🧭 Find your weakest productivity vector in 5 minutes: free assessment
🎓 Productive, Not Busy, my cohort on Maven: join the waitlist
Why AI Adoption Failure Starts With a Broken System
Companies buy the tool, train the team, and wait for the hours to come back. The hours rarely show up. Chris Carolan has spent years inside the plumbing of businesses, first as a HubSpot implementer and now as a transformation advisor, and he has watched enough rollouts to name the pattern: AI adoption failure traces back to the same root cause every time, a process that was broken before anyone added a model to it. On episode 87 of ProductiviTree, Carolan walks through what he’s seen inside hundreds of companies, why the software is rarely the part worth fixing, and what leaders should look at before they sign another contract.
The Real Cause of AI Adoption Failure
Carolan’s career started in HubSpot implementation, setting up the software that was supposed to make sales and marketing teams faster. That job put him inside the engine room of a lot of companies, and it gave him a view few consultants ever get: the gap between the workflow on the diagram and the workflow people actually run day to day.
That gap is where AI adoption failure begins. A company buys a tool to fix a symptom, but the underlying process, the handoffs, the approvals, the person who quietly holds the whole operation together in their head, stays exactly the same. The software changes. The system underneath it does not.
Learned Helplessness Inside a Broken System
One of the clearest patterns Carolan describes is what happens to people who work inside a process nobody has ever questioned. They stop looking for better ways to do things, because the system has already taught them that better ways don’t survive contact with reality. He tells the story of an order confirmation email, a small detail buried in a client’s workflow, that ended up revealing an entire operation built on workarounds nobody had written down.
That is the failure mode he says shows up in nearly every company he walks into: a gap between what leadership believes is happening and what is actually happening on the ground. Buying a tool is a decision a leader can announce on Monday. Redesigning how work actually moves through a company is not, so leaders keep reaching for the tool instead.
Transformation Versus Optimization
Carolan draws a hard line between optimization and transformation, and he’s upfront that it’s a line his own industry blurs constantly. Optimization makes the current process a little faster. Transformation changes what the process is. Optimization is cheaper, faster to sell, and nobody loses their job over it, which is exactly why leaders keep picking it even when the deeper problem never gets touched.
His answer to the obvious objection, that transformation is an overused word in consulting, is to point at the deliverable: a working system the client owns and can run without him, not a slide deck and an invoice. Whether that line holds depends on the engagement, but it’s the right question for any leader evaluating a vendor selling the word.
What Stays Human When AI Holds the Context
Carolan’s own firm runs on a deliberate split. AI holds memory, preparation, pattern recognition, and rhythm. Humans keep every relationship and every judgment call that actually matters. He’s aiming for something close to 95 percent automation on the operational side of his business, with a small, deliberate 5 percent kept human on purpose.
That split sounds clean on a website. It gets harder in a real company, especially for the person whose entire value used to be holding context that nobody else had. Carolan’s answer isn’t that the role disappears. It’s that the person gets to spend their time on judgment and relationships instead of on remembering where things are, which only works if leadership actually redesigns the job instead of just handing them a chatbot.
Ownership When a System Prepared the Decision
Carolan also tells a story from his own HubSpot rollout days about ownership: what happens when a system prepares a decision and a human signs off on it without really checking the work. Approving what a model produced can look exactly like thinking, from the outside. He argues the leaders who get this right build in a slower, deliberately human step at the exact points where a wrong call is expensive, even when the machine could technically move faster.
What Leaders Should Fix Before They Buy More Software
The takeaway from Carolan’s work isn’t that AI doesn’t work. It’s that AI amplifies whatever system it gets dropped into. A clear process gets clearer and faster. A broken one gets broken faster, at scale, with a bigger invoice attached. Before a company signs another AI contract, the better question is whether anyone in the room actually knows how work gets done, not how the diagram says it gets done.
That matches the wider data. A 2025 MIT analysis of hundreds of enterprise AI deployments found that the large majority delivered no measurable financial return, and the gap traced back to the same place Carolan points to: a process nobody fixed before the AI arrived (Forbes, August 2025).
Carolan’s framework, which he calls the AI-Native Shift, walks companies through five stages: diagnosis, unified systems, integrated context, a platform the company owns rather than rents, and the internal capability to keep running it without him. The order matters. Skip the diagnosis and the rest of the stages just add AI to the same broken operating model.
For more on building systems that hold up under pressure instead of just looking good on a roadmap, check the free productivity tools and templates, take the free productivity assessment, or join the ProductiviTree newsletter for more episodes like this one.
Listen to the Full Episode
Chris Carolan goes deeper on all of this, including the exact split his own firm runs between AI and human work, on episode 87 of ProductiviTree.
Guest links: Chris Carolan, Value-First Team, AI-Native Shift, Chris Carolan on LinkedIn.