Otis Kim: Iris, long week — I ended up in this rabbit hole of AI workforce data at eleven p.m. on a Tuesday and genuinely could not tell if I was reading a crisis or a non-event.
Iris Holm: Which report?
Otis Kim: Both of them. CompTIA's AI Skills Tracker — published July 21st, thousand-plus leaders surveyed — and the Kyndryl People Readiness Report. And they're telling what looks like the same story from opposite ends, and I can't figure out if the story is 'we have a problem' or 'we have a measurement problem.'
Iris Holm: Give me the numbers.
Otis Kim: Eighty percent of professionals use AI tools multiple times a month. Twenty-nine percent say they're highly familiar with AI. And then Kyndryl: fifty-seven percent of organizations have AI embedded in core processes now — up from thirty-five last year — but only twenty-three percent believe their workforce is ready. That's a six-point drop in readiness confidence in a single year, while deployment went up twenty-two points. Those two lines are moving in opposite directions. That's what I couldn't shake at eleven p.m.
Iris Holm: Right. That's not a skills gap signal. Or — actually, it's not *only* that. When confidence drops as deployment accelerates, you're not watching a training problem. You're watching an accountability problem.
Otis Kim: Okay, but Seth Robinson — CompTIA's VP of Research — framed it as exactly a readiness problem. 'Critical success factor as organizations move beyond experimentation.' That's the official diagnosis.
Iris Holm: CompTIA runs a certification business. The diagnosis and the product happen to be the same thing. That doesn't disqualify the data — but it should shape how you read the framing.
Otis Kim: Fair — but that conflict-of-interest flag only gets you so far. You still have to explain the twenty-nine percent familiarity number against the eighty percent usage number. That gap exists independent of who's selling what.
Iris Holm: Think about it like buying a treadmill. Eighty percent of people in this survey are using the treadmill. But more than half of them are using it to hang laundry, not to run. The organizations, meanwhile, just installed treadmills on every floor of the building. That's the gap — not that people can't run, but that the machine is now load-bearing infrastructure and most people haven't needed to run on it yet.
Otis Kim: The laundry thing is — yeah, that lands. But what's the actual number?
Iris Holm: More than half of respondents say business-related activities are twenty percent or less of their total AI use. That's the CompTIA AI Skills Tracker. Most usage is personal, experimental. Exploration, not embedment.
Otis Kim: So the familiarity gap isn't a failure signal. It's a — wait, actually, it might just be accurate self-assessment? People know they're hanging laundry.
Iris Holm: Exactly that. And Kyndryl's number makes more sense through this frame — seventy-nine percent of organizations believe AI will outpace their ability to keep up. Seventy-seven percent say generative AI has already scaled across multiple functions. So is the six-point confidence drop a warning sign, or is it leaders becoming honest about complexity they previously ignored?
Otis Kim: I mean, we went through this with cloud. With PCs. With mainframes. Deployment always outran readiness and the world didn't actually end — organizations just muddled through until the learning caught up. So is this a crisis or is this just Tuesday?
Iris Holm: It's Tuesday — until it isn't. The historical parallel holds right up to the pilot-to-production moment. That's the inflection point. When AI moves from experimental into core workflows, the readiness deficit stops being theoretical. It becomes operational.
Otis Kim: And fifty-seven percent of Kyndryl's respondents say they're already past that point. Core processes. Load-bearing treadmills, on every floor.
Iris Holm: Load-bearing, yes — but who told us the load was that heavy? CompTIA publishes the inaugural AI Skills Tracker on July 21st. Kyndryl drops the People Readiness Report. Aon releases workforce analysis in June. All three land in the same six-week window. All three say the same thing: invest more in structured readiness.
Otis Kim: And all three sell the cure. CompTIA sells certifications. Kyndryl sells AI consulting and deployment services. Aon sells workforce consulting. That's not independent evidence — that's a sales pitch with footnotes.
Iris Holm: The methodology doesn't help. The CompTIA survey is self-reported. June 2026, thousand-plus leaders. Social desirability bias is baked in — nobody tells a surveyor 'yes, we deployed AI badly.'
Otis Kim: Right — but the directional finding is still real, even with that caveat?
Iris Holm: Directionally, probably. The survey asks whether workers are 'highly familiar' with AI. That's a confidence metric. Not a competence metric. Those are not the same number.
Otis Kim: Okay, let me make that concrete. Compliance analyst, regional insurance company. Wednesday afternoon. Her company rolled out an AI document-summary tool three months ago — official training module, mandatory video, nobody watched it. She learned it in forty minutes from a colleague's screen-share. Now she processes claims thirty percent faster than before. Her employer's readiness survey flags her team as undertrained. So — is the survey measuring a skills gap, or is it measuring whether she sat through the video?
Iris Holm: The video. Almost certainly the video. That's the familiarity metric — it's self-reported confidence tied to formal exposure, not actual output.
Otis Kim: Which means Kyndryl's 23% 'workforce ready' number might be measuring compliance with training programs, not whether anyone can actually do the job.
Iris Holm: And that distinction — familiar versus competent — is exactly where the data starts to crack. There's a 47% familiarity figure among workers actively building job skills versus 29% overall, and that gap is where informal learning gets interesting. But what it can and can't carry into production workflows — that's the part we need to get to.
Otis Kim: That 47 versus 29 number — it stands out. CompTIA's own data shows workers who are deliberately building job skills hit 47% high familiarity. The overall population sits at 29. That's not a small gap. That's almost a different category of person.
Iris Holm: It's a motivation sorting mechanism. The CompTIA AI Skills Tracker isn't measuring a skills gap there — it's measuring who decided to care.
Otis Kim: Right — but the method matters. Over six in ten workers are learning AI through general social tools. YouTube, Reddit, Slack threads. Not corporate LMS platforms. Not certification tracks.
Iris Holm: And it's working. That's the part the readiness-crisis framing wants to skip.
Otis Kim: Okay, I want to test that. Picture a hospital billing coordinator — not a tech role, not adjacent to tech — learns the new AI coding-assist tool entirely from a TikTok walkthrough and a Discord channel for medical billers. No corporate validation. No certification. She's faster, more accurate. The CompTIA survey flags her team as undertrained. Actually, no — wait. The survey would flag her as unfamiliar, not undertrained. Those are different indictments.
Iris Holm: Different indictments, yes. The barriers workers actually cite are limited time for training and uncertainty about which AI skills matter most. That's structural friction — not motivational failure.
Otis Kim: I'll grant that. Informal learning holds — until it's running payroll. Ismail Amla and Dr. Vishnu Nanduri at Kyndryl are explicit that the gap isn't uniform. Some organizations navigate the pilot-to-production transition differently. A colleague's screen-share can close a workflow gap. It cannot close a governance gap.
Iris Holm: That's the concession I'll take. The informal learning case holds. Right up until it's the system of record.
Otis Kim: We started this conversation with eighty percent usage and twenty-nine percent familiarity and called it a crisis. Aon says readiness is the competitive differentiator — that's June 2026, that analysis. CompTIA says certify. But if the compliance analyst learned the tool in forty minutes from a colleague and outperformed everyone who sat through the training video — I mean, what exactly are we measuring? And if organizations stop treating readiness as a checkbox and start designing around how people actually learn — what does that system even look like?
Iris Holm: I don't know. That's the part I can't close.
Otis Kim: Neither can I. And I've been sitting with it since eleven p.m. Tuesday.
Iris Holm: The question's worth leaving open.
Otis Kim: Yeah. Thanks for not pretending you had an answer.
