Skip to main content

Why IT Keeps Getting Handed an AI Training Problem It Can't Solve Alone

Emily Mabie
Zapier

A few months ago, an IT director at a 2,000-person company looked at me on Zoom and said, "I have no idea how I became the person responsible for this." His CEO had told him to make sure everyone gets AI training by Q2. He had three engineers, no curriculum, and a Slack channel full of people asking how to write better ChatGPT prompts.

He's not alone.

A recent Zapier survey of more than 500 enterprise C-suite executives and decision-makers makes that pattern visible in numbers. 77% of leaders say their teams need AI skills urgently. 64% say their organization plans to train current employees rather than hire new ones. So far, so reasonable. The part that surprised me is who's been put in charge: 34% of those leaders say IT and engineering own the AI skills mandate. Learning and Development or HR own it at 7% of organizations.

That's roughly five-to-one in favor of the people who understand the tools, over the people whose actual job is teaching adults how to learn new ones.

I've spent enough time around AI rollouts to know why this happens. When a new technology is unfamiliar, leaders default to whoever can get it to work. That's usually IT. But there's a quiet assumption in that handoff: knowing how a tool works and teaching a workforce to use it are the same skill. They aren't. Knowing the API and knowing how adults learn a new way of working are different disciplines, and treating them as interchangeable is how you get rooms full of people who watched the demo and still don't use the tool a month later.

The Shelf Life Problem

The other thing the survey surfaced is speed. Among the 78% of leaders who name a barrier to building AI skills, the most common one (18%) is that AI is changing fast enough to make training obsolete almost as soon as it's built. Anyone who has tried to write internal documentation for a model that ships a new version every six weeks knows the feeling.

This compounds the IT ownership problem. If you're responsible for training and the curriculum is out of date by the time it's reviewed, you fall back on what you can sustain: a wiki page, an office hours channel, maybe a recorded webinar that ages poorly. None of that builds real fluency. It builds a small group of confident users and a much larger group of people who quietly stop trying.

What Works Better than Another Curriculum

In the work I do, I get to watch how teams actually learn AI in their daily work. The patterns that hold up under real conditions don't look like training programs. They look like changes to how people already work.

A few things consistently help.

Start from friction, not features. Most rollouts begin with someone explaining what a tool can do. The teams that get further start by asking where work feels slow, repetitive, or mentally draining. AI introduced as relief from a real annoyance gets traction. AI introduced as a capability tour does not.

Embed the learning where the work already happens. A short demo at the start of a weekly standup beats a separate workshop almost every time. So does walking through how someone used AI to prepare for a decision in a project retro, or sharing a rough draft and the prompts behind it in a team channel. The signal these rituals send is that AI is part of how the team thinks, not a separate skill to certify in.

Make imperfect normal. Everyone has tried to use an AI tool, gotten a moderate result, and quietly stopped using it. They worry about looking like they don't know what they're doing. But when a manager shows their team the rough draft AI helped them write, complete with the clunky parts they're still fixing, it tells everyone else that practicing in public is fine. That practice is what actually builds judgement.

Invest in champions, not heroes. Effective adoption rarely hinges on one expert. It depends on a small group of people close to the work who are willing to learn out loud and help others. They don't need deep technical backgrounds. They need credibility with peers and permission to experiment. Over time, they become translators between what's possible and what's actually useful for their teams.

The Piece It Can't Carry Alone

None of this contradicts IT's role. IT teams have to make sure AI tools are deployed responsibly, the data flowing into them is governed, and the systems they touch stay reliable. That's plenty. What IT can't reasonably also do is design and run the human side of adoption, especially in functions IT doesn't sit inside every day.

The survey hints at why so many leaders feel stuck. About three-quarters of them say they're confident their organization already has the right people and skills to hit their AI goals. That confidence is hard to square with the gaps in the same data: unclear ownership, training that goes stale in weeks, employees with tools and no guidance. Something has to give.

