How to Get Your Insurance Team to Actually Use AI

Where insurance teams are on the AI journey

Chaz Perera, BevayaSo a lot of insurance teams are in different stages of their AI journey. Some are just starting to explore AI. Some are mid pilot and a growing number have an agent live in production today. Wherever you are on that path, today is for you because getting live is the part everyone plans for and what comes after is the part almost no one plans for.

That's true whether go live is a year out or if it happened last month. Everyone asks, "How do we get something live?" Almost nobody asks the harder question. How do I turn a tool that's live into something my people will use and trust and lean on every single day? So that's what today is about.

The milestone versus the gap that opens the morning after

There's a milestone here and that's and then there's the actual gap and most teams only ever plan for the milestone. The milestone is technical. Can we get this built, integrated, and running on real work. The system is ready.

That part closes with technology, with a vendor, with a project plan. It's hard, but it's a known kind of hard. The gap opens the morning after. The tool is live and it is now.

It is not yet part of how your people work. Nobody gets assigned to that gap and it isn't necessarily on the project plan. There's no go live date for it. And that's the one that quietly stalls rollouts even when the technology is performing exactly as designed.

The dashboard says green. The usage probably tells a different story. This gap is a leadership challenge, not a technology challenge. And that's good news because it's the part you can really, really shape.

Roles change first, and if you don't define them people will

So, where does that leadership challenge start? With roles, the first thing that changes after go live isn't titles. It's people's actual day-to-day. Here's why that matters.

If you don't define the new role, people will define it themselves and usually not well. Frontline staff build workarounds all the time. Leads guess what to manage. Managers keep measuring the old thing because nobody told them otherwise.

By the way, none of this is malicious. It's just what happens. So, let's get the roles right first. And a lot of the downstream friction will never show up.

Frontline staff: from doing the task to reviewing the work

Frontline staff importantly will feel it first. Think about the person in the mail room who used to open and sort and index claims mail by hand. Once AI handles the first pass, they're not indexing anymore. They're reviewing its work, handling exceptions, and picking up higher value tasks they never had time for.

A workers comp carrier that is a customer of ours went through this in their mail room. AI took over high volume indexing and saved them more than 22,000 hours. But here's what mattered for adoption. Their goal was never just about speed.

It was making sure nobody ended up overworked or sidelined. So the work changed that people stayed.

Team leads: from managing volume to managing judgment

Team leads have a different shift and it's the one people most underestimate. They move from managing task volume to managing judgment. Knowing what the tool does well, what it needs a human for, and how it reads its outputs well enough to answer the team's questions. That's really a new skill and most leads are learning it live in front of their people.

So investing in leads first is really important because they set the tone for everyone underneath them.

Managers: from whether work got done to whether it got done right

And managers jobs shift from tracking whether work got done to tracking whether it's done the right way. Automating intake should free underwriters to spend time on what they're paid for. Evaluating risk servicing agents and brokers helping the end customer. Put an AI agent on a COI process for example and it's the same shape as the mail room.

The people who used to grind through manual work now handle exceptions and client relationships while the routine volume runs on its own. So whatever function you're in, the pattern holds. But that only happens if the manager actively steers people there. It doesn't happen on its own.

And the message to your people has to be plain. This isn't about preserving jobs. It's about careers evolving. People can hear that. what they can't handle or hear is silence.

And when you do get silence or push back or someone dragging their feet, that's worth understanding because it's rarely what it looks like on the surface. So, let's also talk about that.

Why people hold back, and why the reasons don't look alike

When people hold back, resistance is usually the word we use for it. What actually looks like, sorry, what looks like resistance is almost always a reasonable reaction to something specific. I see five reasons for resistance and they don't look alike.

Skepticism, fear, workflow friction, falling behind, past scars

The first is skepticism about accuracy. This is often from your most experienced people applying the same standards that make them good at their job. They've seen many tools overpromise. They're not blocking anything in reality.

They just won't trust the output until they've watched it be right on real files over and over again. That's what you call professional judgment doing its job. The second is fear of being replaced. This shows up as a quiet disengagement.

