EZLynx and Vertafore both document useful built-in AI on their own pages: EZLynx's EVA drafts emails, summarises accounts and answers routine client texts, and Vertafore's Velocity agents handle policy intake, policy checking and document research. Before switching, ask which features are generally available in your version, who approves changes, and exactly what you can export.
The vendors publish impressive time-saving figures, but nearly all of them are "up to" numbers from early testing on one task, and neither vendor lists a price on the pages I read. That makes data export, not AI, the part of the decision you can verify before you commit. Score the two together, on what each vendor states in writing as of November 2026 rather than on what a demo implies.
What EZLynx says its AI does
EZLynx, which describes itself as an Applied company, calls its assistant EVA (EZLynx Virtual Assistant). The first set of features was announced on 30 January 2025, and a further batch followed on 8 October 2025 at Applied's annual user conference. The table lists each item with the vendor's own status wording.
| Feature | What the vendor says it does | Status wording |
|---|---|---|
| EVA Email Content Assist | Drafts client and prospect emails from prompts about subject, tone and length; you review before sending | Announced January 2025 |
| EVA account summaries | Condenses emails, texts and agent notes from the past week, month or three months into a client history summary, with suggested next steps | Announced January 2025 |
| EVA summaries in your inbox | The same summaries inside Outlook and Gmail, including whole email threads | "Now in Outlook and Gmail" (October 2025) |
| EVA Product Assistance | Answers how-to questions inside the platform with step-by-step instructions | Announced January 2025 |
| Automated and after-hours text replies | Watches incoming client texts and answers routine requests (ID cards, coverage, vehicle or driver details), including outside business hours | After-hours replies announced October 2025; listed on the AI page now |
| Quote application autofill | Pulls data from PDFs and handwritten notes to fill a quote application | Beta (October 2025) |
| Coverage recommendations | Suggests lines of business and coverages for less familiar commercial risks from industry classification codes | Announced October 2025; availability not stated |
| EVA Recon Import | Reads carrier commission statements in several file formats and imports them, so nobody re-keys the figures | Listed on the AI page; no date given |
Keep the scale in mind. Email drafting and summarising save minutes per person, worth having but unlikely to justify a migration on their own. Commission-statement import can be a bigger block of work, especially with many carriers, so if it matters to you, ask to see it run on two of your own carrier statements. The pages I read do not state pricing or plan tiers for EVA.
What Vertafore says its agents do
Vertafore announced its "Agentic Agency" strategy on 23 September 2026, built on its Velocity AI Platform, with four agents named at launch. The details below come from Vertafore's own press releases from September and October 2026. All the percentages are the vendor's "up to" figures from early adopter testing.
| Agent | Where it runs | What it does | Reported status |
|---|---|---|---|
| Document Intelligence Agent | ImageRight | Pulls relevant details from documents into one answer, with document and page citations, so staff do not review files one by one | Generally available 28 September 2026; up to 90% less review time in early testing |
| Coach Agent | AgencyOne with AMS360, and Sagitta | Answers how-to questions from Vertafore's documentation and, once configured, your own standard operating procedures | Available from 20 October 2026; up to 88% less time finding answers |
| Policy Intake Agent | AgencyOne with AMS360; early access for Sagitta | Extracts policy details from PDFs, declarations pages and binders, maps them to AMS fields, flags new, changed, missing or conflicting items | Generally available 15 October 2026; you approve before it updates records; up to 87% less time on a typical commercial property and liability package |
| Policy Check Agent | AMS360 | Compares issued policies with AMS data and source documents, recommends corrections, logs review history | Generally available 15 October 2026; you approve corrections first; up to 90% less checking time per line of business |
Two more, a Data Collection Agent and a Carrier Routing Agent, are due to enter early adopter programmes in winter 2026-27. Vertafore's April 2026 announcement also introduced a Reconciliation Agent (it ingests carrier statements, matches transactions and flags exceptions in AMS360) and an Email Agent (it turns incoming Outlook emails into actions in AMS360), both with general availability expected in summer 2026. The pages I read do not confirm when those two went live, so ask which agents run in the exact product and version you would use.
