Three days of keynotes at ITC Vegas will give you more good ideas than you can carry home. You hear 40 of them, you nod along, and by Monday you can name maybe two. So at our Brainshare session this year at Mandalay Bay, Jason Cass and I tried to fix that problem with the same tool everyone at the conference was talking about.
We took the transcript of each speaker's talk and gave it to Claude with two instructions. Pull out the 3 main points, then give us the contrarian take, the strongest argument against what the speaker said. Both went up on the screen, the drinks were cold, and a room full of independent agency owners and producers took it from there. Jason's one rule: talk it through and keep it friendly.
Here's what the AI said, where the room agreed, and where it pushed back.
Ryan Hanley: find your team's "easy mode"
Ryan's talk was about the work every person does so naturally it looks like cheating. His point was that a job title hides talent. He told the story of a CSR on his team who was struggling until she was matched with pickup truck contractors, and then became a top producer. The AI's summary added the practical piece: use AI to remove the low-value work that keeps people away from their strengths, and skip the shiny tools you don't need.
The contrarian take was fair. Not every agency can build roles around strengths, because someone still has to answer the service calls and do the data entry, and real growth often comes from work that isn't easy.
The room answered with how they use AI to take out the work that quietly makes a team less efficient. One owner has every employee write down what they did in 15-minute blocks, once a year, and then looks for where the time is going. Double entry "only takes 5 minutes," until you stack it across 50 quotes a month. Then the math gets uncomfortable: an owner who won't pay $150 a month for a tool is choosing to pay salary for 4 people to retype data for 3 to 5 hours a month each. His advice was to stop hunting for the big two-hour win and look for the micro moments. How do you give your team 25 minutes back every day?
Someone else shared a small workflow that made a big difference. The CEO's receipts used to pass through 3 or 4 people before the accountant saw them. Now he emails them to one address, an automation files them and creates an activity in Epic, and the accountant just confirms.
The big learning here: technology works when it's built into the process itself, as part of how the work gets done every day. That receipt workflow exists because the agency hired someone whose whole job is technology and process. If you want to scale, I'd think hard about making your next hire that person.
Shawn Michael Walker: replace yourself
Shawn talked a lot about calibration and about replacing yourself, which is a different goal from just growing faster. The AI's pushback: calibration sounds wise but is hard to measure without clear numbers, and an owner can call hesitation "prudence" while competitors are already making bold moves with AI.
The room went straight to a decision a lot of owners avoid. When the principal is also the agency's best salesperson, the team can't reach them, no one has time to find better systems, and that owner is usually the most stressed person in the building. One principal picked a lane, went full producer and hired an operations person to run the agency. Six months later he called to say it was the happiest he'd been in his insurance career. Jason put it plainly: agents are not cheap about cars and houses, so we need to stop being cheap about the operations hire that makes us more money.
Billy Wagner: autonomy is built one handed-off task at a time
Billy's line on the slide was "autonomy is built one handed-off task at a time," and the AI asked the question every small agency is thinking: what's a realistic first step?
Billy's answer is one of the best things I heard all week. Keep a document that lists every responsibility in the agency, with a person's name next to each one. Early on, every line has your name on it. Each time you hire, you peel off the tasks that fit that person's highest and best use. His agency gives every new hire a $25 Working Genius assessment to find out what they love doing, and every year he plans 3 hires or changes and decides exactly what he's handing off. He was also honest about timing. In the first 5 years you should have your head down selling and building systems, and some roles only make sense once revenue supports them, so know your targets.
Billy's agency also gets 300 or more candidates for a single job post. A few things make that happen:
- The whole team shares every job post on their own social media.
- The job ads include video, which gets seen far more than a text post.
- Billy knows the owners of several local restaurants and asks them to refer staff who want a career, because people from hospitality do well in insurance.
Because of all this, he always has about 10 people waiting for a spot to open. The same hiring problem shows up in how agencies train new producers, and the fix starts the same way.
Ciara: big agencies can outspend you
Ciara's message was that your brand, the specific thing that makes you different, removes the "they're bigger and have more money" excuse. Her best tactic, and the one an attendee called his biggest takeaway of the day: market to the advisors your ideal clients already take advice from.
The AI's counterpoint said consistent content takes time a small agency may not have, which got a laugh, since AI built a slide about not having time to make content. The best answer came from an owner with 20 years in insurance marketing. Office TikTok dances get 50 likes and zero leads. A video answering "what is workers comp?" gets zero likes and 50 leads. People who believe you can solve their problem don't hit like. They hit the button that says "can you help me with my insurance?"
As for the time problem, someone pointed out that most agencies have young people on staff who want to learn these tools and are being treated as nothing more than a producer or CSR. Look inside your agency first.
