Scaling AI Without Breaking Your Culture

This one got into the part of AI adoption nobody really wants to talk about. Ryan Deeds runs AI at ALKEME, an insurance company that bought its way from seven agencies to ninety-three in about four years, and his whole argument is that the technology was never the hard part. The hard part is getting fifteen hundred people to actually trust a tool and use it. We talked about the golden path, why every rollout fails for the same reason, the eighteen months it really takes to hit adoption, and the bot pipeline that lets anyone at the company build their own software. Watch the full conversation above.📩 Subscribe to Leadership in ChangeOutline(00:11) – Scaling without breaking culture(00:49) – Seven agencies to ninety-three in four years(02:07) – From data guy to Head of AI(03:41) – The golden path across 150 workflows(05:32) – “AI is a bunch of hype” and the cultural work underneath(08:17) – Why every rollout fails for the same reason(12:54) – The 85% that fail, and what they all skipped(14:21) – Fear, mistakes, and fifteen dead initiatives(15:05) – Eighteen months to 70% adoption(19:43) – Why the model wars are wasted noise(21:14) – There’s no easy button(24:05) – The bot pipeline that lets anyone build(27:19) – Scaling AI without it blowing up in your face(29:23) – The most overrated AI adviceMy TakeawaysThe failure is always culture. Ryan has built nine CRMs and implemented seven more, and he says the result never changes. “Got nothing to do with technology. It’s culture.” I brought up the stat that 85% of AI implementations fail, and his read matched what I see all the time: people get handed a tool, get told to use it, and decide not to, either because they think it’s there to replace them or they don’t trust the people who bought it.There’s no easy button. Ryan’s least patient moment was about the model wars. “You cannot judge current state AI in current state,” he said, and the people arguing over which model beat which this week are usually the ones not actually building anything. What matters to him is boring on purpose: how many people used it, how many minutes did it save. “If I was judging what I created, I’d be freaking Elon.” That’s the line I keep coming back to, because most of us are measuring the wrong thing.Eighteen months to seventy percent. I appreciated how honest he was about the timeline. He told his CEO up front it would take about eighteen months to reach full adoption, and that full adoption, in his world, is 70%. The pushback he plans for doesn’t come from the technology. It comes from moving people out of a culture where nobody was accountable and into one where everybody can see your numbers.Closest to the work is clean. This was the part that genuinely surprised me. ALKEME has fifteen hundred employees and a build team of nine, so Ryan built a pipeline instead. An employee emails an idea, a bot pins down the narrowest version of success, another bot builds the MVP on the company’s own infrastructure and sends back a working link to iterate on. “I would way rather empower you, employee E, to bring your concept to fruition.” The person closest to the problem builds the fix, inside the guardrails.Don’t pour a bucket of AI water on everything. For an insurance company, a broken tool isn’t a small thing, so I asked how he moves this fast without it blowing up. His answer was discipline about scope. “It’s not just pour a bucket of AI water on everything and see where it hits.” They ran fifty thousand historical submissions through one tool before they trusted it, locked permissions to each user, and pushed everything through a pipeline that gets checked. “We’re just very careful about where it touches.” It’s the same discipline ALKEME brings to the coverage side, where a single missed risk like a cyber attack is exactly what their clients are paying to be protected against. If your business carries that exposure, their team handles cyber insurance directly.If you’re the person inside your company trying to get AI actually adopted and not just installed, that’s most of what I work through with teams. You can start at jsalinas.org and book a free call with me.If you rolled out an AI tool tomorrow and nobody used it, would you call that a technology problem or a culture problem? Sit with which one you’d actually fix first.Watch the full conversation above, and go connect with Ryan Deeds on LinkedIn and see what ALKEME is building.About Ryan DeedsRyan Deeds is Head of AI & Enablement at ALKEME, an insurance company that has grown to around 93 agencies and 1,500 employees in roughly four years. He spent 30 years in the insurance agency space, hosted The Digital Broker podcast from 2017 to 2019, and now builds the internal tools and adoption systems that let ALKEME scale AI without losing its culture. Connect with him on LinkedIn.About ALKEMEALKEME is a national insurance company built from more than 90 independent agencies, serving clients across commercial, personal, and specialty lines. One of those lines is cyber insurance, coverage that protects businesses against cyber attacks, data breaches, and the operational fallout that follows. If your organization needs to think seriously about cyber exposure, talk to ALKEME.About meJoel Salinas is an Executive AI Coach for leaders at small and mid-sized businesses and nonprofits. 1:1 coaching, team workshops, and AI strategy work built around amplifying what your team is already good at. Creator of the AI Leadership Triad. He writes Leadership in Change. If you want help thinking through your own AI strategy or online presence, start here.Written by a human, for humans. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit leadershipinchange.com/subscribe