AI Submission Intake: Sixfold vs Federato vs Kalepa

The math on commercial lines submission intake is brutal. A typical commercial underwriter receives 50-80 submissions per week. Of those, maybe 20-30% are in appetite, and only 10-15% will actually bind. That means 85-90% of submissions result in no revenue, but each one still takes 15-30 minutes to review, triage, and decline. For a team of 10 underwriters, that’s 500-800 hours per month spent on submissions that won’t produce premium.

AI submission intake tools exist to fix this ratio. The pitch is simple: let the machine read the ACORD apps, loss runs, and supplemental data, score each submission against your underwriting guidelines, and surface only the in-appetite risks to your underwriters. The out-of-appetite ones get auto-declined or routed to a different appetite.

We evaluated three platforms that take different approaches to this problem: Sixfold, Federato, and Kalepa. Each has genuine strengths, and each requires something different from your organization to work well.

How AI Submission Intake Actually Works

Before comparing the three platforms, it’s worth understanding what they all do at a basic level. The workflow has four stages:

1. Document extraction. The platform ingests submission documents (ACORD applications, loss runs, supplemental questionnaires, SOVs) via email integration, API, or manual upload. OCR and document parsing extract structured data from unstructured documents.

2. Risk characteristic identification. The extracted data is mapped to underwriting-relevant fields: occupancy type, building construction, years in business, loss history, revenue, employee count, and whatever other factors your guidelines consider.

3. Appetite matching. The extracted risk characteristics are compared against your underwriting guidelines. This is the step that separates these platforms from generic AI. You supply the rules; the machine applies them.

4. Scoring and routing. Each submission gets a score (Sixfold uses 1-5, others use different scales) and is routed to the appropriate underwriter, queue, or auto-decline workflow.

The quality of the output depends entirely on step 3: how well your guidelines are encoded. If your underwriting guidelines are a 40-page Word document with phrases like “generally avoid frame construction in wind-prone areas,” the AI will struggle to apply them consistently. If your guidelines are structured as rules (Class code 5812, max TIV $5M, loss ratio below 60%, no more than 2 losses in 5 years), the platforms perform much better.

Comparison Table: Sixfold vs Federato vs Kalepa

Sixfold: Submission Scoring Built for Volume

Sixfold’s approach is the most straightforward of the three. You upload your underwriting guidelines, the platform converts them into scoring rules, and every incoming submission gets a 1-5 score. A 5 means the submission is a strong match for your appetite. A 1 means it’s clearly out of appetite and can be auto-declined.

What Sixfold Does Well

Fast triage on high volume. If your team processes 300+ submissions per week across standard commercial lines, Sixfold’s scoring system gets the best risks to the front of the queue. Underwriters see the 4s and 5s first, work the 3s when they have time, and the 1s and 2s get declined automatically or routed to a referral queue.

Clear scoring logic. Each score comes with an explanation of which guideline factors drove it up or down. An underwriter can see “Score: 4. Positive: class code in appetite, 7 years in business, clean loss history. Negative: TIV slightly above preferred range.” That transparency matters for adoption; underwriters won’t trust a black box.

Auto-decline with templates. For submissions scoring 1-2, Sixfold can send templated decline communications to the broker. This alone saves significant time. A team declining 40 submissions per week at 10 minutes each gets back 6-7 hours weekly.

ACORD extraction accuracy. In our testing, Sixfold correctly extracted data from ACORD 125/126/130/140 applications about 88% of the time without manual correction. Handwritten ACORD apps dropped that to about 65%.

Where Sixfold Falls Short

No portfolio context. Sixfold scores each submission in isolation. It doesn’t know or care that you’re already concentrated in Florida frame construction or that your WC book is heavy on class code 8810. If you need to manage portfolio concentration, Sixfold won’t help.

Commercial lines only. Sixfold is built for standard commercial lines. If your book includes specialty, excess, or surplus lines, the scoring models aren’t trained for those risk types.

Guideline maintenance. Your score accuracy is only as good as your encoded guidelines. If your appetite changes quarterly (as it does at many carriers), someone needs to update the rules in Sixfold. The platform doesn’t auto-learn; it applies the rules you give it.

Federato: The Portfolio-Aware Approach

Federato takes a fundamentally different approach. Instead of scoring individual submissions against static guidelines, it scores submissions against your entire portfolio. The platform’s core question isn’t “is this risk in appetite?” but rather “does adding this risk improve or worsen our portfolio’s risk profile?”

