How does an insurance agency / broker use inTusell? A daily operations guide

In the pillar article you trained the inTusell AI. You loaded your branch and coverage information, your premium and scope notes and your frequently asked policy questions into the knowledge base, then set the persona and the boundaries specific to insurance. Training was a one-time setup. Now the real work begins: how does an insurance agency / broker use inTusell — that is, how does using inTusell for insurance work in daily operations? This article shows step by step what the assistant you trained does across a workday, from the incoming quote request to the meeting appointment, and from the policy renewal reminder to follow-up and human handoff.

This article is the second part of the insurance chapter of our sector-by-sector "how to train your AI" series. The first part, how to train the inTusell AI, explained the setup; this article covers turning that setup into daily operations.

Quick answer

The flow of a day

The typical day of an agency that has set up inTusell looks very different from before. A representative used to arrive in the morning and go through dozens of unread WhatsApp messages and unanswered DMs one by one, sorting which was a new quote, which a renewal, which a claim. Now the messages that came in overnight are already answered, requests are pre-classified, some meeting appointments are booked, and renewal reminders have gone out. The backbone of a day looks like this:

The team no longer spends its day scanning messages and sorting "new vs. renewal vs. claim." It works on the conversations that genuinely require a representative — risk assessment, premium negotiation, running claims.

Who is it for?

This usage model is especially meaningful for these insurance businesses:

• Agencies receiving heavy quote requests via WhatsApp and Instagram — those who can't keep up with dozens of "how much is comprehensive/traffic/earthquake?" messages a day.
• Agencies and brokers missing renewals — those losing a customer to another agency because the policy expiration date passed.
• Businesses losing after-hours demand — those who respond late to evening, weekend and urgent requests like "I need traffic insurance today."
• Agencies managing many branches and many channels with a small team — those forced to track WhatsApp, DM, phone and the renewal list separately.

If your demand is more than one representative can keep up with, inTusell meets it like an experienced insurance assistant with clear boundaries — not like a chatbot.

The most important boundary: the AI does not give risk and claims advice

For an insurance agency, this is the rule that comes before everything. In a concrete situation, inTusell does not give a risk assessment, a claims strategy, or definitive advice like "you'll lose out if you don't get this coverage." That is a professional boundary, and it also keeps a wrong steer from putting the customer and the agency at risk. The binding assessment is made by the licensed representative.

In practice, what the AI does and doesn't do is separated like this:

When a customer asks a concrete question like "Will I be deemed at fault in this accident, will my claim be paid?", the AI does not run through the possibilities. It says "This depends on your policy's scope and the file details; let our representative assess it" and routes them there. We explain how to set this boundary during training in the how to train the inTusell AI article.

Incoming request: the customer asks for a quote, the AI collects info and directs

The flow starts when the customer sends a message. Imagine someone writing on WhatsApp "I want to get comprehensive auto for my 2019 car, can I get a price?" The AI reads the message and pre-classifies which branch the request falls under — here, comprehensive auto. It then asks the quote questions you defined during training, in order: vehicle make/model/year, usage, claims history, and the license plate if available. Finally it directs the customer to a quote meeting or appointment.

Two points are critical. First: the AI does not state the premium out of thin air. It collects the quote request in full but does not produce an exact premium, because the premium depends on a risk assessment, and the licensed representative does that. Second: the AI replies in whatever language the customer writes in — a practical convenience for agencies with foreign customers.

This way the request reaches the representative complete, and nobody has to question the customer over and over. You'll find the details of the quote intake and renewal engine in the quote and renewal automation article; and how all channels are combined into a single inbox is explained in the Instagram and WhatsApp automation article.

Premium and coverage questions: answered from the knowledge base, no invention

In insurance, most incoming messages are not a new customer but a recurring scope question: "Does my comprehensive policy cover glass damage?", "Is earthquake insurance mandatory?", "Can traffic insurance start today?" The same questions come back all day and eat up the representative's time. inTusell takes that load by answering from the knowledge base.

The most critical rule here is this: the AI does not invent premiums and coverage. It gives its answer only from the source you've loaded into your knowledge base:

If someone asks about information that isn't in the knowledge base — the scope of a product you didn't load, say, or a personalized exact premium — the AI doesn't invent anything. It says it doesn't know and offers to connect the customer to a representative. That keeps you from giving a customer the wrong coverage and then having to say "actually that scope wasn't included." A false statement puts both the customer relationship and the agency's liability at risk.

This is why you keep the knowledge base up to date. inTusell feeds it from a few sources and indexes them for pgvector-based search (RAG): product/coverage files (PDF, Excel, CSV — branch, limit, included/excluded, standard premium), web URLs (agency/product/FAQ pages) and free text (for example "this month, cash discount on traffic"). Because conditions change frequently in insurance, when you update a product condition you also refresh the knowledge base; the AI speaks based on the most recent source you loaded.