The version that works, in my experience, is the one where IT, L&D, and the functions actually using AI build the program together. IT brings the systems view. L&D brings the way adults actually learn. The teams using AI bring the friction that makes the training matter. None of those three can do the job on their own, and the moment one of them is asked to, the program starts to wobble.

If you're the person who keeps getting handed this, it's worth saying out loud: this is a design problem the IT team can't solve alone. Name the gap. Pull in the people whose job is teaching adults to learn. Start with one team, one real friction point, one working example. That's how this actually gets built.

Emily Mabie is AI Automation Engineer at Zapier

Hot Topics

The Latest

For fifteen years, observability lived downstream of everything else. Code shipped, something broke, an engineer went to the dashboards. The job was forensic. The pillars we built, such as logs, metrics, and traces, were designed for that role: tell a human what just happened, fast enough that they can make it stop. That role has quietly ended ...

Hybrid IT has become the standard operating model for enterprises — but that companies are still looking for the right hybrid IT mix, according to the 2026 State of the Data Center Report from CoreSite. After years of cloud migration and hybrid adoption, organizations are shifting their focus from deciding whether to use cloud, colocation or on-premises infrastructure to determining which workloads belong in each environment ...

Pilots are everywhere, stakeholders are seeking results, businesses are pushing for new tools, and IT teams are being asked to make AI secure, reliable, and useful at scale. But as organizations move from testing AI to operationalizing it, many are discovering that the biggest barrier is not the model, the use case, or even the budget. It is the file data foundation within ...

Fast or cheap? For most of my career in engineering, speed and quality sat on opposite ends of a seesaw. The "OR" in "fast or cheap" was non-negotiable. It was expected that pushing for faster releases meant that something in quality would give way. Tightening quality controls meant the schedule slipped. Every engineering leader I know has lived some version of that tradeoff ... The seesaw is starting to level out ...

I have been building enterprise software for more than 20 years ... One thing stays true across all of it: You do not find out your foundation is wrong during the crisis. You find out when the debt comes due. For a lot of organizations, that bill is arriving now. New research ... puts hard numbers on something practitioners have been sensing for a while. The telemetry problem isn't coming. It's already here ...

The rapid growth of AI workloads is pushing traditional log management approaches to their limits, according to The State of Log Management 2026 report from Dynatrace. Modern logs have become critical to understanding, validating, and securing AI-driven decisions, helping organizations ensure reliability, compliance, and performance at scale. However, the volume and complexity of AI telemetry are overwhelming legacy tools ...

For years, secure connectivity has relied on a familiar pattern: route traffic back to centralized gateways, inspect it, and then allow access. This model worked when applications lived in a handful of data centers and users were largely confined to offices. That model is now under strain. Applications are distributed across clouds, users connect from everywhere, and real-time workloads demand performance that centralized inspection points struggle to deliver. As traffic volumes grow and latency expectations shrink, routing everything through a small number of control points has become both a performance bottleneck and a resilience risk. The future of secure connectivity requires a different approach ...

The AI experimentation phase is over, and the private cloud is where enterprise AI workloads are being deployed for security and scale, according to Private Cloud Outlook 2026, a new report from Broadcom ... 2026 marks an acceleration into a full AI tipping point. The shift is being shaped by three forces — costs, complexity, and control — that public cloud environments are increasingly failing to address for production AI at scale. Key findings from the report include ...

44% of organizations have reported an outage in the past year tied to suppressed or ignored alerts, and 78% had at least one incident where no alert was fired at all ... Engineers learned about failures from customers. That gap between what our tools report and what our customers experience is the problem DevOps teams have been quietly solving with GenAI tooling, even as most enterprises continue to run their NOCs on manual alert triage ...

Cloud outages are usually described as technical failures. When a service goes down, a dependency breaks, or a region has issues, the focus immediately shifts to infrastructure. But if you look closely at how these incidents actually unfold, the root cause is rarely the technology itself. It is almost always tied to decisions made earlier, during design, implementation, or day-to-day operations. The system behaves the way it was built. The real question is how it was built ...

