We’ve Been Talking About AI Taking My Job Forever and I Still Have a Job to Do
When I worked for the Commonwealth of Massachusetts, I had a sign on my wall. It said: We’re going to keep having meetings until we can figure out why no work is getting done. I put it up as a joke. It stopped being funny pretty quickly.
I’ve been thinking about that sign a lot lately.
Over the last few years, I have spent more time in conversations about how we’re going to use AI to work better and faster than I’ve spent actually working better and faster. I mean, have you been to a conference lately? It almost feels like they’re working on a quota where if they don’t say AI at least 50 times an hour, they have to pay a fine.
The meetings have different titles now. The slide decks are prettier… mainly because people are using Claude to build them. But the output is remarkably unremarkable… a lot of discussion and not a lot of results. It feels like busy work with the sole purpose to just look good.
I don’t think I’m alone in this. I think this is happening everywhere, to everyone, and most people are too busy attending the next AI strategy session to process their thoughts or say it out loud.
Or they’re afraid to be labeled anti-AI, A procurement dinosaur, or just the negative person.
I don’t think I am any of those… I think AI has great potential, probably uncapped potential. I think bad business is its only real hindrance.
So, I’ll say the inside thought outloud… And I’ll bring references.
Eighty percent of enterprise AI projects fail to reach meaningful production deployment.
That’s not a hot take. That’s the conclusion of RAND Corporation research, which also found that AI projects fail at roughly twice the rate of traditional IT projects.1 S&P Global went further: 42% of companies abandoned most of their AI initiatives in 2025, up from just 17% the year before.2 Forty-two percent! Gartner is now predicting that more than 40% of agentic AI projects will be abandoned by 2027 due to unclear ROI, cost overruns, and governance gaps.3
The investment is enormous. The results are, to put it generously, mixed.
IBM’s 2025 CEO study found that only 25% of AI initiatives had delivered expected ROI, and only 16% had scaled enterprise-wide.4 Seventy-four percent of companies showed no tangible value from AI investments despite collectively spending $252 billion on it in 2024.5 BCG surveyed over a thousand senior executives across 20 sectors and found that only 4% of companies have what they’d describe as cutting-edge AI capabilities. Four percent!
And yet the meetings continue.
Here’s what I think is actually going on.
Companies got excited about a technology before they were remotely ready to use it. One of the downsides to being unmedicated for ADHD is you get very attracted to the shiny object. Rather than admitting that the company lacks medication, they scheduled more meetings. The problem is not that AI doesn’t work. The problem is that most organizations aren’t in any condition to make it work. Their data is a mess, their governance doesn’t exist, their cost controls are nonexistent, and nobody has a clear answer for what “managing this long-term” actually looks like in practice.
Only 7% of enterprises say their data is completely ready for AI, according to a joint study by Harvard Business Review Analytic Services and Cloudera. Seven percent!6 Seventy-three percent of organizations say they actively struggle with AI data preparation. Informatica’s 2025 CDO Insights survey found that 43% of organizations cite data quality and readiness as their single biggest obstacle to AI success.7 Publicis Sapient put it plainly in their 2026 industry report:
“AI won’t fail for lack of models. It will fail for lack of data discipline.”8
That’s not a technology problem. That’s a homework problem.
The technology showed up and companies realized they hadn’t done any of the foundational work that would make it useful. Rather than doing that work, many of them hired consultants and scheduled workshops and formed AI task forces. Which is exactly what you do when you don’t know what else to do but can’t admit it. Leadership need to look good in front of their peers. Confidence is more important that one may think….
And the governance situation is, if anything, worse. Seventy-eight percent of enterprises are unprepared for their EU AI Act obligations.9 Forty-two percent of companies still lack formal governance frameworks for AI data accuracy and control.10 Gartner estimates that less than 10% of large enterprises had formal AI risk management programs in place as recently as 2023.11 We are deploying systems we don’t fully understand into organizations that haven’t built the controls to manage them. And then we’re writing that off as a strategy.