An additional security layer: the AI is constrained by a sector-specific system instruction; a guardrail above it scans responses and flags risky statements (for example "definitely paid," "guaranteed compensation"). This layer runs in observation (shadow) mode by default: it doesn't hard-block, it detects and flags. The system instruction is what actually steers the behavior.

Meeting and quote appointment: slot, duration and conflict prevention

When the customer wants a detailed quote or coverage meeting, the AI doesn't offer a time out of thin air. It looks at the three things you defined during training: the duration of the meeting type (for example "Quote Meeting 30 min"), the representative's business hours, and whether that slot is taken. If it finds a suitable slot, it creates the meeting. This flow suits branches that need explaining, such as life, private pension or corporate packages; for simple traffic/comprehensive requests, collecting information is often enough on its own.

The most critical safeguard here is conflict prevention: two meetings cannot be booked for the same time. Before creating an appointment, inTusell checks your business hours and existing appointments. If a conflicting slot is requested, it won't open it, and cancelled or missed appointments free the slot up again. An automatic reminder goes out for each appointment (for example 1 day and 2 hours before) and reduces no-shows. The customer manages their own appointment through the /manage-appointment/{token} link and can cancel or reschedule with one click. Appointments are copied one-way to Google Calendar (inTusell → Calendar). This module can be enabled in every package; if you don't need it, you don't use it.

Policy renewal: the follow-up engine's most valuable job

In insurance, the customer you lose most often is the one whose policy quietly expires and who renews somewhere else. Renewal is far cheaper and more profitable than finding a new customer, which makes this the follow-up engine's most valuable job at the agency. inTusell tracks policies whose expiration date is approaching with a proactive follow-up flow and sends the customer a timely renewal reminder.

The reminder goes out over the channel the customer wrote on; if they can't be reached, SMS kicks in, and if that fails, email. The message is not a dry "your policy is expiring" alert; it directs the customer either to the new quote flow directly or to a short renewal meeting. For an interested customer, the AI collects the necessary up-to-date information (for example, has the vehicle changed, is the address the same) and hands the quote over to the representative complete.

One boundary should be clear: inTusell is not a separate product that keeps a managed, personalized "renewal agenda." What it does is offer an opportunistic re-quote or reminder, based on the expiration date, to suitable WhatsApp customers in follow-up. In the morning you can see which policy is near renewal and which is awaiting a callback, so you no longer keep a renewal list by hand and work down it call by call. We explain in detail how this engine combines with quote intake in the quote and renewal automation article.

Claim reporting: front-line reception, then the representative

The customer most often writes at their most stressful moment — after an accident, fire or theft: "Someone hit my car, what do I do?" inTusell acts in a balanced way at this moment. It doesn't comment on the claims outcome, doesn't say "will be paid/won't be paid"; it lists the first steps and required documents you entered into your knowledge base (for example a report, photos, policy number) and escalates the situation to a representative along with its urgency.

So in a claim, the AI is a front-line reception and routing layer: it keeps the customer calm, has the first information collected, and moves the file to the right person. Running the file, the loss adjuster process and coordination with the company stay with the representative; the representative takes over the file not from scratch, but with the information the AI has collected.

Following up the quiet customer: proactive follow-up and CRM

In insurance, most sales are lost to the customer who says "let me compare prices" and disappears, or who keeps postponing the renewal. inTusell does not forget these leads. A proactive follow-up message goes out to a customer who got a quote or information and then went quiet. Every conversation is recorded in the CRM, so the customer's history, the branch they're interested in, their existing policies and the stage they're at all stay on record. The next contact starts from where it left off, not from scratch.

One boundary should be clear here too: inTusell does not keep a managed "waitlist." What it does is offer an opportunistic re-quote to suitable WhatsApp customers in follow-up — for example, a reminder along the lines of "The comprehensive auto quote you asked about is still valid, shall we continue?" to a customer who got a comprehensive quote and went quiet. In the morning you can see which lead is warm and which is awaiting follow-up, so there's no need to keep a manual reminder list.

The insurance pipeline

Every prospect moves along a pipeline, and the AI advances them to the right stage as the conversation develops: new quote request → information collected → quote presented → meeting/negotiation → policy issued (won) or dropped (lost). At a glance you can see which prospect only asked about price and which is awaiting a policy decision. Renewals run on a separate follow-up flow keyed to the expiration date, so new business and renewals progress in the same panel without getting mixed up.

Human handoff: working modes

You determine how autonomously the AI works. There are three working modes:

In hybrid mode, the AI hands off to the representative when a question needs a risk assessment, when a claim is reported, or when a scope question falls outside the knowledge base. At the moment of handoff, the team can see the entire conversation history, so the customer doesn't have to start over. You can change the mode at any time — for example, ai_only during a busy renewal period, human_only for a complex corporate quote.