The approval step matters more than the percentages. Vertafore's releases say users approve proposed changes before the agent writes to the system, and that the review history is logged. That is the design you want from any system that edits policy records, and it is what to demand in a demo from every vendor.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowA scorecard that puts AI and exit on one sheet
The mistake is scoring AI first and treating export as an afterthought. Put both on one sheet with weights that add to 100, and agree the weights before the demos so a good presenter cannot move them. Here is an illustrative agency, twelve people, scoring two unnamed systems out of 5. They are not EZLynx or Vertafore; use the structure, not the numbers.
| Criterion | Weight | System A | System B | A weighted | B weighted |
|---|---|---|---|---|---|
| AI features generally available for your tasks | 15 | 4 | 3 | 60 | 45 |
| Human approval and audit trail on AI actions | 10 | 3 | 4 | 30 | 40 |
| AI data handling (training, provider, switch-off) | 10 | 3 | 3 | 30 | 30 |
| Export completeness (policies, contacts, notes, documents) | 20 | 2 | 4 | 40 | 80 |
| Export mechanics (format, fee, timing, API) | 10 | 2 | 3 | 20 | 30 |
| Fit with your carriers, rating and accounting | 15 | 5 | 4 | 75 | 60 |
| Migration support and cost | 10 | 3 | 3 | 30 | 30 |
| Price and contract terms | 10 | 4 | 3 | 40 | 30 |
| Total (out of 500) | 100 | 325 | 345 |
System B wins by 20 points although its AI features are less mature. It gains 60 weighted points on export completeness (40), export mechanics (10) and the audit trail (10); System A's 40-point lead on AI, carrier fit and price is not enough to cover them. If you disagree, change the weights, but do it before you see the demos. An agency that scores AI at 15 per cent and export at 30 per cent is saying something about its priorities, and should be able to defend it to its partners.
The export test: leave before you have joined
Ask each vendor for a sample export of five test clients, or run one on your current system to see what you would be asking the new vendor to import. Then open the files yourself. This is a filled-in result from an illustrative test, with the fixes you would ask for.
| Item | What came back | What to ask for |
|---|---|---|
| Contacts | CSV with names, addresses, phone, email | Nothing; usable |
| Policies | Current terms only; endorsement history missing | Full transaction history, or a statement of what is not exportable |
| Notes and activities | One long text field per client | Date, author and type kept as separate fields |
| Documents | Flat folder, files named by internal ID | Folder per client, filenames with policy number and date, plus a mapping file |
| Certificates and tasks | Not included | Written list of what is excluded and why |
If a vendor will not run a sample export before you sign, record that as a low score on export mechanics. Also ask what the export costs, how long it takes, and how long after cancellation you can still request it. These are contract questions, so ask your adviser to read the clause rather than relying on a sales call.
Demo requests that expose weak AI
Do not accept a scripted demo. Send the vendor five requests in advance, using documents from your own book with names removed, and score the live result.
| Ask the vendor to | A good answer | A warning sign |
|---|---|---|
| Run intake on a messy declarations page from one of your carriers | Shows proposed fields beside the source text, flags what is uncertain | Uses a clean sample and will not run yours |
| Show what the agent does when two documents conflict | Flags the conflict and asks a human | Picks one silently |
| Show the approval and log screen | Records who approved, when, and what changed | Log exists only in a vendor-side report |
| Switch the AI feature off for one user group | Setting exists and takes effect at once | Needs a support ticket or a plan upgrade |
| State where documents go for processing and whether they train a model | Written answer naming the provider and the retention period | "We take security seriously" |
The $1,000,000 limit that should have read $500,000
An illustrative case: a six-person commercial agency trials an intake agent on a package policy. The proposed per-occurrence limit reads $1,000,000 when the declarations page says $500,000, because a similar figure sits nearby in the aggregate section. The reviewer catches it only because the approval screen shows the source snippet beside each proposed value. Without the snippet, the figure would have gone into the record and surfaced at renewal or, worse, at claim. The lesson is to insist on source-visible approval for any limit, deductible or date, and to review the first fifty proposals closely before sampling.
What a switch really costs, beyond the licence
Licence prices are not on these pages, so build the part you can estimate: your own time. An illustrative twelve-person agency runs the old and new systems in parallel for three weeks, and each person loses four hours a week to double entry and training. That is 12 x 4 x 3 = 144 hours. At an internal cost of $45 an hour, it is $6,480, before any conversion fee.
