This One Chart Exposes Why Most Companies Are Failing At AI

Summary

Marketing Against The Grain – Episode Summary

Episode Title: This One Chart Exposes Why Most Companies Are Failing At AI
Hosts: Kipp Bodnar (HubSpot CMO), Kieran Flanagan (HubSpot SVP of Marketing)
Date: March 10, 2026

---

Episode Overview

In this engaging episode, Kipp and Kieran dissect a widely-shared chart from Anthropic that illuminates a massive gap in AI adoption across industries. Their central thesis: the winners in AI won't be determined by who has the best models, but by who transforms their business processes to become truly AI native. They introduce a new, practical framework for closing the “adoption gap” and deliver actionable insights for marketers and business leaders navigating the era of AI-driven transformation.

---

Key Discussion Points & Insights

1. The Anthropic Chart – A Misinterpreted Lightning Rod

• [01:34] Kieran explains a viral chart by Anthropic showing the “theoretical AI coverage” (blue) versus “observed AI coverage” (red) across industries.

  • The blue area: how much of an industry can theoretically be automated by AI.
  • The red area: what’s actually been automated in practice.
  • Insight: The focus should be on the gap between theory and reality, not on improvements in model performance.

• [03:38] Kipp elaborates:

  “Everybody’s looking at this chart and they’re looking at this blue area, which is theoretical, right? And they're like, oh gosh, AI is going to wipe out a lot of the jobs… what’s most interesting… is that there’s this massive gap between the red and the blue...”
  —Bodnar, 03:38

2. Why Great Models Alone Are Not Enough

• [01:08] Kieran introduces the episode’s core argument:

  “We’re kind of here to argue in this 12 to 15 minute video that it doesn’t even matter, that this model does not even matter… Model capabilities are not the important thing right now in the AI industry.”
  —Flanagan, 01:08

• [03:38] Kieran notes:

  • Most models are “already very, very good.”
  • Further advances in pure capability aren’t holding companies back.

3. Historical Parallels: Electricity’s Slow Adoption in Factories

• [04:31] Kieran tells a story comparing AI’s current adoption to electricity’s rollout in the 19th century:

  • Edison commercialized electricity in 1881, but by 1900, only 5% of American factories used electric motors meaningfully.
  • Factories that simply swapped steam for electricity, without redesigning their workflow, failed to capture its true benefits.
  • The productivity explosion came only when factories restructured around electricity—a metaphor for the change needed in today’s organizations to leverage AI fully.

  “The productivity explosion… only happened when they redesigned their factory floor around electricity... That's going to be the fundamental shift… Not AI model capabilities, but redesigning the company to be an AI native company.”
  —Flanagan, 05:36

4. The Real Bottleneck: Humans and Business Processes

• [07:24] Kipp makes a key point:

  “Honestly, one of the most important points we’ve ever made on this show is that the bottleneck is humans. That human bottleneck is going to take—I don’t know, Kieran, what do you think—decades to fully get through?”
  —Bodnar, 07:24

• [08:20] Kieran cites striking statistics:

  • “84% [of people surveyed] had never used AI. Right. That's how early we are.”
  • “Only 8.6% of companies have even deployed an AI agent in production.”

• The adoption gap between top-performing companies and the rest is vast and growing.

5. The Massive Opportunity: Closing the Gap

• [08:56] Kipp:

  “The single biggest opportunity on this chart is actually the gap between the red and the blue. If you can help any company transform and become AI native, then there’s a lot of opportunity, a lot of money to be made there.”
  —Bodnar, 08:56

6. The Rising Cost of Scaling AI Internally

• [10:00] They discuss the current intense subsidization of model costs and the real, often large, expenses involved in deploying AI at scale.
  • The economic reality: “It’s going to be tens of thousands of dollars [to close the gap], not hundreds.”