Why IT Keeps Getting Handed an AI Training Problem It Can't Solve Alone

Emily Mabie
Zapier

A few months ago, an IT director at a 2,000-person company looked at me on Zoom and said, "I have no idea how I became the person responsible for this." His CEO had told him to make sure everyone gets AI training by Q2. He had three engineers, no curriculum, and a Slack channel full of people asking how to write better ChatGPT prompts.

He's not alone.

A recent Zapier survey of more than 500 enterprise C-suite executives and decision-makers makes that pattern visible in numbers. 77% of leaders say their teams need AI skills urgently. 64% say their organization plans to train current employees rather than hire new ones. So far, so reasonable. The part that surprised me is who's been put in charge: 34% of those leaders say IT and engineering own the AI skills mandate. Learning and Development or HR own it at 7% of organizations.

That's roughly five-to-one in favor of the people who understand the tools, over the people whose actual job is teaching adults how to learn new ones.

I've spent enough time around AI rollouts to know why this happens. When a new technology is unfamiliar, leaders default to whoever can get it to work. That's usually IT. But there's a quiet assumption in that handoff: knowing how a tool works and teaching a workforce to use it are the same skill. They aren't. Knowing the API and knowing how adults learn a new way of working are different disciplines, and treating them as interchangeable is how you get rooms full of people who watched the demo and still don't use the tool a month later.

The Shelf Life Problem

The other thing the survey surfaced is speed. Among the 78% of leaders who name a barrier to building AI skills, the most common one (18%) is that AI is changing fast enough to make training obsolete almost as soon as it's built. Anyone who has tried to write internal documentation for a model that ships a new version every six weeks knows the feeling.

This compounds the IT ownership problem. If you're responsible for training and the curriculum is out of date by the time it's reviewed, you fall back on what you can sustain: a wiki page, an office hours channel, maybe a recorded webinar that ages poorly. None of that builds real fluency. It builds a small group of confident users and a much larger group of people who quietly stop trying.

What Works Better than Another Curriculum

In the work I do, I get to watch how teams actually learn AI in their daily work. The patterns that hold up under real conditions don't look like training programs. They look like changes to how people already work.

A few things consistently help.

Start from friction, not features. Most rollouts begin with someone explaining what a tool can do. The teams that get further start by asking where work feels slow, repetitive, or mentally draining. AI introduced as relief from a real annoyance gets traction. AI introduced as a capability tour does not.

Embed the learning where the work already happens. A short demo at the start of a weekly standup beats a separate workshop almost every time. So does walking through how someone used AI to prepare for a decision in a project retro, or sharing a rough draft and the prompts behind it in a team channel. The signal these rituals send is that AI is part of how the team thinks, not a separate skill to certify in.

Make imperfect normal. Everyone has tried to use an AI tool, gotten a moderate result, and quietly stopped using it. They worry about looking like they don't know what they're doing. But when a manager shows their team the rough draft AI helped them write, complete with the clunky parts they're still fixing, it tells everyone else that practicing in public is fine. That practice is what actually builds judgement.

Invest in champions, not heroes. Effective adoption rarely hinges on one expert. It depends on a small group of people close to the work who are willing to learn out loud and help others. They don't need deep technical backgrounds. They need credibility with peers and permission to experiment. Over time, they become translators between what's possible and what's actually useful for their teams.

The Piece It Can't Carry Alone

None of this contradicts IT's role. IT teams have to make sure AI tools are deployed responsibly, the data flowing into them is governed, and the systems they touch stay reliable. That's plenty. What IT can't reasonably also do is design and run the human side of adoption, especially in functions IT doesn't sit inside every day.

The survey hints at why so many leaders feel stuck. About three-quarters of them say they're confident their organization already has the right people and skills to hit their AI goals. That confidence is hard to square with the gaps in the same data: unclear ownership, training that goes stale in weeks, employees with tools and no guidance. Something has to give.