A side note: a file such as a registration photo, an old policy image, a claim photo or an ID is not on its own a reason for human handoff. The AI takes the attachment, understands the context and continues the flow; it only hands off if there's truly a situation requiring a decision (for example an active claim).

Privacy, KVKK and sensitive data

In insurance, the incoming information is often sensitive: identity, address, vehicle information, and in some branches a health declaration. inTusell protects this at three points:

1. Explicit consent record: The customer's explicit consent is recorded with its time and type (kvkk_consent_at / kvkk_consent_type).
2. Sensitive field control: Fields such as health declarations and identity are flagged with the highest sensitivity; they are not stored automatically and require manual approval.
3. Encrypted storage and masking: Personal data is end-to-end encrypted; PII masking is applied to audio recordings.

This makes it easier to answer the question "what did the customer accept, and when?" with evidence in case of an audit or dispute.

Reports: what's working, what should be fixed?

To make operations visible, you track core metrics in the panel: how many messages came from which channel, how many the AI answered, how many went to a representative, which branches are asked about most, and how many renewal reminders turned into a new quote. Seeing these numbers serves two purposes:

1. Finding knowledge-base gaps: If there's a branch or coverage where the AI often says "I don't know" and hands off, that information is missing from the knowledge base. Once you add it, the handoff rate drops.
2. Improving response rules: You review the labels and correct misclassified conversations — a renewal logged as a new quote, for example. The AI learns from these corrections. You can also compare two different renewal-reminder approaches with an A/B test.

So reports are not a summary; they are a feedback loop that sharpens the assistant over time.

What it isn't

To put inTusell in the right category, let's clarify what it isn't:

• It is not an insurance company or licensed representative. It does not perform a risk assessment, produce an exact premium, or make a claims decision; the agency's licensed representative handles these.
• It is not a comparison/quote portal. It does not pull instant premiums from companies and compare them; it collects the quote request and moves it to the representative, and the agency issues the policy.
• It is not a collection or policy-issuance system. It does not collect the premium or issue the policy; it is a communication and coordination layer, not a seller or intermediary.
• It is not an insurance-specific priced product. The pricing model is based on messages and voice minutes, and does not change by sector. The scheduling and follow-up modules can be enabled in every package.

One more note: Instagram comment automation is rolled out gradually depending on Meta approval; the first step is always WhatsApp and Instagram DM, which are approved. Keeping these boundaries clear from the start saves you from setting the wrong expectation. For product and pricing details, see the solutions and pricing pages.

Frequently asked questions

Does the AI sell the policy on its own?

The AI collects the quote request, answers the customer's coverage and scope questions from the knowledge base, and takes them all the way to a meeting/quote appointment. Finalizing the premium, risk assessment, issuing the policy and collection stay with the agency. In risky or out-of-knowledge-base situations, the AI hands the conversation off to a representative. The agency always issues the policy.

Does the AI invent premiums and coverage?

No. The AI states premium, coverage limit and scope information only from the source you have loaded into your knowledge base. It will not invent a coverage, exclusion or instant premium it does not know; it says it does not know and hands off to a representative. This is the most critical trust rule in insurance.

How does it handle policy renewal reminders?

It tracks policies whose expiration date is approaching with the follow-up engine and sends the customer a timely renewal reminder over the channel they wrote on (WhatsApp, DM, Telegram, Messenger); if they can't be reached, SMS kicks in, and if that fails, email. If the customer is interested, they are directed to the new quote flow or a meeting appointment.

Can the customer be connected to a representative?

Yes. In hybrid mode the AI runs the normal flow and escalates the conversation to a representative when needed. In human_only mode all conversations go straight to the team, and in ai_only mode the AI manages all of them. You can change the mode at any time.

What does the AI do when a claim is reported?

The AI does not give advice at the moment of a claim; it lists the first steps and required documents from your knowledge base and escalates the situation to a representative along with its urgency. Running the claims file and coordinating with the loss adjuster and the company stay with the agency; the AI here is a front-line reception and routing layer.

How are customer data and KVKK protected?

Customer data is end-to-end encrypted; explicit consent is recorded with its time and type. Sensitive fields such as identity and health declarations are flagged with the highest sensitivity and are not stored automatically. When needed, the conversation is handed off to a representative along with the full conversation history.

Next step

If you haven't trained your assistant yet, start first with the how to train the inTusell AI article — the foundation of these operations is established there. For the channel side, move on to Instagram and WhatsApp automation. For the details of the quote intake and policy renewal engine, see quote and renewal automation. The solution built for insurance agencies is on the insurance page, and the whole series is on the all articles page.

If you're curious how the same logic is set up in another regulated field with a "does not give advice" boundary, clinic AI training and legal AI training offer a good comparison; and in tourism, the flow is in the tour agency AI training article.

To see live how it works in your own agency, use Get a demo or write directly to hello@intusell.com. In a 20-minute session, we'll open your inbox together and test the first quote request and a renewal reminder in the system.