Insurance Agencies Run on Spreadsheets: What an AI Audit Reveals

An insurance agency owner I audited last year had a full-time employee whose primary job was logging into carrier portals. Eight portals. Every Monday morning, she started at 7:30 AM and finished the data pull by noon if nothing went wrong. If a portal had updated its login flow or added a new authentication step, it took longer. This was her core function - not serving clients, not supporting producers, not anything that created agency value. Four and a half hours per week, every week, pulling renewal dates and commission statements that could not be retrieved any other way.

That is not an edge case. It is the default state of a mid-size P&C agency in 2026.

When I explain what an AI audit looks at for an insurance agency, I start with this: the most valuable AI applications for agencies require centralized, clean, current data. The single biggest obstacle to that is a carrier ecosystem that was never designed for data interoperability. Fixing that gap - or working around it strategically - is what the audit is for.

The Portal Problem: Why It Exists and What It Costs

Most independent agencies represent 6-12 carriers. Progressive, Nationwide, Travelers, Hartford, Markel, Employers, Cincinnati - the list varies by line of business and agency specialty. Each carrier has its own portal. Each portal has its own login, its own data structure, its own export formats (if export exists at all), and its own update cadence.

None of these portals have public APIs. Full stop. There are isolated exceptions - some carriers have developed limited API access for high-volume agency partners or aggregators, and some are building out integration capabilities slowly. But for the typical independent agency today, the carrier portal is a human-operated interface that requires manual login, navigation, and data extraction.

The time cost is immediate and measurable. The agencies I audit typically have 25-40% of administrative staff time dedicated to portal-related work: pulling renewal lists, checking policy status, downloading loss runs for remarketing, reconciling commission statements. For a 10-person agency with three admin staff at $45,000-$55,000 annual salary, that is $33,000-$66,000 in annual labor cost for work that produces no client value. It just moves data from where the carrier stores it to where the agency needs it.

The intelligence cost is less visible but larger. Renewal predictions, cross-sell triggers, at-risk client identification, commission forecasting - every AI capability that an insurance agency should want requires aggregated carrier data that nobody has bothered (or been able) to centralize. The spreadsheet tracking renewals is not searchable. It is not connected to the policy management system. It does not trigger any workflow. It just exists on a shared drive, updated manually, used reactively.

The Systems Layer: What Is Connectable and What Is Not

Let me go through the actual platforms an independent agency typically runs on, and give direct assessments of connectivity.

Applied Epic is the dominant agency management system for mid-to-large independent agencies. It has a mature REST API. Policy data, client records, producer data, activity history, and accounting information are all accessible via the API. Applied Epic is a genuine connectivity point - if your agency management system is Epic, you have a solid foundation for building on top of it.

Hawksoft is popular among smaller and regional independent agencies. It also has a REST API. More limited in scope than Epic’s, but the core objects - clients, policies, renewals - are accessible. Functional connectivity exists.

AgencyZoom is a CRM and workflow tool built specifically for insurance agencies, used alongside the AMS rather than replacing it. It has integration capabilities with some AMS platforms and carrier APIs where they exist. Its value for AI is in the workflow and communication data it captures - quote activity, outbound calls, follow-up sequences.

EzLynx serves as both a comparative rater and, for some agencies, a basic AMS. API access exists for the comparative rating components. The AMS functionality is more limited from an API standpoint.

Now the hard truth: none of this AMS connectivity solves the carrier data problem. Your Applied Epic database contains what your staff entered into it. Renewal dates that were pulled from carrier portals and keyed in. Commission figures that were reconciled from portal downloads and manually posted. Policy status updates that were checked on the portal and recorded in the system. The AMS is not a live reflection of carrier data. It is a record of what someone checked recently.

Carrier Portals: The RPA vs. API Question

The AI blueprint for an insurance agency has to directly address the carrier data problem - and there is no simple solution. The audit’s systems assessment produces a carrier-by-carrier breakdown that classifies each carrier into one of three categories.