The version that works, in my experience, is the one where IT, L&D, and the functions actually using AI build the program together. IT brings the systems view. L&D brings the way adults actually learn. The teams using AI bring the friction that makes the training matter. None of those three can do the job on their own, and the moment one of them is asked to, the program starts to wobble.

If you're the person who keeps getting handed this, it's worth saying out loud: this is a design problem the IT team can't solve alone. Name the gap. Pull in the people whose job is teaching adults to learn. Start with one team, one real friction point, one working example. That's how this actually gets built.

Emily Mabie is AI Automation Engineer at Zapier

Hot Topics

The Latest

For fifteen years, observability lived downstream of everything else. Code shipped, something broke, an engineer went to the dashboards. The job was forensic. The pillars we built, such as logs, metrics, and traces, were designed for that role: tell a human what just happened, fast enough that they can make it stop. That role has quietly ended ...

Hybrid IT has become the standard operating model for enterprises — but that companies are still looking for the right hybrid IT mix, according to the 2026 State of the Data Center Report from CoreSite. After years of cloud migration and hybrid adoption, organizations are shifting their focus from deciding whether to use cloud, colocation or on-premises infrastructure to determining which workloads belong in each environment ...

Pilots are everywhere, stakeholders are seeking results, businesses are pushing for new tools, and IT teams are being asked to make AI secure, reliable, and useful at scale. But as organizations move from testing AI to operationalizing it, many are discovering that the biggest barrier is not the model, the use case, or even the budget. It is the file data foundation within ...

Fast or cheap? For most of my career in engineering, speed and quality sat on opposite ends of a seesaw. The "OR" in "fast or cheap" was non-negotiable. It was expected that pushing for faster releases meant that something in quality would give way. Tightening quality controls meant the schedule slipped. Every engineering leader I know has lived some version of that tradeoff ... The seesaw is starting to level out ...

I have been building enterprise software for more than 20 years ... One thing stays true across all of it: You do not find out your foundation is wrong during the crisis. You find out when the debt comes due. For a lot of organizations, that bill is arriving now. New research ... puts hard numbers on something practitioners have been sensing for a while. The telemetry problem isn't coming. It's already here ...

The rapid growth of AI workloads is pushing traditional log management approaches to their limits, according to The State of Log Management 2026 report from Dynatrace. Modern logs have become critical to understanding, validating, and securing AI-driven decisions, helping organizations ensure reliability, compliance, and performance at scale. However, the volume and complexity of AI telemetry are overwhelming legacy tools ...

For years, secure connectivity has relied on a familiar pattern: route traffic back to centralized gateways, inspect it, and then allow access. This model worked when applications lived in a handful of data centers and users were largely confined to offices. That model is now under strain. Applications are distributed across clouds, users connect from everywhere, and real-time workloads demand performance that centralized inspection points struggle to deliver. As traffic volumes grow and latency expectations shrink, routing everything through a small number of control points has become both a performance bottleneck and a resilience risk. The future of secure connectivity requires a different approach ...

The AI experimentation phase is over, and the private cloud is where enterprise AI workloads are being deployed for security and scale, according to Private Cloud Outlook 2026, a new report from Broadcom ... 2026 marks an acceleration into a full AI tipping point. The shift is being shaped by three forces — costs, complexity, and control — that public cloud environments are increasingly failing to address for production AI at scale. Key findings from the report include ...

44% of organizations have reported an outage in the past year tied to suppressed or ignored alerts, and 78% had at least one incident where no alert was fired at all ... Engineers learned about failures from customers. That gap between what our tools report and what our customers experience is the problem DevOps teams have been quietly solving with GenAI tooling, even as most enterprises continue to run their NOCs on manual alert triage ...

Cloud outages are usually described as technical failures. When a service goes down, a dependency breaks, or a region has issues, the focus immediately shifts to infrastructure. But if you look closely at how these incidents actually unfold, the root cause is rarely the technology itself. It is almost always tied to decisions made earlier, during design, implementation, or day-to-day operations. The system behaves the way it was built. The real question is how it was built ...