7. Introducing RAPID5: The AI Adoption Playbook

• [10:44] – [12:57] Kipp unveils a practical framework, RAPID5, distilled from analysis of 20+ leading AI transformation models:

  • R: Reveal – Assess current workflows and the “jagged frontier” of where AI fits.
  • A: Architect – Design a target operating model, including before/after workflow mapping.
  • P: Prove – Run two-week real-world pilots, measuring across three horizons: efficiency, capability, and transformation.
  • I: Ingrain – Shift from tool adoption to identity shift, making AI central to work culture.
  • D: Dynamize – Reassess quarterly, adjusting for AI’s rapid advancement.

  “The new framework is called RAPID5… what I had to do was go and research current best practices in AI transformation, look at all the current frameworks, rethink them and actually make a new framework for us..."
  —Bodnar, 11:00

• Inputs needed: Team profiles, workflow documentation, current AI state, transformation goals, leadership/culture context, risks, competitor benchmarks, skills inventory.

8. How to Get Started

• [14:12] Kieran suggests:
  • Record videos of team workflows (using Loom), transcribe them, and feed into the AI “skill” to map out automatable processes.
  • The skill enables companies to self-conduct “forward deployed AI engineering”—not just relying on external specialists.

---

Notable Quotes & Memorable Moments

• "You cannot walk into the future if you are looking back at the past."
  —A founder at Kieran's dinner, echoing the need for rethinking business with AI (04:39)

• "Model capabilities are not the important thing right now in the AI industry."
  —Kieran Flanagan, 01:08

• "There’s this massive gap between the red and the blue… that’s the most interesting thing.”
  —Kipp Bodnar, 03:38

• “Only 8.6% of companies have even deployed an AI agent in production.”
  —Kieran Flanagan, 08:20

• "The real gap for a lot of businesses is from where they are to where they could be."
  —Kipp Bodnar, 15:21

---

Important Timestamps & Structure

• [01:08] – Argument: Model improvements don’t matter as much as process redesign
• [03:38] – Interpretation of the red/blue adoption gap
• [04:31] – The electricity metaphor and history lesson
• [07:24] – The human bottleneck and AI’s slow real-world adoption
• [08:20] – Surprising statistics about enterprise AI adoption
• [10:44] – Introduction and breakdown of the RAPID5 framework
• [14:12] – Practical tips for documenting/transforming workflows
• [15:21] – Call to action and summary

---

Tone & Style

Kipp and Kieran maintain a candid, practical, “no-BS” approach. They push listeners not to chase model upgrades, but to “redesign their factory floor,” so to speak, embracing identity-level transformation. Their focus is empowering marketers and business leaders with actionable frameworks—no fluff or recycled Twitter advice.

---

In Summary

The main lesson: If you want to win at AI, it’s not about having a smarter model, but about rethinking your workflows, team structures, and company culture to become truly AI native. The real opportunity is in closing the gap between AI’s theoretical potential and its practical, everyday application—and that requires new skills, continual adaptation, and a strategic approach like RAPID5.

“Huge opportunity. You might be ahead of your competitors, but you’re probably behind the market opportunity. Big market opportunity.”
—Kipp Bodnar, 15:21

---

If you want a step-by-step playbook for AI transformation, keep an eye out for future episodes and the upcoming Marketing Against the Grain newsletter, where Kipp and Kieran promise to share refined tools, frameworks, and real-world examples.

Transcript

[A]

On today's show, we're going to show you the single most important chart in AI that explains why your company isn't going to be a winner just because you use the best models. You heard that right. The best AI models are not going to dictate who wins and who loses in AI. We're going to give you the actual winning formula and how you can implement it at your company today. All of that and more on this episode of Marketing against the Green.

[B]

Here's a quick word from HubSpot. HubSpot helped Tumblr solve a big problem. They needed to move fast to produce trending content. But their marketing team was stuck waiting on engineers to code every single email campaign. Now they use HubSpot's customer platform to email real time trending content to millions of users in just seconds. The impact, three times more engagement, double the content creation. Want to move faster like Tumblr? Visit HubSpot.com.

[A]

All right, Kip, we are here with yet another AI model. GPT 5.4 is out. Another model beats lots of benchmarks. Apparently one of the best models on the planet. And we're kind of here to argue in this 12 to 15 minute video that it doesn't even matter, that this model does not even matter. Doesn't matter how good it is. Model capabilities are not the important thing right now in the AI industry.

