice), extract into a typed schema per class with a confidence score per field, and write into the AMS through its API or a validated import. Anything below the confidence threshold lands in a review queue with the PDF on one side and the extracted field highlighted on the other, so a human confirms in eight seconds instead of typing for 40 minutes. Loss runs get normalized into a real claims table, so five years across three carriers becomes a queryable loss ratio by class of business instead of a stack of PDFs. Be honest with your vendor about accuracy: nothing hits 100 percent. Design for the exception queue. We aim for straight through processing on the 70 to 80 percent of clean, machine generated documents and route the rest to a person.

Certificates and endorsements eat your service team alive

One contractor account, 240 certificate holders. A general contractor wants an additional insured with a specific blanket endorsement form number and a waiver of subrogation, wants it on their own form, and wants it before the crew shows up Monday. Your account manager opens Epic, issues an ACORD 25, attaches, emails. Then does it again. At renewal you reissue all 240 and lose a week.

Epic and AMS360 do issue certificates and hold holder lists, so this is not a gap in their feature list. The gap is that the insured and the holder cannot serve themselves, and nothing checks whether the requested wording is actually supported by an endorsement on the policy. A custom portal lets the insured or the holder pick the project, pulls the bound policy, verifies the AI and waiver wording against the endorsements on file, and issues in under a minute with a full audit trail. When the wording is not supported, it does not issue: it opens a task naming the exact gap, and that record is your E&O defense two years later when a claim lands. Renewal reissue becomes a batch job you approve, not a week of a person's life.

Commission statements never match what you booked

Thirty carriers, thirty statement formats. A PDF from one, a CSV from another, a portal login for a third, and direct bill sweeps that arrive on their schedule. Then producer splits: 40/60 new business, 25/75 renewal, house accounts carved out, a mid term endorsement that changes premium and therefore commission and therefore comp, retroactively. In the agencies we have rebuilt this for, someone in accounting burns four to six days a month on this, and the answer to "is this producer's book profitable" arrives 45 days late.

Epic and AMS360 have direct bill reconciliation and it works when the statement matches your policy record exactly, which is roughly never on E&S and program business. A custom module ingests every format, extracts the lines, matches against your policy data on a fuzzy key of named insured plus policy number plus effective date, auto-reconciles the exact hits and queues only the variances with a visible diff. Producer compensation then calculates from reconciled data instead of a spreadsheet. The side effect is the one your CFO cares about: a commission received versus expected view by carrier, which surfaces the market that has been short paying you for eight months.

What this costs and how long it takes

These are Digital Heroes delivery bands across 2,000 plus projects, not a market survey. A focused first release typically runs $60,000 to $130,000 and ships in 12 to 16 weeks. For a brokerage, that first release is usually the submission tracker, the carrier email parser, the renewal workbench, and a read-only sync from the AMS. A full platform, meaning documents, certificates, commissions, client portal and two-way AMS writeback, runs $150,000 to $400,000 phased over 6 to 12 months.

What drives price up specifically in this category: the number of carrier and wholesaler touchpoints, and how many of them have no API at all, because those get built around email and portals. AMS API access is the big one, since Applied and Vertafore gate integration through their partner programs on their timelines and their fees, which is calendar time you must budget for and cannot compress. ACORD and Ivans data mapping. Commission accounting, which is the single most underestimated module in every estimate we have seen. Surplus lines filing and stamping office rules, where each additional state is real work, not a config toggle. Data migration of 15 years of Epic attachments and activity history. And compliance: SOC 2 Type II, the NAIC insurance data security model law as your states adopted it, and NY DFS Part 500 if you write New York.

Build versus buy: when Epic or EZLynx is genuinely the right answer

Do not build an agency management system. That is a position, not a hedge. If you are under roughly $2 million in commission, mostly personal lines, one or two locations, EZLynx or HawkSoft plus Indio for applications will beat anything you commission, and it will beat it for the next five years. Buy, configure, move on.

The signals it is time to build: you have three or more people whose real job is moving data between systems. Your marketing process lives in Excel and you cannot answer "which submissions are stale" in ten seconds. More than a quarter of your revenue sits in E&S, program or specialty business where download does not reach. You run a vertical program where your edge is a data model no vendor will ever build for you. The vendor answer to your top request has been "it is on the roadmap" for two years. Or your combined annual spend on AMS seats, rater, certificate tool, e-signature and workflow bolt-ons already crosses $150,000, at which point you are funding a build without owning one.

Build around, not instead of. Epic or AMS360 stays the record of policy and the accounting engine, because replicating download, ACORD form generation and agency bill accounting is a multi-year mistake. Your system owns the pre-bind pipeline, the service layer, the document pipeline and the reporting your leadership actually reads.

How to choose a developer for insurance broker software

Make them draw the data model before you sign anything. Account, policy, policy term, line of business, coverage, endorsement, market, submission, quote, claim, certificate, holder, transaction. If they model policy as one flat row with no terms and no endorsements, they have never shipped in this industry. Then ask the killer question: how does a mid term endorsement that changes premium flow through to commission and then to producer compensation, retroactively, without a manual journal entry. Watch what happens.

Ask what they have shipped against Applied Epic or AMS360 APIs, and what they did where no API existed. The credible answer includes partner program timelines, email parsing, SFTP drops, and occasionally a supervised job against a carrier portal. Anyone promising clean API integration across all your carriers has not spoken to your carriers.

Push past "we are secure" on compliance. Ask specifically about encryption of nonpublic personal information at rest, retention and purge rules, an audit log at record level, role separation so one producer cannot pull another producer's book, and how the system would answer an E&O discovery request or a state market conduct exam. Ask whether they have carried a client through SOC 2 and what the auditor made them change. The second answer tells you whether the first one is real.

Settle ownership and exit on day one. Code in your repository, infrastructure in your cloud account, a documented data model, and a handover document good enough that another firm could pick it up in a week. Ask to see the handover doc from their last project. If it does not exist, you are not buying software, you are renting a dependency.

The agency that can answer "which of my submissions went quiet this week" in three seconds is not smarter than yours. It just stopped storing that answer inside a person.