API available (rare, conditional). A small number of carriers have developed API access, typically for agency management system vendors with formal partnership programs. Applied Epic has carrier integrations with select carriers that allow direct data pulls. If your carriers are among the supported integrations, this is the cleanest path.

RPA feasible. Robotic Process Automation - software that operates carrier portal interfaces like a human would - can automate the manual portal work for carriers where web automation is technically stable. This is not the same as an API. It is brittle: if the carrier changes their portal layout, the bot breaks. It requires maintenance. But for carriers with stable portal interfaces where API access does not exist, RPA reduces 4-hour Monday morning processes to automated overnight scripts. The time savings are real even if the solution is imperfect.

Manual - no good option. Some carrier portals actively block automation attempts, require multi-factor authentication that defeats RPA, or have interfaces that change frequently enough to make RPA uneconomical. For these carriers, the audit documents the manual workflow cost and recommends requesting API access through the carrier’s agency relations contact - a request that occasionally succeeds for large-volume agencies.

The carrier-by-carrier classification is one of the most practical deliverables in the systems assessment. Agencies often have assumptions about which carriers can be automated - usually wrong - and the actual assessment changes how they prioritize the work.

The Data Layer: What Exists and What Does Not

Stepping back from the carrier portal problem: what data does a typical insurance agency actually have, and where does it live?

Policy data. In the AMS, entered at bind and updated when changes are made. Reasonably complete for bound business. Accuracy depends on how consistently staff enter changes they learn about.

Renewal pipeline. Typically a spreadsheet maintained by whoever owns the renewal process. Sometimes also in the AMS with renewal dates populated. The spreadsheet exists because the AMS’s renewal workflow features are often not set up to match how the agency actually works.

Loss runs. Downloaded from carrier portals as PDFs. Stored in a client folder on the shared drive. Unstructured, unsearchable, not connected to anything. When a producer needs the loss run for a commercial lines remarketing, they find the PDF and send it manually.

Commission data. Reconciled monthly from carrier statement PDFs and portal downloads. Posted to accounting software (QuickBooks, usually). The reconciliation process is a major source of admin time and occasional error.

Client communication history. In AgencyZoom (if the agency uses it), in email, in the AMS activity log. Variable completeness. High-value client relationships often exist mostly in the producer’s personal history and institutional memory.

Prospect data and quote history. In the CRM or AMS. Variable completeness based on how disciplined producers are about logging activity.

The pattern: structured policy data exists in the AMS, but the surrounding intelligence - renewal risk signals, cross-sell opportunities, client relationship context - is either in spreadsheets, in portals, or in people’s heads.

What the Audit Maps and What It Finds

The process map for an insurance agency follows the client lifecycle: prospect to quote, quote to bind, bind through renewal, renewal through potential churn or coverage change. Each stage gets mapped for data creation, data storage, and data flow.

The quantified waste report for a typical 10-20 person P&C agency consistently surfaces four major categories:

Portal data retrieval time. As discussed: 25-40% of admin hours. Quantified in annual labor cost.

Manual renewal pipeline management. Spreadsheet maintenance, reminder chasing, follow-up tracking that does not connect to outbound communication tools. Time is moderate but the miss rate - renewals that fall through cracks due to spreadsheet inaccuracies - carries real revenue consequence.

Loss run and remarketing preparation. Pulling, organizing, and sending loss run PDFs for commercial lines remarketing is largely manual. For agencies with significant commercial books, this represents meaningful producer and CSR time every month.

Commission reconciliation. The monthly process of matching carrier statement PDFs to expected commissions and posting to accounting. Errors are common. Disputes require going back to portal data. Time cost is 8-15 hours per month for mid-size agencies.

The Intelligence Layer: What Becomes Possible

Once carrier data is centralized - via API where available, RPA where feasible, and clean manual entry discipline where neither exists - the AI applications that actually move the needle become buildable.