Michael Weaver: AI as a coach
Michael's talk was about using AI to coach producers and CSRs. One owner in the room uses ChatGPT's voice mode for role-play. Teammates role-playing with each other tend to say yes too fast, while the AI will hold the objection as long as you tell it to.
This is also where I talked about carrier playbooks. A lot of agencies update a playbook every week with market appetite and who's writing what. If you're already doing that work, put it in OneDrive or SharePoint and connect Copilot or Claude to it, so a new hire can ask "I've got this new risk, where should I go?" and get your agency's answer. Put one person in charge of keeping it current. AI won't mysteriously fix a file that no one maintains.
Then the recording question. Jason asked the room who records calls and meetings, and his view was that it's the number one way to use technology right now. My morning brief runs on it. At 6:50 every morning, a Teams chat with my executive assistant has a summary of every call and meeting from the last 24 hours, plus the action items I owe.
The AI's contrarian take was a good one: recording and scoring every call drives accountability, but it can feel like surveillance. How do you keep your team's trust? The room's answer was visibility and leaderboards, because A players want to be on A-player teams and want to be judged on what they do. Jason's example: his personal lines producers weren't bringing up life insurance, so he asks Claude which of yesterday's calls never mentioned it. If someone had 6 calls and asked once, that's a coaching conversation. Most of the time it's a training issue, because people want to do it right.
The data privacy argument
This was the part of the night that got a little heated. Ryan's warning was direct: unless you're running an open-source model on hardware you own, assume that what you push through any AI platform is stored on their servers, business plan or not. He still told everyone to use the tools. He just wants owners to stop assuming they're protected, and he admitted our industry was never perfect with data to begin with. How many Social Security numbers have lived on sticky notes on a desk?
An attendee disagreed. Business plans have a setting that keeps your data out of training, and legally that matters, though he thinks the terms could change down the road. My take: read the data terms on the plan you're paying for, turn that setting on, and decide what stays out until you're comfortable.
Charles: people development is your edge
Charles closed his talk with a line from his kid's martial arts: a black belt is a white belt who never gave up. The slide that got the room going was "as AI becomes available to everyone, people development becomes your edge. An intern at a spare desk costs almost nothing."
An owner who came from the same organization explained how it works. They recruited college juniors and seniors from the entrepreneurship club once called SIFE, now Enactus, where students already compete in sales pitch contests. Students are used to cramming, so they get licensed in about 2 weeks. They shadow a good producer, start by quoting, move to sales calls, and by graduation they're a fully trained producer. The only thing missing is inbound leads.
Marie: is it OK to fail?
Marie from Molina talked about failure leading to success. The room reframed it, and I think the reframe is right. None of us want our people chasing failure. What we want is for failure to carry no punishment when someone is chasing something good for the agency, a higher close rate or a faster process. That's on the leader. One owner also pushed back on keeping those lessons inside one-on-one meetings, because the whole team learns more when someone says out loud, "I tried this, it didn't work, how do we fix it?"
Paul: the CSR is your future producer
Paul from HawkSoft made the point that the CSR stuck on repetitive work is your future producer, once AI takes the repetitive part off their plate. The room added that new hires now ask what AI tools you use. One owner said he'd almost question a candidate who didn't ask.
4 takeaways for your agency
- The wins that sound small are the ones that compound. Receipts to one inbox, an auto-reply on web quote requests, 25 minutes back per person per day. Using AI well can be that simple. Pick one annoying task, fix it properly, and let it pay you back every day.
- Use what's already in your building. Your playbook, your call recordings, the young CSR who wants to learn. Connect them before you buy anything new.
- In most cases, the owner is still the bottleneck. Billy Wagner's system makes delegating easier: one document with every responsibility in the agency and a person's name next to each one. Each hire takes the tasks that fit them best, and each year he plans the next 3 hires or changes and what each one will take off his plate.
- Your next hire might be a tech and process person. Somebody has to build the workflows, or they never get built.
New learning from this
The contrarian prompt is what made the whole exercise work. Left alone, Claude tends to agree with whoever is typing, which is how an owner ends up hearing their own ideas back. Asking it to argue with you is how you get a perspective your team didn't bring to the table, and almost every good idea of the night came from someone disagreeing with the AI.
You can do the same thing after any brainstorm or team meeting. Teams and Zoom both give you a transcript, so paste it in with this prompt:
Here is the transcript of our team meeting. What did we miss? Then give me the strongest contrarian argument against each idea we agreed on, in 2 sentences each.
Bring the answers to your next meeting and talk through the ones that sting. Some of them will be wrong, and that's fine. The point is that your team gets to decide with the other side of the argument in front of them.
Which contrarian take would you have argued with? Drop it in the comments.
Want this kind of room? IndieTech is February 2 and 3 in West Palm Beach, built for independent agents, with an invite-only Brainshare the day before. Details at aiindietech.com.