I keep thinking about RPA.
Remember when robotic process automation was going to change everything?
Every conference had a keynote about it. Every vendor had a platform for it. Every organization had a pilot program. RPA was going to eliminate the repetitive, manual work that was slowing everyone down and free humans up for higher-value tasks. In fairness, RPA did deliver real value in specific, well-defined use cases. Automating invoice processing, handling structured data entry, managing high-volume repetitive tasks where the inputs don’t change.12
But most organizations discovered, usually after significant investment, that RPA is fragile. Change the format of an input and the bot breaks. Try to stretch it into complex decision-making, and it falls apart. As of a few years ago, only 13% of large firms had actually industrialized RPA at scale, despite years of enthusiasm and investment. Most were still stuck in pilots.13
Does that trajectory sound familiar?
I’m not saying AI is RPA. It’s not. The capability ceiling is genuinely different and the potential is real. But the organizational pattern looks identical. Excitement before readiness. Investment before infrastructure. Strategy sessions before the actual strategy. Then, when results don’t materialize on the timeline the slide deck promised, a slow, quiet pivot to the next thing, with the previous initiative left running somewhere in a corner of the organization that nobody officially manages anymore.
What’s different about AI is the scale of the investment and the speed of the hype cycle. Companies spent $684 billion on AI initiatives globally in 2025.14 That’s just what people are willing to admit to or had to report. More than $547 billion of that, over 80%, failed to deliver intended business value. That is not a rounding error. That is a structural problem dressed up as innovation strategy.
I want to be clear that my frustration isn’t with the technology itself. My frustration is with the theater surrounding it.
The AI steering committees.
The use case inventories that never get prioritized.
The proof-of-concept that lives in a PowerPoint for six months because nobody will make a decision about whether actually to deploy it.
The vendor demo that becomes a three-month evaluation process that produces a thirty-page report recommending a second evaluation process.
MIT research found that 95% of enterprise generative AI pilots failed to deliver measurable P&L impact, mostly due to integration, data, and governance gaps, not anything wrong with the models themselves.15 … I mean, companies still struggle implementing things like SAP and Coupa and those are time-tested at this point.
The technology is not the bottleneck. It us… the organization and end users that are the bottleneck. I’ve seen a lot of organizations be very good at scheduling meetings to discuss the bottleneck without actually moving it.
The companies that are getting real results from AI are doing something different.
They’re committing to fewer use cases, not more. They’re investing in data readiness before deployment. They’re building governance structures before they need them, not after something goes wrong. BCG found that leading companies expect 2.1 times greater ROI by focusing on fewer, better-defined use cases rather than spreading AI across every possible function at once.16
The ones winning are the ones who did the boring work first.
None of that requires a steering committee. It requires someone with authority to say: we are not ready, here is what ready looks like, and here is what we are going to do about it before we have another conversation about use cases. To be the C-level executive that says “we aren’t ready”... That takes a special kind of leader and those ones are rare.
I don’t know if this is a bubble. The investment is real, the underlying technology is genuinely capable, and unlike some past technology cycles, the large language models actually do something impressive. But impressive technology deployed into unprepared organizations with messy data and no governance is not transformation.
It’s an expensive hobby.
The sign is still relevant. We’re going to keep having AI strategy sessions until we figure out why no actual AI work is getting done.
Sources
1. RAND Corporation (2024). AI project failure analysis: more than 80% of AI projects fail, at roughly twice the rate of non-AI IT projects. Referenced in WorkOS, “Why Most Enterprise AI Projects Fail.” https://workos.com/blog/why-most-enterprise-ai-projects-fail-patterns-that-work
2. S&P Global Market Intelligence (2025). Survey of 1,000+ enterprises across North America and Europe: 42% of companies abandoned most AI initiatives in 2025, up from 17% in 2024. Referenced across multiple sources including Neuwark and WorkOS analyses.
