328. Why AI makes people more productive, but companies aren’t getting ROI

You’re more productive with AI: your presentation gets done quickly (and it looks pretty too!).

But then it gets stuck on your boss’s desk or your client’s inbox. Nothing moves.

Even though they use AI too!

Sound familiar?

Research from Stripe Economics shows that AI has raised our individual productivity, but the efficiency of the wider economy has stalled.

This episode is about why that happens, and what to change in your organisation so your AI investment turns into real results.

In this episode:

• Why dropping AI into a broken process just moves the bottleneck, usually into someone’s inbox
• What the companies getting real value from AI did differently
• A simple way to move towards AI ROI this week 
• Why fixing this is a product thinking skill, and how it makes you the person who bridges the gap

This episode is for you if:

• You use AI every day but don’t feel your team or company moving any faster
• You’re a leader whose CFO is asking where the return on AI spend is
• You advise or sell to organisations and want to show clients where the value is

Resources mentioned in this episode:

• Alice Labs: AI Implementation Case Studies: Real Enterprise Results 2026
• Goat: Does AI Increase Productivity? Unpacking the Impact of Artificial Intelligence
• Stripe Economics: AI and productivity - The story in the data so far (briefly)

Timestamps:

• 00:00 – Why AI tools aren't making organizations faster
• 01:12 – Halloween update and intro to today's topic
• 02:12 – Stripe Economics data on US productivity since 2022
• 06:54 – Why organizations treat AI like plug-and-play software
• 09:18 – McKinsey: what high AI performers do differently
• 10:50 – Why change management matters more than AI spend
• 11:44 – 3 practical steps to fix your AI bottlenecks
• 14:46 – Why asynchronous workflows reduce bottlenecks
• 16:32 – Why process redesign is the real innovator identity 

Follow and Review:

We’d love for you to follow us if you haven’t yet. Click that purple '+' in the top right corner of your Apple Podcasts app. We’d love it even more if you could drop an honest review on Apple Podcasts. Simply select “Ratings and Reviews” and “Write a Review” then a quick line with your favorite part of the episode. It only takes a second and it helps spread the word about the podcast.

Listen to our podcast on:

• Apple
• Spotify
• YouTube
• Audible
• Pandora

Connect with Sophia Matveeva on LinkedIn: https://www.linkedin.com/in/sophia-matveeva-556365a/?isSelfProfile=true

Transcript:

[00:00] Tell me if this sounds familiar: you've got AI tools, you're more productive individually, and yet somehow nothing moves faster. Your memo sits on your boss's desk, and your clients don't respond to your proposals any sooner, even though you know they're also using AI. That's not a technology problem — this is a people and processes problem. This episode is about why that happens and how to fix it.

Welcome to Tech for Non-Techies. This is a podcast for business leaders and non-technical founders building the future in the age of AI. Whether you've been in business for a hundred-plus years and you're looking to modernize, or whether you're building something new, this is the show for you. You'll learn how to come up with new ideas and make them come to life, no matter the size of your organization. I've taught tech and innovation at Oxford University, advise companies like Microsoft, and written for the Harvard Business Review. You're going to hear the frameworks and the thinking that I've built, tested, and taught at the highest level. And now it's your turn. Let's get started.

[01:12] Hello, smart people, how are you today? How was your Halloween? Did you wear something fabulous, go to a fabulous party? Halloween in London is usually very fun, but it's not something I've been indulging in, because I'm definitely in my mom era — I've been going to bed around 8:30pm. Until my son is old enough to go trick-or-treating, I think Halloween is going to be something I remember with fond memories. But the parties in London for Halloween are wild — if you're in the phase of your life where you want a wild party, Halloween in London is the place, also New York, even better. But I digress.

We're going to talk about a serious subject today: whether AI is actually increasing productivity, where it's working, and where productivity gains haven't been happening at all. There's actually a lot of media coverage on this right now — I saw on Instagram that the chief economist of a major bank was saying they'd studied the data and found that, despite all this investment into AI, it hasn't increased profits, there hasn't been much productivity gain.

[02:12] I want to share an article I read from Stripe Economics — they've got really good research, it's the economics arm of Stripe, so they've clearly got interesting payments data, but also some clever economists. Their newsletter is really good. They published data on US productivity since 2022, and it tells an interesting story.