What Federato Does Well

Portfolio-level optimization. Federato’s RiskOps platform knows your current book composition. When a new submission arrives, it evaluates not just the individual risk characteristics but how that risk fits within your portfolio’s concentration, geographic exposure, and aggregate limits. A Florida property submission might score well individually but poorly if you’re already concentrated in Florida wind exposure.

Dynamic appetite. Because the scoring is portfolio-aware, your effective appetite adjusts automatically as your book changes. Early in the year when you need to grow, more submissions score favorably. As you approach aggregate limits or concentration thresholds, the platform tightens its scoring. This is something static guideline-based systems can’t do.

Claims reported 89% time-to-quote reduction. Federato claims its platform reduces the time from submission receipt to quote delivery by 89%. We couldn’t independently verify this number across a large book, but the logic is sound: if underwriters only see pre-qualified, portfolio-appropriate submissions, they can move faster on each one.

Cross-line visibility. For carriers writing multiple lines (property, casualty, specialty), Federato can show how a submission affects the overall portfolio across lines, not just within the specific line being quoted.

Where Federato Falls Short

No auto-decline. Federato routes and prioritizes, but it doesn’t auto-decline out-of-appetite submissions. The philosophy is that underwriters should see everything and make decisions; the platform just helps them decide faster. If you’re looking for automated triage that removes low-scoring submissions from the workflow entirely, Federato doesn’t do that.

Longer deployment. Because Federato needs to ingest your entire portfolio (not just your guidelines), deployment takes 8-12 weeks. You’ll need to provide historical policy data, claims data, and portfolio composition data before the system can score new submissions in context.

Higher complexity. Federato is a more complex tool to configure and maintain. The portfolio optimization features require ongoing data feeds (new bindings, claims, renewals) to keep the portfolio model current. A team without dedicated analytics support may find this overhead significant.

Pricing. Federato’s annual license pricing reflects the platform’s complexity. It’s designed for carriers with $100M+ in premium where the portfolio optimization pays for itself. A $30M MGA writing standard commercial lines would likely find the ROI harder to justify.

Kalepa: Specialty Lines Focus

Kalepa was built specifically for specialty and hard-to-place lines where the underwriting complexity is higher and generic scoring models don’t work. Workers compensation, fleet auto, and professional liability are Kalepa’s primary focus areas.

What Kalepa Does Well

Specialty lines depth. Where Sixfold and Federato work across broad commercial lines, Kalepa goes deep on specific specialty segments. For workers compensation, the platform can ingest medical records, OSHA logs, and experience modification worksheets alongside the standard ACORD application. For fleet auto, it processes MVRs and vehicle schedules.

Appetite guardrails. Kalepa’s guardrails system flags submissions that fall outside defined appetite parameters but explains exactly why. For a WC submission, it might flag: “Governing class code 5183 (plumbing) is in appetite, but 38% of payroll is in class 5190 (electrical) which is restricted.” That specificity helps underwriters decide whether to make an exception.

Loss run automation. Kalepa’s loss run parsing is particularly strong. The platform extracts individual claims from carrier loss runs, categorizes them by type, calculates development factors, and presents a loss analysis summary. For specialty lines where loss history is the most important underwriting factor, this saves significant analysis time.

Integration with appetite adjustments. When your specialty appetite changes (e.g., you stop writing WC for roofing contractors in Texas), the guardrails update immediately and begin flagging out-of-appetite submissions. The feedback loop between appetite changes and submission scoring is tight.

Where Kalepa Falls Short

Narrow line focus. Kalepa is not a general commercial lines platform. If you write a diversified book with BOP, property, GL, WC, auto, and umbrella, Kalepa only helps with the specialty portions. You’d need a second platform for the standard commercial lines.

Smaller integration ecosystem. Kalepa integrates with Applied Epic and offers custom API connections, but the integration list is shorter than Sixfold or Federato. If you’re on Duck Creek or Guidewire, integration may require custom development.

Less market maturity. While Kalepa has been around since 2018, its market presence is smaller than Sixfold or Federato. Fewer industry case studies and fewer published customer references make it harder to evaluate production-scale performance before committing.

Integration Reality

The marketing materials for all three platforms list impressive integration partners. The reality is more nuanced.

