WC Class Code Mapping: 3 Tools Compared for 2026

Assigning the correct NCCI class code to a new workers' compensation policy is one of those quiet tasks that decides whether a commercial account is profitable or a problem. Get it right and the premium matches the exposure. Get it wrong and you create an audit dispute, an E&O exposure, and a frustrated insured all at once. The work is repetitive, rules-heavy, and yet still done by hand at most independent agencies. This guide compares three tools agencies actually reach for and then walks through the workflow recipe that removes the manual lookup entirely.

Key Takeaways

• Class code mapping is a rules problem, not a judgment problem — which makes it an ideal automation candidate once the source data is structured.

• NCCI publishes more than 600 active class codes according to NCCI (2024), so manual recall is unreliable past a handful of common industries.

• NCCI's Riskworkstation, Applied Epic, and Tarmika each solve a different slice; none covers the full intake-to-bind path alone.

• A reliable automated workflow needs a clean classification rule set, a confidence threshold, and a human review queue for low-confidence matches.

• US Tech Automations sits above the rating tools, orchestrating the lookup, validation, and write-back so your raters and AMS stay in sync.

Why class code accuracy decides commercial profitability

A workers' comp class code maps an employer's operations to a base rate that reflects the loss history of similar businesses. The classification drives the manual premium before experience modification, so an error here compounds through the entire calculation. The commercial lines that depend on this — workers' comp, general liability, and package policies — represent a large and growing book for independent agents.

The independent agency channel is not a niche; it is the backbone of commercial distribution. Independent agencies write roughly 62% of U.S. commercial P&C premium according to the Big "I" 2024 Agency Universe Study. That share means the accuracy of class code assignment at the agency level affects a meaningful slice of the entire commercial market, not just a single carrier's book.

The macro stakes are large because the P&C market itself is large. U.S. P&C direct written premiums exceeded $900 billion according to the Insurance Information Institute 2025 Fact Book. Even a fractional error rate in classification, applied across that volume, produces audit adjustments, premium leakage, and disputes worth far more than the cost of fixing the workflow.

What a misclassification actually costs

When an auditor reclassifies a policy at renewal, three things happen. The insured receives an additional premium bill they did not budget for. The carrier loss ratio for the original class was distorted for a full term. And the producer spends hours defending the original assignment instead of selling. None of those costs appear on a quote, which is exactly why agencies underinvest in getting the classification right the first time.

The labor math reinforces the point. Insurance customer service representatives earn a meaningful hourly wage — the median CSR wage exceeds $22 per hour according to the U.S. Bureau of Labor Statistics (2024) — and class code research on a complex commercial account can consume a large fraction of an hour per policy. Multiply that by a steady new-business flow and the manual lookup quietly becomes one of the more expensive recurring tasks in the service department, all to produce a result that a rules engine could generate in seconds.

Why insurers see this as an automation priority

The carriers and large brokers are already moving here, which is a signal for the agency channel. The industry is shifting routine, rules-based work to software so skilled staff can focus on judgment. Insurers are among the heaviest adopters of process automation precisely because so much of underwriting and servicing is rules applied to structured data, according to McKinsey & Company (2024) research on insurance operations. Class code mapping is a textbook example: a defined rule set applied to a structured business description, with a clear right answer in the great majority of cases.

Who this is for

This guide is written for commercial-lines CSRs, account managers, and agency principals at independent agencies handling a steady flow of new workers' comp and BOP submissions across multiple states or industries.

Red flags: Skip this if you write fewer than 10 new commercial policies a month, operate single-state with three or four repeat industries you know cold, or have no agency management system to write results back into. At that volume, a well-maintained spreadsheet and the NCCI lookup tool are genuinely faster than building automation.

The three tools commercial agencies actually use

Before automating anything, understand what each tool does and where its responsibility ends. None of them is a full intake-to-bind solution, which is the gap automation closes.

NCCI is the rating organization. It maintains the classification system, the scopes manuals, and the lookup utilities most states reference. Applied Epic is an agency management system — it stores the policy, the insured, and the activity log, but it is not a classification engine. Tarmika is a commercial rater and marketing platform that pushes a single submission to multiple carriers; it consumes a class code rather than authoritatively deriving it.