They tracked two things — linked in the show notes, as is most of the research I'll mention. First, how much the average worker produces per hour. Second, something called total factor productivity — for those who studied macroeconomics, you might remember this: it asks how the whole economy performs, how much the economy gets out of what it puts in — people, equipment, money, all of it. It measures how efficiently the entire system works, not just how fast individuals or firms work.

Both numbers went down from 2020 to 2022, then started bouncing up in 2023, 2024 — and then in 2024, they split. This is where it gets interesting: output per hour has stayed high, still growing at roughly 2.5% per year. But the efficiency of the whole system has plateaued — by the start of 2026, it was growing at roughly zero.

So the story this tells is: we're producing more per hour, but the system around us isn't getting any more efficient. Think of it as an inner loop and an outer loop — your individual work is more productive per hour, but the outer loop, the entire organization, isn't speeding up. This is really true in large organizations — in startups it's faster, and I'll tell you why.

So, AI is speeding up the individual, but not the organization as a whole. A good example: you can create output faster — a presentation, a memo — let's say you need a memo for your boss, it's done quickly, designed well, you put your brand guide in and get a beautiful design — this is what I'm doing with Claude, love Claude's design. You send this beautiful, thoughtful memo to your boss, and it just sits there. Your boss doesn't respond, and you need their response to take steps forward — but because they haven't responded, nothing moves. AI sped up some of it, but the whole thing is still really slow.

So let's think about what productivity actually is. If we just think of "output per hour," that's useful for economists, but we want to think more holistically. I like this definition from Gloat, also linked in the show notes: "Productivity in the AI era goes beyond the traditional concept of more output per hour. It's about fundamentally changing how work gets done — not just doing existing tasks faster, but reimagining entire workflows to eliminate bottlenecks, reduce errors, and free up human employees for higher-value activities." That sounds pretty good — that's what we want: to be freed up for high-value activities, and bottlenecks eliminated, so when you send your thoughtful memo, your boss actually responds and the whole thing moves quickly.

[06:54] So before the solution, let's diagnose the problem. What I'm seeing is that many organization leaders are used to buying software — software as a service — and thinking, "We've got Calendly, we've got Salesforce, we plug it in, upload our data, and good things will happen." Many organizations treat AI like a plug-and-play SaaS tool you just drop in and expect results from. Or, like the memo example — you're using AI to create a thoughtful memo or presentation, but it's going into the same broken process. The process is still: it has to get to your boss, who has to review five other memos, has been put on some committee, there's an emergency — so even though your boss genuinely wants to read your memo, it's frankly not going to happen given everything else going on.

So the process is broken. Putting a great AI tool into a broken process just moves where the process is broken — AI moves the bottleneck from you working on the presentation for ages, to your boss's inbox. The bottleneck moves up the value chain, it doesn't disappear. If the process requires four sign-offs and two committee meetings, making the document faster won't help. This especially applies to big companies, and especially governments — I've had government clients, I love working with governments, you get the biggest impact, working with long-term-vision people — but the bureaucracy is insane. My government listeners will definitely identify with "four sign-offs and two committee meetings" — probably more like six.

[09:18] So let's look at what high performers did, so we can copy them. McKinsey studied this in 2025 and found that companies getting real benefits from their AI investments are three times more likely to have redesigned workflows before introducing the tool. According to McKinsey, workflow redesign was the single strongest predictor of EBIT impact — outweighing which model you used, whether free or cheap open-source Chinese models, or paying for ChatGPT. That didn't matter much. What mattered was whether you laid the groundwork, thought through the process, before implementing the tool. Interesting, right? Workflow redesign also outweighed training spend — training everyone on AI doesn't matter much if the underlying process is still bureaucratic.

[10:50] There's more research from Aliff Labs I really want you to remember, because it's wild: for every one dollar spent on AI technology, nine dollars needs to go into change management, training, and workflow redesign. My jaw literally dropped reading this — this is what excites me in my forties now, the ratio of change management spend to AI spend. Who would have thought — my 16-year-old self would be very surprised. Anyway: for every dollar spent on the tech, you should be spending nine times more on people and processes — change management, training, workflow redesign.

So in this whole AI era, where we think AI can do anything — and it genuinely can do a lot — what's actually standing in the way of technological progress is, as always, us.