3. Gartner (January 2025). Prediction: more than 40% of agentic AI projects will be abandoned by 2027 due to unclear ROI, cost overruns, and governance gaps. Eluminous Technologies, “RPA vs AI in 2026: Enterprise Insights.” https://eluminoustechnologies.com/blog/rpa-vs-ai/
4. IBM CEO Study (May 2025). Only 25% of AI initiatives delivered expected ROI; only 16% scaled enterprise-wide. Neuwark, “Enterprise AI Failure Rate.” https://neuwark.com/blog/enterprise-ai-failure-rate-why-85-percent-of-ai-projects-fail
5. BCG, “Where’s the Value in AI?” (October 2024). Survey of 1,000 CxOs across 59 countries: 74% of companies show no tangible value despite $252.3B in 2024 AI spending; only 4% have cutting-edge AI capabilities. Talyx analysis. https://www.talyx.ai/insights/enterprise-ai-implementation-failure
6. Harvard Business Review Analytic Services and Cloudera (March 2026). “Taming the Complexity of AI Data Readiness.” Only 7% of enterprises say their data is completely ready for AI; 73% struggle with AI data preparation. https://www.cloudera.com/about/news-and-blogs/press-releases/2026-03-05-only-7-percent-of-enterprises-say-their-data-is-completely-ready-for-ai-according-to-new-report-from-cloudera-and-harvard-business-review-analytic-services-reveals.html
7. Informatica CDO Insights Survey (2025). 43% of organizations cite data quality and readiness as top obstacle to AI success. Quick Launch Analytics. https://quicklaunchanalytics.com/bi-blog/why-80-of-ai-projects-fail-before-they-start-its-your-data-foundation/
8. Publicis Sapient, 2026 Guide to Next Industry Trends Report (November 2025). “AI won’t fail for lack of models. It will fail for lack of data discipline.” https://ppc.land/data-governance-gap-exposes-ai-confidence-crisis-across-industries/
9. Vision Compliance / Optro AI Governance Stats (2026). 78% of enterprises unprepared for EU AI Act obligations. https://optro.ai/blog/ai-governance-stats
10. Clari Labs Research (January 2026). 42% of enterprises still lack formal governance frameworks; 87% of enterprises missed 2025 revenue targets despite record AI investment. https://www.businesswire.com/news/home/20260114516767/en/New-Clari-Labs-Research-Reveals-87-of-Enterprises-Missed-Revenue-Targets-in-2025-Despite-Record-AI-Investment
11. Gartner prediction: by 2026, 50% of large enterprises will have formal AI risk management programs, up from less than 10% in 2023. Quinnox, “Data Governance for AI in 2025.” https://www.quinnox.com/blogs/data-governance-for-ai/
12. RPA use cases and limitations: Linford Co, “Is RPA Dead?” (February 2025). https://linfordco.com/blog/is-rpa-dead/
13. Only 13% of large firms had industrialized RPA at scale. Neuronimbus, “The Future of RPA in 2025-2027.” https://www.neuronimbus.com/blog/the-future-of-rpa-robotic-process-automation-in-2025-2027/
14. Global enterprise AI investment of $684 billion in 2025; 80%+ failed to deliver intended business value. Pertama Partners, “AI Project Failure Statistics 2026.” https://www.pertamapartners.com/insights/ai-project-failure-statistics-2026
15. MIT Project NANDA (July 2025). 95% of enterprise generative AI pilots delivered zero measurable P&L impact, primarily due to integration, data, and governance gaps. SR Analytics. https://sranalytics.io/blog/why-95-of-ai-projects-fail/
16. BCG AI Radar. Leading companies focusing on fewer use cases expect 2.1x greater ROI. Neuwark, “Enterprise AI Failure Rate.” https://neuwark.com/blog/enterprise-ai-failure-rate-why-85-percent-of-ai-projects-fail