The internet is on fire discussing some recent stories about how the vast majority of enterprise generative AI pilots fail and/or don’t produce measurable value. For example, the New York Times wrote that companies are pouring billions into AI which has yet to pay off, referencing a McKinsey report from earlier this summer which says nearly 80% of companies using gen AI report no material bottom-line impact. Another report from MIT Media Lab claims 95% of gen AI pilots fail, and although that report has serious flaws, it’s fueling the larger conversation nevertheless.
McKinsey is calling this moment of massive adoption with minimal results the “gen AI paradox.” Their advice is just to wait for AI agents to mature so companies can then “reinvent workflows from the ground up.” But what McKinsey (and most of the internet) are missing is that the AI-induced workflow reinvention they’re waiting for is already happening—just not where they’re looking!
The McKinsey report perfectly diagnoses why IT-deployed gen AI disappoints:
Horizontal gen AI use cases (copilots built into productivity apps, enterprise chatbots, etc.) are widely (but thinly) spread across organizations. While they are legitimately useful, this is not fundamentally transformative AI. These horizontal tools are just the latest & greatest tech applied to the existing stack, which is really no different than building a mobile app ten years ago or a web app twenty years ago.
The other type of IT-deployed gen AI, which McKinsey calls “vertical” and actually has the potential to drive deeper & measurable impact, is unfortunately mostly stuck in pilot purgatory, with 90% never making it to production. (McKinsey cites the classic reasons: too complex, too expensive, too much organizational inertia around the status quo…)
So if neither horizontal (wide) nor vertical (narrow & deep) IT-led gen AI solutions are showing real value, what’s a company to do? This is where McKinsey’s suggestion to use gen AI agents which autonomously execute complex workflows comes into play. In fact their report includes several case studies of this with real results, including a bank using human-AI-agent “digital factories” to rebuild and transform old apps, and a research firm that automated data quality using a series of distributed agents.
It’s refreshing to read about genuine positive results while the internet is shouting that gen AI is all hype. However, I can’t overlook the fact that these case studies are for pretty heavy projects, as they detail new architectures, new tech stacks, new governance frameworks, and fundamental workflow redesigns which were needed to get those results. (I assume more than a few McKinsey consultants were also needed.)
So while I agree with McKinsey that the future will be autonomous agents operating at scale, (eventually), I think they’re missing what’s right in front of them.
The revolution is already here
While McKinsey is talking about preparing for that future AI agent revolution, your workers are already living it. They’re just not using the agents McKinsey envisions.
Instead, your workers are using ChatGPT, Claude, Gemini and other widely-available gen AI tools as their agents. As I wrote a few weeks ago, these tools are not simple chatbots, but genuine workflow transformation tools:
This isn’t horizontal productivity spread thin, or vertical complexity stuck in pilot mode. These are real, function-specific transformations happening worker-by-worker, task-by-task, day-after-day, figured out on their own without external consultants or guidance.
The brutal irony is that while companies invest millions into IT-deployed custom gen AI architectures that might maybe work someday, they’re actively blocking the agent-like tools that actually work today.
McKinsey’s future where “agents supercharge operational agility” is literally what your workers are trying to do right now with today’s consumer AI tools. But enterprises force them into absurd workarounds:
Again, McKinsey’s AI agent examples and case studies are compelling. But they’re also all top-down, IT-led initiatives requiring months of planning and millions in investment.
Meanwhile, the conversations around your company’s coffee machines are how individual workers are hacking their way to similar gains with $20/month ChatGPT subscriptions. The transformation McKinsey promises with future agents is being attempted right now by workers with today’s AI. The only thing stopping them is corporate policy.
McKinsey’s idea to build an “agentic AI mesh” by restructuring workflows and deploying sophisticated agent orchestration isn’t wrong, it’s just looking too far ahead. There are immediate opportunities around enabling the agent-like capabilities your workers already use, assuming you can:
The work your employees are doing with AI tools today is the bridge that gets to the future world McKinsey predicts.
There are a lot of challenges which need to be solved to allow your workers to use their AI tools of choice in your environment. But so what? There are also challenges trying to deploy and get measurable value from traditional IT-led horizontal and vertical gen AI deployments. So if you’re going to make the effort and investment, why not focus on the area that has the highest chance of success (and where partial success still provides value)?
Your workers aren’t waiting for the perfect agent platform or for IT to finish its pilot. They’re just using tools like ChatGPT to do their jobs better, and if they accidentally “reimagine” a workflow in the process, then so be it.
Taking a step back and looking at this whole moment, it seems like the real “gen AI paradox” isn’t massive AI deployment with no measurable value. It’s companies pouring money into tomorrow’s gen AI moonshot “maybe” transformations while blocking the much more feasible massive transformation that could be enabled today.
The enterprises who recognize the AI revolution is here today, led by workers, will be the ones who come out ahead. Yes, it’s messy. Yes, it’s happening in the shadows. Yes, there are still big challenges to solve. The question is whether you’ll help your workers use their gen AI tools of choice to continue to transform their workflows, or keep shooting for the expensive gen AI moon while your workers route around you.
Read more & connect
Join the conversation and discuss this post on LinkedIn. You can find all my posts on my author page (or via RSS).
My upcoming public talks
Brian Madden is a VP & futurist at Citrix. He writes about the future of work, AI in the workplace, and the evolution of Citrix.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。