[B]

Yeah, we're going to instead tell you what the most important thing instead of the models getting better, what that is and how you can leverage that for your business to actually change growth in this new era. And Kieran, based on that, there's a chart that's going viral from Anthropic that I think was the lightning rod moment for the conversation we wanted to have. And a lot of people are interpreting it one way. We have like a very different interpretation of it.

[A]

So GPT5.4, great coding model, great intelligence model, beat em and Tropic and lots of benchmarks. But we've been here before. Every single model that comes out kind of goes a little bit more up the benchmarks. Now there are ways that they kind of align themselves to benchmark to do really well. But we are believers that model capabilities are already very, very good. And actually given people even more capable models is not going to make much of a difference right now because of this chart. So this chart is being shared pretty widely around X and it's from Antropic. So fair play to Antropic. They're putting out these to try to show what the impact of AI could be across industries. So what it's showing you is theoretical AI coverage, which means how much of that industry could be theoretically automated with AI. And there are no surprises here. AI is really good at coding and math. It's very good at finance, it's very good at engineering type roles, it's very good at legal, it's very good at like arts and media. Pretty great at all of the office and admin work. So it can get you a lot of coverage in a lot of different industries. And the red is the observed AI coverage. Just how much AI is being deployed within this industries and how much of that work it is automated. And I think what you and I talked about on Slack is we don't believe that for the average AI user, the model is already good enough. It really is not going to matter this year if they get another model. Another model, Another model. The really hard thing about AI is actually integrating it into your existing workflows.

[B]

I think one important thing to understand is like, I think everybody's looking at this chart and they're looking at this blue area, which is theoretical, right? And they're like, oh gosh, AI is going to wipe out a lot of the jobs in all of these categories. And like, that's very theoretical. Might happen. Might happen a year from now, might happen a decade from now. We don't know what's the most interesting. I think what you and I are interpreting this differently is that there's this massive gap between the red and the blue. And there's some work that's happening in coding and in business process and in sales and some of these customer service, these really important markets, but it's like still small relative to what the perceived opportunity is. And that blue is really kind of the perceived opportunity. So if we all watching the show today agree that like, oh, there's some theoretical, much bigger opportunity versus where we are now, what the heck does that mean? Is the big question.

[A]

We just dropped the AI adoption playbook. It gives you the exact framework to actually redesign your business with AI. It's a proven roadmap that helped one company book 11,000 meetings and another to resolve issues 39% faster. Get it for free. Click the link in the description. So there's a good story that this has happened, you know, once before in time. So I was giving you a good quote from a dinner I was at and I'm with a bunch of founders and one of the founders in there is incredible. And he had this incredible quote where he said, you cannot walk into the future if you are looking back at the past. And I think this story I'm going to go through kind of explains what's happening, which AI is being compared a lot to electricity, right? It's like a fundamental thing that everyone is going to access and get a lot of value from. And Thomas Edison built electricity generating stations and in Manhattan and London in 1881. And within a year, electricity was being sold as a commodity. So we had electricity as a commodity in 1881, this incredible new invention. But by 1900, less than 5% of mechanical drive power in American factories came from electric motors. And so companies had not integrated into how they were doing work. And if you looked at that, it's because those factories kept their old layouts. They just swapped steam for electricity and then ran the same processes. So think about that in terms of AI. The productivity explosion of electricity when in factories only happened when they redesigned their factory floor around electricity, right? So they started by stepping into the future with electricity, by looking back at the past, the same old processes. I'll keep the same processes. I will swap steam for electricity and I'll just do that thing. And There was only 5% adoption, and it got wide adoption when all the factories, when they redesigned their exact factories around AI. That, I think is going to be the fundamental shift that happens in companies. Not AI model capabilities, but redesigning the company to be an AI native company. Team structure, skill sets, how people do work. And that's going to take way more time, way more these companies to come out with another model. Another model. There's a cool thing you can do now. Go to perplexity, go to ChatGPT, go to any one of these models and ask it to build a table of open jobs in Google, DeepMind, Claude, Entropic and OpenAI and tell it to show you who they're hiring for. Because if you ask who they're hiring for, it tells you a lot about their strategy. And I do this all the time. They are hiring for Ford deployed engineers, right? People who can help companies integrate it into that business, but not swap steam for electricity. But how you start to redesign your business around these models, I think this