Easy, by the way, to mistake for apathy. Nobody raises a hand in a team meeting to say they're worried about their job. So, they don't engage. and you read it as indifference when in reality it's fear. The third is workflow disruption.

This is the person who isn't against the tool at all. Their system worked before and now there's an extra step in the middle. The resistance is to the friction, not to the artificial intelligence. The fourth is falling behind.

This is a quiet worry about being the one who doesn't understand it or get it that's often uncomfortable to admit. So it goes unspoken and rarely gets raised as a real concern. So you have to go looking for it. And the fifth is distrust from a past roll out.

Almost every team has a scar from a roll out that just didn't deliver. When the next tool shows up, they assume it'll be the same. That distrust started long before this project did and it isn't about your AI agent at all. So, same behavior on the surface, but really five different causes underneath.

You need to respond to all five the same way and you'll fix none of them. Sorry, what I meant to say is if you respond to all five the same way, you'll fix none of them. So, let's move on to how to resolve these.

Matching a response to each one

Okay, you need to match a response to each because just communicate more is not an effective answer for the accuracy skeptics. Don't assert perfection. Show the receipts. Put the error put the error rate and the review process in front of them.

This is where confidence scoring earns its keep. a score on every extracted value allows the reviewer to go straight to what the human needs to focus on. And when the AI kicks something out for review, that's not the system failing. That's it working as you designed it. There's one more way that turns your biggest skeptic in to your biggest ally.

And that is deciding where to set the threshold. How sure does the AI have to be to allow it to act on its own? It's a risk decision, not a technical one. And nobody is better positioned to make that call than your most experienced people because they understand the downside of getting it wrong.

So give them that decision. Once the team sees the accuracy hold up on their own documents and their own data and their own files and their own work items, they're the ones who choose to push more volume straight through. The trust gets earned, not mandated. That's the pattern you really want to strive for.

Now, for the folks worried about their jobs, say the quiet part out loud before they do. Name the role change directly. Be specific about what the AI takes and what stays. And make the higher value work real, not a slogan.

For workflow disruption, for the workflow disruption group, get specific and get close. Sit with them, watch the new flow, fix the friction. These are these are all the sorry these are almost always small things. A handoff that moved, a screen that added two or three clicks, a notification that stopped firing.

And they feel it's an enormous change because they're living it every single day. So the quicker you solve it, the quicker you get them to buy in. The mistake is treating a workflow complaint as an attitude problem when it really isn't. Now for the people worried about falling behind, say out loud that everyone is starting from the same place, including you.

Make it a safe space to ask a basic question and give people room to practice before real stakes are on the line. And back that up with real training, role-based tied to how their jobs are actually changing, not a generic walkthrough of the system. there's a learning curve. It's not a performance problem. And the way you prove that is by resourcing it, not just saying it.

And finally, for the for the team carrying scars from a past roll out, acknowledge the last one was bad. Don't pretend it's a clean slate and show them one concrete way this time is different. Then let the results do the talking. Now, every one of those responses has something in common.

None of them are a one-time conversation. They only work if you keep showing up. Which brings me to our next slide because none of this works without it.

Communication cadence: weekly early, monthly after, never zero

Okay, so most teams overcommunicate at launch and then they go silent. So you got to stretch the cadence, but don't stop it. In the early weeks, often the anxious period, people are forming their opinion of the tool. That opinion is sticky.

So, your primary channel should be a weekly live check-in person or on a call where you can read the room. Email is the wrong primary channel because announcements go out and questions never come back. Keep a written log for issues, but don't lead with it. and act fast on what you hear because the first ignored problem is honestly the last one raised. Share the wins, but also share the misses.

And as you move to steady state, weekly becomes monthly, but it never goes to zero. The content shifts from check-ins to results, reinforcing what's working and correcting drift. And make those results your team's own number, not vague reassurance. because a specific stat lands harder than you know quote unquote things are going well. That's also where peer voices matter more than yours.

One adjuster telling another that the tool made their week easier does more than anything you'll say from the front of the room. Then on an ongoing basis, the channel has to stay open. Keep one place for questions and repeat it sorry repeat where it lives every few months because people will tend to forget. Treat any new agent or workflow change as its own mini launch.

