Credit analysts

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

• Mercantile ledger1841–1920

  Mercantile ledger and correspondent network (Tappan era)

  The Mercantile Agency and its competitor Bradstreet built the first systematic infrastructure for credit assessment in the United States: a network of local correspondents who submitted written character and financial reports, centralized in New York, accessible to paying subscribers. Credit "analysis" was fundamentally qualitative: was the borrower a person of good character? Did they pay their debts? What was their reputation in the trade? R. G. Dun's ledger volumes, which subscribers could consult in person, contained dense handwritten entries on hundreds of thousands of merchants. These ledgers now live at Harvard Business School and represent the most comprehensive record of 19th-century American commercial creditworthiness ever compiled. The methodology was entirely human judgment applied to soft information, with no standardized financial metrics.

  Effect on the work

  The Mercantile Agency model required a large staff of correspondents nationwide plus centralized clerks in major cities. By 1870, R. G. Dun employed thousands of correspondents; the firm's paid subscriber base included most major wholesalers, manufacturers, and banks in the country. This was the first time credit assessment was a scalable, organized profession.

• Standardized financial ratios1919–1955

  Standardized financial ratios and Annual Statement Studies (Alexander Wall / Robert Morris Associates)

  In 1919, Alexander Wall of Robert Morris Associates (founded 1914) published the first systematic industry benchmarking framework for commercial credit: seven financial ratios derived from more than a thousand financial statements across many industries. Wall's work was revolutionary because it gave credit analysts a shared vocabulary and a comparator: instead of asking only "is this borrower honest?" analysts could now ask "how does this borrower's current ratio compare to others in the same industry?" The Risk Management Association (as Robert Morris Associates was later renamed) published the first Annual Statement Studies in 1925, which became the standard reference for commercial lending for the rest of the century. For the first time, credit analysis was a data-informed profession with standardized tools, and the title "credit analyst" gained institutional meaning.

  Effect on the work

  Standardized ratio analysis increased the productivity of individual credit analysts substantially: a trained analyst using Wall's framework could evaluate a credit more efficiently than a pure-judgment predecessor. The professionalization also raised entry requirements: by the 1930s, commercial banks increasingly required college degrees and accounting coursework for credit analyst roles, separating the function from general banking clerks.

• Statistical credit scoring1956–1989

  Statistical credit scoring (Fair Isaac Corporation, 1956 pilot models)

  Bill Fair and Earl Isaac founded Fair Isaac Corporation in 1956 and began developing statistical scoring models for consumer credit decisions at a handful of bank clients. Their initial work replaced the wholly subjective personal-interview model for consumer lending with a quantitative framework: applicant characteristics (income, employment history, payment behavior) were weighted and summed into a numerical score that predicted repayment likelihood. Through the 1960s and 1970s, Fair Isaac refined its algorithms and expanded to more lenders, though the scores remained internal tools. For commercial credit analysts, the immediate effect was limited: scoring models were applicable to consumer and small-business credits but could not replace judgment in middle-market or corporate lending. The deeper effect was structural: scoring displaced the consumer underwriter from most simple loan decisions, permanently separating the role into an automated consumer tier and a judgment-intensive commercial tier.

  Effect on the work

  Statistical credit scoring ultimately reduced headcount for consumer credit underwriting jobs by automating the decision for standard applicant profiles. By the time the FICO score went nationwide in 1989, lenders reported automating most standard consumer loan decisions, shrinking the population of consumer credit analysts relative to the volume of applications processed.

• FICO score nationwide rollout1989–2005

  FICO score nationwide rollout and credit bureau automation (1989-2005)

  The FICO score launched commercially in 1989, available via all three major credit bureaus (Equifax, TransUnion, Experian). The effect on consumer credit analysis was immediate and dramatic: Fannie Mae and Freddie Mac adopted FICO score minimums for conforming mortgage underwriting in 1995, making the score the de facto gateway to the US mortgage market. Auto lenders, credit card issuers, and personal loan providers followed rapidly. By 2000, the majority of consumer credit decisions in the United States were made by automated scoring models that required no human review for standard profiles. The analyst's role in consumer lending shifted from primary decision-maker to model validator, exception handler, and regulator-facing explainability owner. Commercial credit analysis remained largely judgment-driven through this period, though software tools for financial statement spreading (early versions of Moody's Credit Monitor, Banker's Systems) began reducing manual calculation time.