[B]

is honestly one of the most important points we've ever made on the show is that the bottleneck is humans. That human bottleneck is going to take, I don't know. Kieran, what do you think decades to fully get through?

[A]

I think it depends if we get agents that get deployed as employees and can actually do a bunch of that work internally. But for the average customer, you and I have talked about this, like there's like we go to dinners with AI founders and it's a different planet. Everyone's in there like Claude MD file trying to like get my open claw agent to like run a bunch of my stuff. Then there's a category who are kind of doing some stuff with AI and they're like, hey, I'm using ChatGPT on a day by day basis, asking it some questions, maybe getting it to write an email. And then there's a group of people are like, what the are you all talking about?

[B]

That's the majority of humans, by the way. There was a study, vastly majority.

[A]

Yeah, There was a study recently maybe from OpenAI or another one of these companies where like normal people, just people surveyed, 84% had never used AI. Right. That's how early we are. If you look at some of the data, only 8.6% of companies have even deployed an AI agent in production. Right? There is such a small amount of usage. And OpenAI had a state of enterprise AI report, actually that was pretty interesting where it just showed the top 5% in terms of intensity, in terms of usage were orders of magnitude way ahead of everyone else. Like the gap between the top percentile and everyone else has never been greater.

[B]

Well one, it shows you that the single biggest opportunity on this chart is actually the gap between the red and the blue. And like, if you can help any company transform and become AI native, then there's a lot of opportunity, a lot of money to be made there because like we said, the models are smart. They can do all of these things.

[C]

Before we continue, let me tell you about a podcast I love Marketing School. Marketing legends Neil Patel and Eric Su bring you daily actionable digital marketing lessons learned from years in the trenches. Marketing School delivers bite sized marketing wisdom you can implement immediately. Whether you have a new website or you're an established business, you'll learn the latest in SEO, content marketing, social media, email marketing, conversion optimization and general online marketing strategies that work today. They just did a great episode called 10 SEO Lessons that Still Work in the AI Era. Listen to marketing school wherever you get your podcasts.

[B]

I think what's going on, Kieran, is a few things. One, first of all, all the cost of these models is greatly subsidized right now. And like if you compared the actual cost to get from red to blue, like in terms of actual costs versus a non, you know, subsidized cost, it would be pretty expensive, right? Like the amount of compute that you're paying for inference that you're paying for to close that gap, it's going to be a lot. It's not going to be tens of dollars or hundreds of dollars. It's going to be tens of thousands of dollars. Right?

[A]

Yeah, I agree with that. Okay. But we are not the episode just to give you the opinions, just to give you like here's what we think. We want to give you the practicalities. Kip, you've been working on a skill that can actually help companies close that gap.

[B]

Okay, so I'm here in Perplexity Computer, which is one of many AI tools. It's kind of inconsequential to what we're using here. And I had to do a task. I gave it a very complex prompt. It ran for well over an hour. And this is my V1. I'm going to edit through this and I'll give you the skill at the end of the day. But what I had to do is basically can it build a skill that you could use in any AI tool? ChatGPT, Claude Perplexity Claude code. What have you any tool to actually help you take your team yourself, your team, your company from being that factory that isn't set up for electricity to your metaphor, Kieran to being an AI native organization to actually understand what it would take. Okay, so Kieran, what I really had to do was go and research current best practices in AI transformation, look at all the current frameworks, rethink them and actually make a new framework for us and a new skill that's going to help people take through. So here's what it did. It researched 20 existing frameworks. It did a bunch of real world case study analysis, it did a failure analysis and current best practices from McKinsey and HBR and Wharton and all the fancy folks. Here's the framework here and I want your opinion on this. The new framework is called Rapid 5 and R is reveal. Assess the team's actual workflows, maturity and map the jagged frontier of where AI helps versus hurts their specific work. A architect design the target AI native operating model with workflow by workflow before and after designs, technology selection and change management. P prove implement through two week sprints on real world pilots, not synthetic measured across three horizons. Efficiency, capability and transformation in grain shift from tool adoption to identity shift through peer alerting, AI first defaults and performance integration. Indeed dynamize build a 90 day reassessment cycles because AI capabilities change quickly. All right, what do you think about the framework here?