Renewal prediction and at-risk identification. Using policy history, claims activity, client engagement signals, and market pricing trends to identify which accounts are at risk of non-renewal or carrier non-renewal before the 90-day window. This is the most consistent producer use case I see agencies want.

Cross-sell triggers. When a commercial client adds a vehicle, when a personal lines client has a life event, when a claims pattern suggests a coverage gap - surfacing these moments requires data from the AMS, the carrier, and the communication history to be in one place.

Commission forecasting. With carrier commission data flowing into a central system rather than being reconciled monthly from PDFs, agencies can build real-time commission reporting and forecasting. This matters for producer compensation planning and agency financial management.

Automated remarketing triggers. When a commercial lines account hits 120 days to renewal and has a loss ratio above threshold, the system triggers a remarketing workflow automatically - requesting loss runs, initiating the quoting process, scheduling producer follow-up. Right now, this trigger is a person looking at a spreadsheet.

None of these capabilities require novel AI technology. They require centralized, current, structured data. The audit’s job is to tell you how far away from that baseline you are and what the path looks like.

For a foundational readiness assessment before engaging with any AI project, the AI readiness audit guide covers the four-pillar framework across process, data, team, and budget. For agencies also handling financial advisory or wealth management services, the AI audit for financial advisory firms covers where the regulatory and data constraints look different from P&C operations.

Frequently Asked Questions

Our AMS has “carrier integrations” listed in the features - does that mean the portal problem is solved?

Not usually. AMS carrier integrations typically cover a limited scope: real-time quoting in some cases, basic policy status for select carriers, and download transactions for new and renewed business. They do not typically cover renewal pipeline data, commission statements, loss runs, or comprehensive account-level reporting. The features listed often require separate carrier enrollment and have data scope limitations. The audit maps what your specific carrier set actually supports through the AMS integration versus what still requires portal access.

Is RPA reliable enough for production use in an agency environment?

For carriers with stable portal interfaces, yes - with the right maintenance approach. RPA bots that handle portal data extraction need to be monitored and updated when portals change. Budget for 4-8 hours per quarter of RPA maintenance per carrier. The break-even math is simple: if a carrier portal takes 45 minutes per week to pull manually, that is 39 hours per year. If RPA maintenance costs 4 hours per year, you come out ahead by 35 hours regardless of what you pay for the RPA platform.

We are a smaller agency with 3-5 staff - is an AI audit relevant at this size?

The portal problem is proportionally worse at smaller agencies because you have fewer staff to absorb the manual work. A 3-person agency where one person spends 30% of their time on portal data entry is a bigger problem than a 20-person agency where the same task is shared across a dedicated admin team. The audit at this size is typically faster and less expensive - 1-2 weeks versus 3-4 weeks - because the systems are simpler. The findings often have disproportionate impact because the time recovered goes directly to producer activity.

Our top producer tracks everything in their head and email. How do we address that in the audit?

The stakeholder interview with that producer is one of the most valuable parts of the audit process. What we find consistently: the producer has a sophisticated mental model for identifying renewal risk, cross-sell moments, and relationship management. The audit goal is to make that model explicit, document it as decision logic, and then assess whether that logic can be encoded into the AMS or CRM with appropriate data to trigger it automatically. High-performing producers are not opposed to AI when they understand it is amplifying their approach rather than replacing it.

What should we expect to spend to fix the carrier data problem before AI work starts?

Rough ranges, depending on agency size and carrier mix: an RPA deployment covering 5-8 carriers with ongoing maintenance runs $8,000-$20,000 per year for most mid-size agencies (build cost plus platform fees plus maintenance). AMS API integration for data aggregation where available adds $5,000-$15,000 in integration development. Data hygiene in the AMS (cleaning up policy records, standardizing renewal dates, building the renewal pipeline view that currently lives in a spreadsheet) is internal labor: typically 40-80 hours of focused effort from whoever owns the AMS. The audit quantifies what the current manual work costs annually and compares it against the remediation investment. For agencies spending $40,000+ per year in admin labor on portal work, the math for automation is straightforward.