Now set that against the AI gains. If EVA-style email and summary tools save 30 minutes per person a month (an assumption to test, not a vendor figure), twelve people save 72 hours a year, about $3,240. That takes two full years just to repay the parallel running, before any licence or conversion fee. Vertafore-style intake is a better case: suppose the agency writes 150 commercial package policies a year and each takes 90 minutes to create and review. The headline 87% saving would release 195 hours, about $8,800 a year. If the agency realises only 40 per cent, it saves 36 minutes per policy, which is 90 hours, about $4,050 a year. Those are assumptions, not forecasts; time ten real policies before you use them.
The conclusion for most agencies is sober. Switching just for AI works when a specific agent removes a big, measured block of work and the system also passes the export and fit tests. For smaller gains, use the AI your current system already has and revisit the question in a year. The tutorial on whether to switch CRM for better AI features applies the same logic to sales systems, and the year-one AI cost for an insurance brokerage shows where the other costs sit.
Testing a coach-style assistant on your own procedures
Coach and product-assistance features answer how-to questions from documentation and, in Vertafore's description, from your own standard operating procedures. They are only as good as the documents behind them, so test them with ten questions you already know the answers to. A filled-in log for an illustrative agency might look like this.
| Question asked | Answer correct? | Note |
|---|---|---|
| How do I add a driver mid-term? | Yes | Cited the right screen |
| What is our rule for certificates on request? | No | Gave the generic product answer; our own procedure was not loaded |
| Where do I log a client complaint? | Yes | Matched the procedure |
| How do I reinstate a lapsed policy? | Partly | Right steps, missed our approval rule |
Two of four fully right is not a pass. The fix is usually the content, not the tool: load the missing procedures, rerun, and decide in advance what score counts as acceptable, say eight of ten, before the trial starts. If the assistant cannot show which document an answer came from, treat every answer as unverified.
A four-week plan if you do switch
- Week one: scope and export. Run the sample export on both systems, freeze the scorecard and list every integration (rating, accounting, e-signature, phone) that must keep working.
- Week two: conversion rehearsal. Convert a copy of ten clients, including a messy one, and compare record by record. Log every field that moved wrongly.
- Week three: parallel running. Two producers and one service person work live in the new system while the rest stay in the old one. Keep AI features switched off for the first few days so you can separate AI faults from conversion faults.
- Week four: AI on, one feature at a time. Switch on the email assistant first, then summaries, then any agent that writes to records, with approval required and a daily log check.
Keep the old system read-only for at least one renewal cycle. The records you will want in an argument with a client or an insurer are the old ones, and the cost of paying for a quiet read-only licence is small beside losing them.
The email to send both vendors
Hello [first name],
We are a twelve-person agency considering a move. Please answer in
writing so we can compare vendors line by line:
1. Which AI features are generally available in [product and version]
today, and which are beta, early access or planned?
2. For each feature that changes records, where do I approve changes
and where is the log kept?
3. Can I switch each AI feature off by user group?
4. Is our data used to train any model? Which providers process it?
5. Can you run a sample export of five test clients before we sign?
6. What does a full export cost, how long does it take, and how long
after cancellation can we request it?
Thanks,
[your name]Score the replies for specificity before you score the features. A vendor that answers each numbered point with a sentence is easier to work with for the next five years than one that sends a brochure.
One last point on timing. Vendors tend to announce agents in waves, and a feature listed as early access in autumn may be generally available by spring. If a promising agent is not yet live in your product version, put a dated commitment in the proposal rather than scoring it as if it existed, and score only what you saw working.
Contract points to raise in plain words
None of this is legal advice, and your adviser should read the wording. These are the points worth raising before you sign, in the order they usually cause trouble.
- Data ownership and export. The agency owns its client and policy data, with a right to a complete export in a stated format within a stated time, including after termination.
- AI controls. You can disable each AI feature, by user group, without penalty or a forced upgrade.
- Model training. Written confirmation that your documents and messages are not used to train models, and a list of the AI providers that process them.
- Change notice. Notice when an AI feature changes materially or moves to a paid tier, as vendors sometimes move features between plans.
- Approval and logs. Features that write to records must require approval and keep a log you can export.
- Price protection. A cap on increases over the term and a clear statement of what is included versus billed by usage.
Tell your errors and omissions adviser which AI features you run, and ask whether cover has conditions about automated changes to records. For a fuller view of what an outside consultant would scope here, see hiring an AI consultant for an insurance agency.