[A]

Yeah, so I say so that's changes quarterly and that's important because it actually understands that things are changing in a fluid way. So this is really good, what you're really doing and what we're showing folks here. And I think if you really want us to go deeper on this specific use case, you're building a 4 deployed AI skill for deploy engineers. So people, our listeners understand they are going into companies, they are working for OpenAI, Claude, working for these different AI products and they go in and they implement AI into your business. And so when I've talked to a lot of AI founders, their biggest problem is that companies are struggling to integrate AI and agents into their workflows, which is what that chart is saying. So they hire these specialists who go in and do that for you. What you're actually kind of building is a skill for the average business to be able to do for deploy AI engineering within their own company. And so this here is like a pretty good framework. What is the inputs to run this skill?

[B]

The skill requires essential team profile, core workflows, 5 to 10 current AI state and transformation goals. Like that's the bare minimum important data, systems, environment, leadership, culture, context constraints, risks, and then helpful competitor benchmarks, individual skills, inventory and prior performance data.

[A]

Okay, so a lot of this is going to be tricky to get actually. I think the takeaway for folks here is if you want to close the gap between the red and the blue for your business because you think there's value in integrating AI across a certain team, that you can actually ask AI to actually start to build a skill that will allow you to kind of give it some input around that team. And I think to your point, one of the suggestions would be have your team just do a bunch of looms of how they work and then take those transcripts and then give it to your skill. And the skill extrapolates that using some sort of framework like you have here, and then creates the ways that AI can automate those things. And maybe we'll show that in practice. Maybe we'll do an episode where we

[B]

show let's do a part two of this where we'll like really do the screen share, we'll try to simplify the inputs and you and I will just keep iterating on the skill like offline and part of that episode until we get it like really, really tight and then we'll give it to everybody. So I'm sure there's some people who want the skill now. We will give the skill away Once we think is like really good. We want to go through a bunch of fictional examples and then we want to do a couple real world examples and then we'll know. It's like pretty good.

[A]

Cool. Okay, that's the episode.

[B]

Huge opportunity. You might be ahead of your competitors, but you're probably behind the market opportunity. Big market opportunity. I think it's funny that Anthropic put that chart out. There's a bunch you could talk about it. But I think the most interesting thing is that there's a real gap for a lot of businesses from where they are to where they could be. Appreciate you watching Hit like hit subscribe. We'll see you real soon on Marketing against the Grain. Foreign. You know, Kieran and I have been doing the podcast for a while now. We've been at this for a couple years. We love it. We could not be happier to be doing this, but we wanted to take things to the next level. We want to level up the impact we're having with marketing Instagram. So the next step of our journey is something we're really, really excited about. We're going to launch the the Marketing against the Grain newsletter. And Marketing against the Grain newsletter is going to be amazing. If you are a marketing leader practitioner, you're in the trenches doing marketing every day. This is for you. We're going to deliver right to your email inbox and you're going to get all the behind the scenes frameworks, practices, tutorials from us, from guests we have on the show, and from people even beyond the podcast that we think are going to be helpful and and really have an impact on your day to day, week to week doing marketing. You're going to love it. It is something we've been talking about for a while. We're really excited to have it out in the world. We've already got a hundred thousand marketers who are on this newsletter. Please join. It's completely free. We'd love to have you as part of the Marketing against the Grain community. And it's easy. You can click the link in the description below or you can head to marketingagainsthegrain.com subscribe.