  Effect on the work

  Consumer credit analyst headcount declined relative to lending volume through the 1990s as FICO automation took hold. However, the securitization boom of the 1990s-2000s created a parallel demand surge for structured-credit analysts at rating agencies and investment banks, partially offsetting the consumer-side contraction. The net effect was modest overall employment growth through the late 1990s, with a significant compositional shift from consumer to commercial and structured credit work.

• Spreadsheet-based financial spreading2005–2015

  Spreadsheet-based financial spreading and early credit platforms (Excel, Moody's Credit Monitor)

  Through the 2000s, the dominant tool for commercial credit analysis remained Microsoft Excel: analysts spread financial statements by hand into standardized templates, computed ratios manually, and wrote credit memoranda in Microsoft Word. Early dedicated credit platforms (Moody's Credit Monitor, Baker Hill, Banker's Systems Laser Pro) reduced some of the keying burden but remained separate from data sources, requiring analysts to re-enter figures from PDFs. The 2008 financial crisis was a painful demonstration of the limits of spreadsheet-based analysis: complex structured credits (CDOs, CLOs, CMBS) that had been modeled in fragmented Excel files proved difficult to stress-test at portfolio level when markets moved. Post-crisis, banks invested in more systematic spreading tools and enterprise risk platforms, but the basic analyst workflow (read PDF, key into template, write memo, present to committee) remained largely unchanged through 2015.

  Effect on the work

  Moderate productivity improvement relative to manual ledger methods, but no structural displacement of analyst headcount. The 2008 crisis reduced credit analyst employment at investment banks, structured-credit desks, and securitization-focused roles; commercial banking credit teams at community and regional banks were less affected.

• Machine-learning underwriting platforms2015–2022

  Machine-learning underwriting platforms (Zest AI 2015, Upstart 2012, LendingClub models)

  The mid-2010s brought the first generation of machine-learning consumer credit underwriting platforms to mainstream banking. Upstart (founded 2012, lender partnerships from 2013) and Zest AI (founded 2009 as ZestFinance) applied gradient-boosted models and alternative data signals to consumer credit decisions, claiming approval rates higher than traditional scorecards at the same loss rate. These platforms further displaced human underwriters from the standard consumer credit decision, pushing analysts toward exception handling, model governance, and fair-lending compliance. For commercial credit, the impact was indirect but real: platforms like Numerated (founded 2016) began pre-populating commercial loan applications from business credit bureau data, reducing the manual application-intake phase. CECL (Current Expected Credit Loss) accounting, which became effective for large institutions in 2020, added a new workstream: credit analysts were increasingly involved in supply data for loan-loss reserve models, adding a quantitative modeling dimension to the traditional underwriting role.

  Effect on the work

  Alternative-data machine-learning platforms accelerated the structural shift in consumer credit away from human underwriting toward model-based decisions, with human review reserved for exceptions and regulatory compliance.

• AI-native credit analysis platforms2022–present

  AI-native credit analysis platforms and large language model integration (Moody's CreditLens AI, ChatGPT, Claude, Numerated)

  The 2022-2026 period marked the arrival of AI tools capable of automating the core analytical tasks that still required substantial human time: financial statement spreading (Moody's CreditLens AI auto-extracts data from uploaded PDFs and populates normalized spreading templates), credit memo drafting (analysts at regional banks report using Claude or ChatGPT to generate structured first drafts from spreading outputs, cutting a few hours per memo), and portfolio monitoring (Moody's CreditEdge applies probability-of-default models to continuous market signals). McKinsey research found that first-line commercial bankers spend significant time collecting information, performing analyses, and writing memos, and that generative AI tools can handle data extraction, ratio computation, and first-draft memo sections. Oliver Wyman's Credit Risk Assistant tool demonstrated a 30-40% reduction in credit report generation time. Banks using AI spreading tools reported reductions in manual data-entry time per application. The role is not disappearing; it is compressing toward the judgment-intensive, relationship-facing, and regulatory-compliance functions that AI cannot handle.

  Effect on the work

  Early evidence suggests AI tools increase analyst productivity substantially rather than reducing headcount in a single wave. Banks using Numerated for SMB credits report faster processing, meaning fewer analysts can handle the same volume. BLS projects a modest employment decline through 2034 as these productivity gains outpace any demand growth. The analysts most at risk are those whose workdays are dominated by spreading and memo-templating; analysts who shift toward covenant structuring, relationship management, and model governance are better insulated.