Insurance claims clerks

Three parts of your work where AI is already doing real lifting, and what stays yours.

• Hand-written ledgers1850–1890

  Hand-written ledgers, index card files, and policy registers

  From the mid-1800s through the 1880s, every insurance transaction was recorded by hand. Clerks maintained policy registers: large bound ledgers where each policy was entered in sequence with the policyholder's name, address, premium amount, coverage period, and sum insured written in careful copperplate. A parallel index by policyholder surname allowed retrieval. When a claim arrived, the clerk pulled the original policy from the physical file and verified coverage against the ledger entry. Premium calculations used printed actuarial tables and were performed with pencil and paper. Error rates were significant, and large carriers employed teams of checking clerks whose sole job was to verify the arithmetic of the entry clerks.

• Typewriter1880–1920

  Typewriter and carbon paper for policy documents

  The commercial typewriter (Remington No. 1, 1874; mass adoption from early 1880s) transformed insurance clerical work within a decade. Typed policies were legible, reproducible via carbon copies, and far faster to produce than handwritten documents. Insurance companies were among the earliest corporate adopters: the Equitable Life Assurance Society was already operating typewriter pools by the early 1880s. Carbon paper meant that a clerk could produce three copies simultaneously (original for the policyholder, file copy for the carrier, and agent copy) replacing the re-copying that had been a major fraction of clerical labor. The typewriter also enabled the standardized printed policy form, which pre-filled boilerplate and left only variables to be typed.

• Hollerith punch card tabulation1900–1965

  Hollerith punch card tabulation (IBM / Powers)

  The US insurance industry was one of the earliest and most enthusiastic adopters of Hollerith electric tabulating machinery. The Mutual Life Insurance Company of New York was using Hollerith machines to tabulate actuarial data as early as 1895, just five years after Hollerith first commercialized the system for the 1890 Census. By the 1920s, major carriers were running entire premium accounting and policy-register functions through punch card tabulating rooms. A policy or claim would be punched onto cards; the tabulating machine would sort, count, and print totals, replacing dozens of hand-tallying clerks. This was the first wave of true mechanization of insurance clerical work, not all clerks were replaced, but the ratio of output per clerk rose sharply, and the punch card operator became a distinct insurance occupation.

  Effect on the work

  IBM's own historical records note that insurance companies were the largest single industry customer for tabulating equipment in the 1920s-1940s. The efficiency gains meant that carrier premium volumes could grow substantially without proportional headcount growth: the first structural compression of the ratio of clerks to policies.

• IBM System/360 mainframe1964–1985

  IBM System/360 mainframe: batch policy administration

  IBM's System/360 announcement in April 1964 opened the mainframe era for insurance. Within five years, most major US carriers had installed mainframe policy administration systems (CYBERLIFE [life insurance], early precursors to what became OPUS and other property-casualty platforms) that stored all policy records in magnetic tape and disk files and ran daily batch jobs to process premium payments, generate renewal notices, and produce claims payment runs. The clerk's job shifted: instead of manually posting premiums to ledger cards, clerks now keyed transactions into dumb terminals whose inputs fed the batch queue. Mainframes dramatically reduced the labor per transaction while enabling carriers to grow policy volume without proportional headcount growth.

  Effect on the work

  Insurance industry employment studies from the 1970s documented that mainframe adoption held clerical headcount roughly flat even as policy volumes doubled between 1964 and 1980, implying roughly a productivity gain per clerk over the period.

• Online policy administration systems1975–1995

  Online policy administration systems (VDTs and early PAS terminals)

  Through the late 1970s and 1980s, carriers moved from pure batch processing to interactive online access: clerks now sat at video display terminals (VDTs) that provided real-time lookup of policy records, claim status, and coverage details without waiting for the overnight batch run. Systems like CICS (IBM Customer Information Control System) enabled transaction-level updates to policy databases during business hours. For the first time, a clerk handling an inbound inquiry could pull up the exact policy record in seconds rather than retrieving a physical file or waiting for a printed report. The "green screen" became the dominant insurance clerical tool for twenty years.

• ACORD electronic data interchange1988–2005

  ACORD electronic data interchange (EDI) for claims submission

  ACORD (Association for Cooperative Operations Research and Development), founded in 1970, published its first standardized insurance electronic data interchange formats in 1988, enabling agents, brokers, and carriers to exchange policy and claims data electronically rather than by paper mail or fax. The ACORD AL3 format for life/health claims and the ACORD XML standards (late 1990s) eliminated large classes of manual keying: a claim submitted through an ACORD-compliant agency management system could flow directly into the carrier's claims platform without re-entry. This was the first significant elimination of re-keying labor: the most repetitive and error-prone part of the claims clerk's day.