Almost make it normal to a ask a basic question at every point not just you know on the first week. And importantly stay visibly engaged with the tools yourself well after launch. Your attention is the signal everyone will read.

Live versus adopted

So let's define adoption. Live means the tool is technically running. And this is where most rollout tracking stops. Adopted means something else.

Your people use it, they trust it, and they would complain if you took it away. That last part is a real test. Are your experienced people leaning on it for the judgment calls or quietly doing the work the old way on the side and running the AI to check a box? That can happen more than people realize.

And when it does, you've done the hard part of getting it live and left the payoff on the table. One of our customers put a fun spin on this. when their agent went live, the team named it, threw it a pizza party to welcome it into the organization. It's the moment the tool stopped being a threat that got installed on them and became part of how they worked. That that's adoption. and it's a team and behavior thing, not a metrics thing.

No status report tells you whether people have truly changed how they work. Only the people can. So, what do you watch for to see if your team has truly adopted the new way of working with AI?

The signals that tell you adoption is real

Seven signals are here, and every one of them is something you'd notice in a normal week, not something you'd have to go measure. Staff correct the tools output instead of working around it. It comes up naturally in conversation, not just in status updates. People ask to extend it to a new use case on their own.

They troubleshoot before they escalate. Complaints shift from why do we have to use it to why doesn't it do X yet? Nobody's quietly asking to go back and new hires learn the AI assisted way first with no old school workaround taught alongside it. Look at these against your own team and the ones you the ones you can't say yes to yet.

That's where to put your attention.

What to act on, whether you're a week in or three months in

Okay. So, let's try to bring it all together. Five things to act on whether you're you're one week in or three months in. Name the role change for every group before people define it themselves.

Name the resistance you're actually seeing and match your response to it. set a cadence weekly early, you know, weekly early on, monthly after, never zero. Find out who's really using it, not just the dashboard. And find your early believers and give them room to show their peers. None of that requires a new budget or a new project.

It requires showing up for weeks after go live the same way you showed up for the weeks before it. All right. getting your team to actually use AI well doesn't happen on a single date. It builds from small ordinary moments, the kind we covered today. One last thing before I hand it back to Jill for Q&A.

On the Bevaya side, every deployment includes a platform that watches for exceptions, manages model drift automatically, and gets smarter with every correction or interaction with your team. So, you can spend time on your team using it, not on tweaking and improving the AI. If you want to talk through what that looks like, click request a demo. We've done this across more than 120 deployments of our AI in production at scale.

So there's a really good chance we've seen something close to your situation. Thanks everyone for taking time to be with us today. With that, I'm going to hand it back to Jill for Q&A.

Jill Zakarian, BevayaThanks Chaz. That was great. We hope everyone's walking away feeling more confident about helping your team turn AI adoption into everyday success.

Q&A: How long before I can stop actively managing it?

Jill Zakarian, BevayaOkay, Chaz, we do have some questions that came through. So, first question in our queue is how long does this take? When can I stop actively managing it and trust that it's just part of how the team works?

Chaz Perera, BevayaWell, certainly longer than the go live date and shorter than you probably fear, but I but I wouldn't give you a hard number because it genuinely depends on the workflow, the size of your team, your past experience, right? What I call scar t the scar tissues that you and the team are carrying from other types of rollouts. what I what I what I would say is it's more important to watch and try to read the signals because it's more useful than a timeline. Actively managing your team, their engagement, their use of the AI. it's an exercise in just checking in with your people. you know, are they hitting the same friction point two or three times? is anyone quietly working around the new process that you're trying to establish? Right?

If if you're not seeing those things after digging in deep, you've reached real adoption when nobody's like genuinely annoyed by what they're working with. that's I think when you can ease off. so again, not a calendar thing, it's a signaling thing.

Q&A: What if the resistance is a team lead, not the front line?

Jill Zakarian, BevayaAwesome, Chaz. So, next question I see in our queue is, "What if the resistance isn't coming from the front line at all? It's a team lead or a manager who doesn't trust it and is signaling that to their people."