[11:44] If you've been feeling jealous of a competitor's AI budget, you can stop — maybe they're outspending you on the technology, but you can do change management properly, work on people and processes yourselves. Yes, I'd recommend working with a consultant, working with me — but you can also do a lot of this yourself.

So here are practical steps. Number one: take a single process — don't try to reimagine the whole organization, you'll be stuck in analysis paralysis. Take one process with one outcome, handled by one specific team or part of the organization. Look at that process like an operations person — where are the bottlenecks? Really map it out: in order to hire a new consultancy, work with a new vendor, buy something, or even a small acqui-hire — whatever it is — what actually needs to happen between "we want to do this" and "we are doing it"? What are the steps, who's involved, does every step actually need to happen, do you really need all those committee meetings, can they be slimmed down? Optimize that process first, and only then bring in AI — and track what's actually working, because you'll identify bottlenecks you didn't foresee, and then figure out what to do about them. This is something you can start implementing today.

I'm recording this in November, and the CFO is going to be very interested in AI ROI — especially at year-end, they'll be looking closely at those numbers. So this is worth looking at now — even if you're told, "You've overspent on this and we're not seeing results," you can say, "We've actually understood where the issues were, we've mapped out the processes, we've seen where we can genuinely get ROI, and in the last couple of months we've started benefiting more than before." This is a learning process — the AI revolution is still early days, so it's normal to get things wrong, realize the bottlenecks are human, and fix them.

[14:46] Step one: map where the real bottlenecks are. Step two: ask which human handoffs in your process could be eliminated, shortened, or made asynchronous. Instead of a committee meeting requiring everyone's agreement, could you say, "The deadline for everyone's feedback is this day, then we'll incorporate it, and so-and-so signs off"? Just organizing everyone onto the same call in the same time zone can be really annoying — making something asynchronous can already be a huge bottleneck reducer. Start with one process, not the whole organization — pick something measurable and get going.

If you're wondering, "Should I be leading this, who should lead this, will anyone agree?" — my advice is just take the reins and get on with it. We're in the early days of this whole AI revolution — the person who steps up and says, "I have ideas, I know what to do here," is assuming the leadership identity, and very often people will just go, "Okay, that's the person who knows what to do here."

[16:32] The people who will thrive in this transition aren't the ones who know the best prompts and use AI the most — they're the ones who can see where the process is broken and redesign it. This is applying product thinking to your organization — this is becoming an innovator in your organization. You don't need to invent the iPhone to be an innovator who makes an impact — if you redesign a process to flow more smoothly and get serious AI ROI, that's already an important innovation that flows through to the bottom line, and then to your paycheck — or, if you're doing this for clients, that's real value delivered.

This is exactly the innovator identity I really recommend you adopt in this age of AI — seeing what others have missed, and making something new come to life, whether that's a new process, a product, or an innovation. Because AI, as Stripe Economics has shown us, speeds up the individual — but people and processes speed up the organization. Remember that one-to-nine ratio: one dollar spent on the tech means nine spent on change management, training, people, and processes. That's the optimal number.

So my suggestion is: become the person in your organization, or for your clients, who bridges that gap — who can see where the outer loop is broken and knows how to fix it. If you're selling AI services or implementations, this is how you go about it — this is innovative thinking, this is product thinking, which is really what we focus on here on this podcast. I'm actually about to kick off a special series specifically on honing your creativity and product thinking in the age of AI, so you can be the best thinker, the best analyst, at a time when quite a lot of people are losing their heads.

Because right now, in the age of AI, the people who win are the ones sharpening their own human instrument — their brain — not outsourcing it to others. That's what I'm helping you do on this podcast, and how I work with my clients: we never outsource our judgment to AI. We ask AI, but we make the final decision — using our critical thinking to find new solutions, think strategically, and imagine things that haven't been imagined before. That's what winning in the age of AI is all about.

On that note, my dear smart person, I'm asking you to leave this show a rating and a review, because I thought this was a good episode — clever, well-researched, and helpful — and if you're still listening, I'm assuming you agree with me. So give me some love: leave the show a five-star rating wherever you get your podcasts, and if you're feeling extra generous, a note on Spotify or a review on Apple would make my day.

On that note, I'm wishing you a wonderful day, and I shall be back in your delightful smart ears next week. Ciao.