• Document imaging1995–2010

  Document imaging, OCR, and workflow routing systems

  The mid-1990s brought document imaging platforms to insurance: paper claims, medical bills, police reports, and application forms were scanned on entry and stored as digital images in document management systems (FileNET, Hyland OnBase). Early OCR (optical character recognition) software could extract typed text with reasonable accuracy, reducing manual re-keying. Workflow routing software, rules engines that read incoming documents and assigned them to the appropriate claims queue, replaced the physical in-box routing that had required a clerk to physically sort and walk paper to the right desk. This generation of technology eliminated the "mail room" tier of clerical work but left intact the verification and data-entry layer.

• Guidewire ClaimCenter2001–2018

  Guidewire ClaimCenter and Duck Creek: modern core claims platforms

  Guidewire Software (founded 2001; IPO 2012) and Duck Creek Technologies (founded 2000) built the next generation of insurance core systems: cloud-capable policy and claims administration platforms that replaced the green-screen mainframe systems. ClaimCenter embedded workflow rules that automatically assigned incoming claims to the appropriate adjuster queue, verified coverage against the live policy database, calculated reserve amounts based on loss type, and generated required correspondence on a schedule. The claims clerk's job narrowed: instead of operating the full intake-to-payment workflow manually, clerks now managed exceptions, the cases the platform's rules could not automatically route.

• Computer vision damage assessment2014–2022

  Computer vision damage assessment (Tractable, CCC AI)

  Tractable (founded 2014 in London by Alex Dalyac and Razvan Ranca) applied convolutional neural networks to automobile damage photos, building an AI that could generate repair cost estimates from smartphone photos faster and often as accurately as a trained human estimator. By 2024, Tractable's technology was deployed by Tokio Marine, Ageas, and Admiral Group and was processing millions of claims per year. These tools eliminated the photo-review-and-estimate step that had been one of the primary tasks of the auto claims processing clerk.

  Effect on the work

  Tractable and CCC's AI assessment tools automate the photo-to-estimate step that previously required a human clerk to review an adjuster's worksheet and key in the repair line items.

• Lemonade "AI Jim"2016–2023

  Lemonade "AI Jim": end-to-end instant claims payment

  On December 27, 2016, Lemonade Insurance reported that its AI claims bot, internally called "AI Jim", had paid a stolen-coat claim in three seconds, without any human involvement: it reviewed the claim, cross-referenced the policy, ran several anti-fraud algorithms, approved the claim, and wired the payment before deleting the paperwork. Lemonade (founded 2015; IPO 2020) built its entire claims stack around AI-first processing, with human adjusters handling only complex or fraud-flagged claims. For simple renters and homeowners claims, "AI Jim" eliminated the claims processing clerk entirely, not augmented, eliminated. By 2022, Lemonade reported that sizeable share of all its claims were settled instantly without human involvement.

  Effect on the work

  Lemonade's architecture demonstrates that for a significant subset of personal-lines property claims, the claims processing clerk role is fully automatable today, not in the future. The constraint is not technology but regulatory environment, claim complexity, and carrier risk tolerance for AI errors.

• IDP + STP straight-through processing2020–present

  IDP + STP straight-through processing (Hyperscience, Ocrolus, Duck Creek)

  Intelligent Document Processing (IDP) platforms, Hyperscience and Ocrolus chief among them, combined deep-learning OCR with classification and extraction models to pull structured data from claim forms, police reports, explanations of benefits, and medical records. Paired with straight-through processing (STP) rule engines in Guidewire and Duck Creek, these platforms enabled end-to-end automation of claims that met carrier-defined criteria: intake, policy verification, coverage check, payment calculation, and payment issuance, with no human touch. Routine form intake, data extraction, and routing represent the exact workload IDP and STP target.

  Effect on the work

  BLS projects employment decline for 43-9041 in the 2023-2033 period, explicitly citing software automation of routine claims processing. This is a faster projected decline than bookkeeping clerks and one of the steepest in the office and administrative support major group.

• Generative AI for claims correspondence2023–present

  Generative AI for claims correspondence, FNOL triage, and Q&A bots

  From 2023, major carriers began deploying large language models for claims-related tasks that had persisted as human work: drafting status acknowledgement letters, coverage denial notices, and reservation-of-rights correspondence; triaging incoming first-notice-of-loss (FNOL) reports to assign priority and route to the appropriate adjuster queue; and operating policyholder-facing chatbots that could answer claim status questions without clerk involvement. Verisk Analytics, Snapsheet, Sixfold, and CCC Intelligent Solutions all launched generative AI features within their claims platforms by 2024. The remaining human role in routine claims processing narrowed to exception handling, fraud-indicator review, regulatory correspondence verification, and empathic handling of distressed policyholders.