Chaz Perera, BevayaWell, I suppose if that's happening, you know, the risk is that those types of things could quietly sink a roll out. you know, a skeptical frontline person typically only affects their work where a skeptical team leads a team lead affects everybody who reports to them. and they usually do it without really ever saying anything bad against the tool or at least not directly. you know, it probably will show up in small things. you know, like I just double check that anyway. Or, never really referencing the tool as part of their workflow. or just never modeling great leadership. and you know teams have a tendency to read those things instantly because people watch what their manager does and and certainly watch those far more closely than what leadership says or does at an all hands. So you have to address it and what would I say here? you know, I think it's important to probably address it with the team lead privately, but address it directly. you know, the lead doesn't have to love it, but they do have to deal with it and they do have to message it and shape it for it to be successful with their team. So the thing I would say you should dig into would be my guess is nine times out of 10 the lead's resistance is other people's fear disguised as resistance, right?

Folks on their team potentially or maybe some of their peers. and what's driving it? Probably you know a past roll out that turned out poorly. because I'm I'm guessing they had to clean up the mess. So, don't just try to correct the behavior, try to diagnose what's underneath. because if you only manage sort of the surface signaling, it's going to come back to haunt you. you know, just communicate, engage, communicate, engage. and be very transparent.

Q&A: What's the minimum that moves the needle?

Jill Zakarian, BevayaAwesome. All right, Chaz, the next question is, everything you described is a lot of ongoing leadership attention. Realistically, most of us are already stretched. What's the minimum that moves the needle if I can't do all of it?

Chaz Perera, BevayaI suppose if you only do one thing, close the loop on problems really, really quickly in the first month and then you could consider everything else that follows is optional because the single biggest predictor of whether adoption is going to stick is whether people believe that raising an issue will lead to a real answer. And if that's true, they'll keep engaging. If that's true, they'll forgive a lot of the rough edges while things settle. And you get to use that the first 30 days to build that mentality and that practice and way of working into your business your the psyche of your team.

By the way, if they flag a problem and it vanishes into a void, they'll stop talking and then they will quickly but quietly revert to their old way of working. And honestly, no amount of communication three months later is going to win them back because we've you've essentially taught them that speaking up is pointless. So if you're stretched thin, don't try to run the full playbook badly. Just be the person who just be the person who builds the communication plan, responds timely in the first few weeks. that will lead to less intensity over a long period of time.

By the way, the irony is that leaders who obsess over the polished launch type of communication, they're typically the ones that go quiet after the launch and almost never communicate a week to two weeks after the launch. So the idea here, I guess, is responsiveness beats polish. So please pick responsiveness.

Q&A: Genuine skeptic or someone waiting you out?

Jill Zakarian, BevayaNice. All right, I think we have time to squeeze in just one last question. so how do you tell the difference between someone who needs more time to trust the tool and someone who's just never going to use it and is waiting you out?

Chaz Perera, BevayaI suppose like it's a time and behavior and I suppose the tell is very specific, right? The for example, the genuine skeptic engages. They poke at it. they bring real objections. if you got this calculation wrong or perspective wrong, they're going to point it out because they're really testing it against what they know. This person is actually doing you a favor even when it doesn't feel like that.

And I know it's hard. It's it's going to feel really uncomfortable sometimes being, you know, having the person poke and prod. But every objection they raise is a specific opportunity to fix something, explain something, and engage with them. And it's been my experience that eventually they will come around and come around because they were heard.

They had an opportunity to provide perspective and they saw that their their voice and their perspective shaped what came next. Right? The other thing I'd say here is real skepticism is loud and detailed. passive resist passive resistance, which is much harder, is vague. So, if you have somebody who's loud, give them the time. and don't rush them, by the way.

Give them the time and don't rush them. If somebody's falls in that other category of being a passive skeptic, you're going to eventually have to go to them directly. and be very direct in your conversation. you know, don't obviously don't don't be threatening. Just clarify that, you know, this is part of your role now and this is what you have to do and this is why it's important, but address their underlying concerns, right? Don't don't don't let their perspectives go unanswered.

Jill Zakarian, BevayaPerfect. All right, everyone. We are going to wrap up there. We appreciate you making time to join us today.

Also, if you think of any questions later, feel free to email us at resources@bevaya.ai. Thanks so much, and we can't wait to see you all again soon.