The honest read: NCCI owns the rules, Applied Epic owns the record, and Tarmika owns the carrier distribution. The classification decision — turning "framing carpentry subcontractor, no work above two stories" into the right code — still lands on a human unless you build a layer that does it.

Where each tool genuinely wins

NCCI Riskworkstation wins on authority. If there is a dispute about a phraseology, NCCI's own manual settles it, and no third-party tool can overrule that. Tarmika wins on speed-to-market for WC quoting — pushing one submission to a dozen carriers is something neither NCCI nor a generic automation layer replaces. Applied Epic wins on being the system of record your auditors, accounting, and renewals all already trust.

It is worth being clear about classification volume too. Workers' compensation uses different classification systems by jurisdiction — most states follow NCCI, while a handful of independent bureau states (such as California's WCIRB) maintain their own. A multi-state commercial agency therefore manages more than one rule set, and the count of distinct codes a busy producer might touch over a year runs well into the hundreds. State workers' compensation programs cover the vast majority of the U.S. workforce, according to the National Academy of Social Insurance (2024), which is why the classification surface area is so broad and why memory-based assignment fails as a book grows.

The automated class code mapping workflow

Here is the recipe. It assumes you have Applied Epic (or a comparable AMS), access to the NCCI classification rule set for your states, and a rater like Tarmika downstream. The goal is to take a structured business description and produce a validated class code with an audit trail, escalating only the ambiguous cases to a human.

1. Capture the operations description at intake. Standardize the submission form so the producer enters the insured's primary operation, secondary operations, payroll by activity, and any governing-class exclusions in structured fields — not a free-text blob.

2. Normalize the business description. Strip it to the operative phrases (e.g., "residential framing," "no roofing," "ground-level only"). This is where an AI step earns its place, because human-entered descriptions are inconsistent.

3. Match against the NCCI rule set. Run the normalized description against your classification table, returning the top candidate codes with a confidence score for each.

4. Apply governing-class logic. Where multiple operations exist, apply the governing-class and standard-exception rules so the highest-rated applicable code is selected correctly rather than the cheapest.

5. Set a confidence threshold. Auto-assign only when the top match clears a confidence bar (for example, 90%) and the second candidate is well below it. Everything else routes to review.

6. Route low-confidence matches to a human queue. A CSR sees the description, the candidate codes, and the reasoning, then confirms or overrides in seconds rather than starting from scratch.

7. Write the confirmed code back to Applied Epic. Update the policy record, the activity log, and the rating fields automatically so the AMS stays the system of record.

8. Push to the rater. Hand the validated code to Tarmika so the multi-carrier submission carries the correct classification from the first quote.

9. Log every decision for audit. Store the description, candidate set, confidence score, and who approved it, so a future auditor or E&O reviewer can reconstruct the call.

This is where US Tech Automations fits: it is the orchestration layer that runs steps 2 through 9, calling the rules, scoring confidence, holding the review queue, and writing back — while NCCI stays the rule authority and your rater stays the distribution channel.

A worked example

Take a small commercial cleaning contractor. The intake description reads "janitorial services for office buildings, no exterior window cleaning above ground floor, 6 W-2 employees." The normalization step extracts "janitorial," "office buildings," and the exterior exclusion. The match returns the standard janitorial classification at high confidence with the window-cleaning code explicitly excluded by the ground-floor note. Confidence clears the threshold, the code writes back to Epic, and the CSR never touches it. A messier description — "general maintenance and repair, some light construction" — would fall below threshold and route to a human, exactly as it should.

Build versus buy: the realistic comparison

Most agencies considering this ask whether to script it in-house, lean on a point tool, or use an orchestration platform. Here is the honest trade-off, including where the alternatives beat a platform.

The claim-handling parallel is instructive: speed compounds across a book. The average auto P&C claim cycle runs over 14 days according to the NAIC 2024 Claims Processing Benchmark, and the agencies that compress cycle time do it by removing manual handoffs — the same principle that makes class code automation worth building once your volume justifies it.

Setting realistic benchmarks before you automate

Before committing to a build, measure your current state so you can prove the gain. Three numbers matter: average minutes spent classifying a new commercial policy, the percentage of policies reclassified at audit, and the share of new-business turnaround time consumed by classification waiting on a senior reviewer. Most agencies have never measured any of the three, which is why the manual cost stays invisible.