If you decide to stay where you are
Staying is a legitimate outcome. Ask your current vendor the same questions in the same format, because vendors respond differently when a switch is possible. Then fill the gaps with cheaper pieces: a general assistant for drafting and summarising, a phone assistant for after-hours calls, and a tidy intake form. The tutorial on seven AI tools for insurance agents lists the pieces, and the one on AI answering for insurance agents covers the phone line.
Whatever you decide, write down the scores and the reasons. In a year, with new agents released and prices changed, you will be able to rerun the scorecard in an afternoon instead of starting from the beginning. If you would rather build that scorecard with someone, my AI implementation consultation covers it; start at the contact page.
Questions agencies ask when AI becomes a reason to switch
Should AI be the main reason to change agency management system?
Rarely. Vendor figures are mostly for single tasks, and a switch costs weeks of parallel running and data conversion. If your current system's own AI covers email drafts and summaries, switching for marginal gains seldom pays. Switch when exit, carrier fit or support are the real problems.
How do I tell a generally available AI feature from a beta one?
Ask for the status in writing for each feature, by product and version, and ask to see it running on your own sample documents. Vendor pages mix labels such as available now, early access, beta and planned. Only generally available features belong in your savings estimate.
What should I export before I sign anything?
Ask both vendors for a sample export of five test clients, covering contacts, policies, notes, activities, tasks and documents. Open the files yourself. You are checking that history comes out in a usable structure, with documents named by client and policy, not by internal ID.
Does AI in the system change my errors and omissions exposure?
It can. If an agent proposes changes to policy records, your process for approving them becomes part of your risk controls. Tell your insurance adviser which AI features you use, keep approval logs, and ask whether your own cover has conditions about automated record changes.
Further reads
- Can AI Answer Calls and Qualify Leads for an Insurance Agency? — The phone side, which agency systems rarely cover well.
- AI Chatbot for an Insurance Agency: Quote Requests, Not Advice — Website enquiries that should land in the new system.
- AI Insurance Lead Generation: Follow-Up Rules Before You Automate — Rules to set before automating any follow-up.
- Title Insurance AI: Pick a First Pilot and Price It for an Agency — A pilot-first approach to pricing AI in an agency.
- How to Read AI Software Pricing: Seats, Credits, and Usage Fees — Seats, credits and usage fees explained.
- What Can Go Wrong When AI Agents Take Actions for You? — What can go wrong when agents act on records.
- HubSpot for Insurance Agents: Pipeline, Renewals and What Breeze Adds — Step by step: one deal per policy, a new-business pipeline, a renewals pipeline, a 90-day view, and a three-job test of Breeze Assistant.
- Insurance Broker AI App: A Buying Checklist for a Small Brokerage — Thirty-six checks in six groups, a scoring sheet, a pilot script and sample pricing sums for choosing an AI app for enquiries and renewals.
- Zoho CRM for Insurance Agents: Policies, Renewals and Zia — A step-by-step Policies module and renewal rules in Zoho CRM, the edition limits that matter, and what Zia adds for an insurance agency.
- AI for Insurance Adjusters: Claim File Summaries and Status Letters — Claim file summary and status letter templates for independent adjusters, each with a built-in line that keeps coverage and payment decisions human.
- AI for Insurance Brokerage: Reconcile Insurer Commission Statements — A routine for parsing insurer commission statements, matching them to a brokerage's policy ledger, listing exceptions and chasing insurers.
- AI for Insurance Compliance: Licence Tracking and Complaint Logs — A licence and appointment calendar, CE reminder routine and complaint-log summaries for an insurance agency, with human verification of every rule.
- Insurance Agency SOPs With AI: Turn Routines Into Procedures — Record a routine, draft its decision points with AI and publish an SOP that the next account handler can actually follow.
- Zapier for Insurance Agents: Five Intake and Certificate Automations — Five Zaps for an insurance agency, each priced in Zapier tasks per run and per month, with a worked total for a six-person office.
- AI Cross-Selling for Insurance Agencies: Account Review Prep, Not Advice — AI prepares the file summary and the questions; the licensed agent advises. A prep routine, prompts and a sign-off log for account reviews.
- Does Agency E&O Cover AI Mistakes? What to Ask Your Carrier — A per-tool list of questions for your E&O carrier, the exclusion wording to look for, and the written AI policy that gives you evidence. Not legal advice.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: EZLynx artificial intelligence page and EZLynx press releases (30 January and 8 October 2025); Vertafore press releases (April, September and October 2026). Read in November 2026; neither vendor publishes pricing on these pages.