What Actually Connects

“Native integration” means the vendor has built and tested the connection. “API-based” means there’s an API you can connect through, but you (or your IT team) are doing some of the work. Email ingestion is universally available and is how most smaller operations start.

What All Three Need From You

Every platform requires the same foundation to work well:

1. Machine-readable underwriting guidelines. Not a narrative Word document, but structured rules: class codes in/out of appetite, acceptable loss ratios, TIV ranges, geographic restrictions, prohibited operations. The more structured your guidelines, the better any platform performs.

2. Clean data feeds. Historical policy and claims data for calibration (especially Federato), ongoing submission flow, and ideally binding/decline outcome data so the platform can measure its own accuracy.

3. Designated owner. Someone on your team needs to own the platform: update guidelines when appetite changes, review scoring accuracy, manage broker feedback on auto-declines, and troubleshoot extraction errors. This is typically 5-10 hours per week for a mid-size carrier.

Honest Limitations of All Three

None replace underwriter judgment on complex risks. A $50M manufacturing risk with environmental exposure, foreign operations, and a complex loss history will get scored by all three platforms, but the score is a starting point, not a decision. Experienced underwriters add context that no scoring model captures.

Extraction accuracy varies by document quality. Clean, typed ACORD apps extract at 85-90% accuracy. Handwritten apps, non-standard supplemental forms, and scanned loss runs with poor image quality drop accuracy to 60-75%. All three platforms require a human verification step for extracted data.

Auto-decline creates broker relationship risk. Sixfold and Kalepa both offer auto-decline features. While this saves time, an auto-declined submission that should have been written (because the scoring missed a nuance) damages broker relationships. Most carriers start with manual review of decline recommendations and move to auto-decline only after 3-6 months of validated accuracy.

Guideline encoding is the hard part. The platforms themselves work. The challenge is translating underwriting appetite from the informal, experience-based knowledge in your senior underwriters’ heads into structured, machine-readable rules. This is an organizational challenge, not a technology challenge, and it takes longer than most carriers expect.

Verdict by Buyer Type

High-Volume Standard Commercial Lines: Sixfold

If your team processes 200+ standard commercial submissions per week and your primary goal is faster triage and auto-decline of out-of-appetite risks, Sixfold is the most direct solution. It does one thing well: score submissions against your guidelines and route accordingly. The 4-8 week deployment is the fastest of the three, and the per-submission pricing model works for organizations that want to start small.

Portfolio Optimization for Large Carriers: Federato

If you’re a carrier with $100M+ in written premium and your challenge is portfolio concentration, aggregate management, and making sure every new binding improves your overall book, Federato is the right tool. The portfolio-aware scoring is a fundamentally different (and more sophisticated) approach than individual submission scoring. The longer deployment and higher cost reflect the deeper integration with your portfolio data.

Specialty Lines with Complex Underwriting: Kalepa

If your book is concentrated in workers compensation, fleet auto, or professional liability, and your underwriters spend significant time on loss run analysis and class code verification, Kalepa’s specialty-focused approach will deliver more value than a general commercial platform. The loss run automation alone justifies evaluation for heavy specialty writers.

Not Sure Yet: Start with Email Triage

All three platforms support email ingestion as a starting point. Before committing to a full deployment, start by routing submission emails through the platform for 60-90 days. Compare the platform’s scoring against your underwriters’ actual decisions. If the scoring agrees with your underwriters 80%+ of the time, you have a foundation for automation. If it’s below 70%, your guidelines need work before any platform will deliver value.

What We’d Ask in a Demo

If you’re evaluating any of these platforms, here are the questions that reveal the most about real-world performance:

1. What’s your extraction accuracy on handwritten ACORD apps? Everyone claims 90%+ on clean forms. Handwritten forms are the real test.
2. Can I see the scoring logic for a specific submission? Transparency matters. If the platform can’t explain why a submission scored a 3 instead of a 4, your underwriters won’t trust it.
3. What happens when a submission is missing data? Does the platform flag it as incomplete, guess at the missing fields, or score it with a lower confidence?
4. How do you handle appetite changes? Can I update guidelines myself, or do I need your professional services team?
5. What’s the average time from signing contract to production use? Not pilot, not testing; actual production with live submissions.

The answers to these questions will tell you more about the platform’s real-world readiness than any demo deck.