--- Title: "You're bought in on AI. Here's what the first 90 days actually look like" Url: "https://devrev.ai/blog/90-days-of-AI" Published: "2026-08-31" Last Updated: "2026-08-31" Author: "Tami Erwin" Excerpt: "AI is happening whether you participate or not, and the leaders who win will treat it as a business decision, not a technology one. Here's what you can expect. " Reading Time: 4 --- # You're bought in on AI. Here's what the first 90 days actually look like Tami Erwin spent 35 years at Verizon, rising from rep to CEO, and now sits on four public boards. In a recent conversation about AI in the enterprise, she made a case that's hard to argue with: AI is happening whether you participate or not, and the leaders who win will treat it as a business decision, not a technology one. _"People are confusing the technology of AI with the capability that AI delivers."_ It's a clarifying line. But it also surfaces a question: _I agree. So what do I actually do on Monday?_ The interview maps the altitude - the why, the who, and the buy-versus-build tension. What follows is the ground beneath it: what a serious first 90 days looks like once you've decided AI is in fact a leadership priority. [Video](https://www.youtube.com/watch?v=KkassXSlk6o) ## **Start with one painful, measurable problem - not a strategy deck** The instinct after a conversation like this is to convene a committee and write an "AI strategy." That's usually the fastest way to lose momentum, because it makes AI an abstraction that lives in a slide. The better move is to pick one problem that is (a) genuinely painful, (b) measurable today, and (c) owned by a single accountable person. Not "improve customer experience." Something like "our support team takes 14 hours to first-response on tier-2 tickets, and we want it under 2." You'll know within weeks whether AI moved the number. This is the practical version of the point that the leadership job is the outcome, not the technology: "The bigger challenge for AI is what problem are you trying to solve, and that is leadership." One problem. One owner. One number. That's a first 90 days you can defend. ## **Fix the data before you fall in love with the tool** Erwin was emphatic that clean data is step one, and it's the part most teams skip because it isn't exciting: "Data in is either going to give you truth out or it's going to give you garbage out." The nuance worth adding: you don't need _all_ your data clean. You need the data underneath your one chosen problem to be trustworthy. Scoping the data work to the use case is what makes it finishable in 90 days instead of becoming an 18-month enterprise data program that quietly kills the initiative. Three questions to ask of any use case: - Where does this data live, and who owns its accuracy? - Is there a single source of truth, or three systems that disagree? - If we act on this data automatically, what's the cost of being wrong? If those can't be answered, that's the first sprint - not model selection. ## **Redesign the work, don't just bolt AI onto it** The most common failure mode isn't picking the wrong tool. It's dropping a capable tool into a workflow designed for humans doing everything by hand, then wondering why nothing changed. Erwin framed the redesign as people, process, and platform - all three, or you fail the mission. The process leg is where teams get concrete: map the current workflow step by step, then ask which steps are judgment (keep them human), which are toil (hand them to AI), and which shouldn't exist at all once AI is in the loop. Often the biggest wins come from that last category - work you stop doing entirely. ## **Measure adoption as honestly as you measure outcomes** Here's a point the altitude view doesn't always require: a pilot can hit its target metric and still fail, because the team quietly went back to the old way the moment the spotlight moved on. In a first 90 days, two things are worth tracking, not one: - **The outcome metric** - did the number move? - **The adoption metric** - are people actually using it a month in, without being reminded? If adoption is low, the problem is rarely the technology. It's trust, training, or a workflow that never got redesigned. Which leads back to the most durable idea in the interview. ## **The part AI can't touch** For all the talk of tools and platforms, Erwin kept returning to the human core - and it's the right note to end on. "You never replace the humanity that is at the core of really successful businesses." The first 90 days are technical and operational. But the reason they succeed or fail is human: whether people trust that the tool is there to remove their toil, not their job. Leaders who bring people along - who teach them to redefine what success looks like - get compounding adoption. Leaders who impose a tool get a pilot that looks good in a board deck and dies in the field. ## **The 90-day checklist** If there's one thing to take from all of this, it's the list: 1. Choose one painful, measurable problem with a single accountable owner. 2. Clean only the data that problem depends on - not the whole enterprise. 3. Map the workflow and separate judgment from toil. 4. Buy the platform, partner on the solution - don't rebuild plumbing. 5. Track outcome _and_ adoption from day one. 6. Bring people along - trust is the real adoption engine. Erwin's advice is to embrace AI because it's happening anyway. So, make it small, make it measurable, and make it human. That's a first 90 days that earns you the next 90.