With those three numbers, the build-versus-buy decision stops being a gut call. If classification eats more than a few hours of skilled labor a week and your audit reclassification rate is non-trivial, an orchestration layer pays for itself quickly. If both numbers are low, you genuinely do not need to automate yet — and knowing that is just as valuable as deciding to build.

When NOT to use US Tech Automations

If you only quote workers' comp in a single state with a handful of familiar industries, an orchestration layer is overkill — Tarmika's rater plus your own knowledge will be faster and cheaper. If your classification disputes are genuinely about phraseology interpretation rather than volume, you need NCCI's scopes manual and an experienced underwriter, not automation. And if you have no agency management system to write results into, fix that foundation first; automating a write-back to nowhere just moves the manual work downstream.

Common mistakes that automation should prevent

Even a good workflow fails if it encodes the wrong assumptions. The mistakes below are the ones that survive past go-live because they look correct until an audit surfaces them.

• Treating the lowest-rated candidate code as the answer because it produces a cheaper quote — governing-class rules exist precisely to stop this.

• Auto-assigning at low confidence to keep the queue empty, which trades a clean queue for audit risk.

• Ignoring secondary operations that carry their own standard exceptions.

• Failing to log the reasoning, so the assignment cannot be defended at audit.

• Letting the rule set drift when NCCI updates phraseologies, then trusting stale matches.

Glossary

• Class code: A numeric code mapping an employer's operations to a workers' comp base rate.

• Governing classification: The basic classification that best describes the insured's overall business, used to assign standard exceptions.

• Standard exception: A classification (like clerical office) applied separately from the governing class.

• Phraseology: The official NCCI text describing what a class code includes and excludes.

• Scopes manual: NCCI's detailed guide interpreting each classification.

• Manual premium: Premium before experience modification, driven by class code and payroll.

• Experience mod: A multiplier adjusting premium based on the insured's loss history.

• Confidence threshold: The score above which an automated match is auto-assigned rather than reviewed.

TL;DR: Class code mapping is a rules-driven lookup that agencies still do by hand. NCCI owns the rules, Applied Epic owns the record, Tarmika owns distribution — but none derives the code. An orchestration layer like US Tech Automations normalizes the description, scores candidate codes, escalates the uncertain ones, and writes the result back, cutting manual lookup time while preserving an audit trail.

For agencies ready to map the workflow against their own volume, the US Tech Automations pricing page lays out tiers by document and workflow volume. You can also explore the broader agentic workflows platform to see how the orchestration layer connects to your existing rater and AMS.

For related workflows, see our guides on new client onboarding in Applied Epic with DocuSign, Applied Epic vs AMS360 for mid-sized agencies, the e-signature workflow with DocuSign and NowCerts, and the broader state of insurance automation.

FAQs

What is a workers' comp class code?

A workers' comp class code is a numeric code that maps an employer's operations to a base premium rate reflecting the loss experience of similar businesses. NCCI maintains the classification system in most states, and the code drives manual premium before any experience modification is applied.

Can class code lookup be fully automated?

Mostly, but not entirely. The lookup and the rule application can be automated reliably, and clean, well-described businesses can be auto-assigned. Ambiguous descriptions should still route to a human reviewer. The goal is to automate the routine 80% and escalate the genuinely uncertain cases rather than forcing every policy through one path.

Does NCCI provide an API for class codes?

NCCI offers classification tools and data products through its Riskworkstation and related services, which are the authoritative source for class definitions and phraseologies. An orchestration layer consumes that rule set rather than replacing it, so NCCI remains the source of truth for the definitions your automation matches against.

How does this connect to Applied Epic and Tarmika?

Applied Epic stores the policy and insured record, and Tarmika handles multi-carrier WC submission. The automation sits between them — it derives and validates the class code, writes the confirmed value back to Applied Epic, and hands the validated code to Tarmika so the submission carries the right classification from the first quote.

What confidence threshold should we use for auto-assignment?

Start conservative, around 90%, and require a clear gap to the second candidate. Auto-assign only when both conditions hold; route everything else to review. After a few hundred policies you will have data showing where overrides cluster, and you can tune the threshold by industry rather than guessing once and leaving it.

Will automating class codes reduce audit disputes?

It should, because the largest source of disputes is inconsistent human classification and missing documentation. A workflow that applies the same rules every time and logs the reasoning gives you a defensible record at audit, which is often what turns a dispute into a quick resolution.