AI Adoption Challenges in Canva.

In May 2026, Canva’s Chief Customer Officer wrote a confession in Fortune. The company had handed all 5,000 of its employees a full week off their day jobs to do nothing but learn AI. They received the best tools available, guided sessions, role-specificworkshops, a closing hackathon, and 26,000 logged hours of hands-on experimentation.

They expected overnight transformation and significant rises in AI adoption and fluency. But what they got instead was a huge lesson in change management and AI business transformation.

Their CCO, Rob Giglio, recently said:

"The bottleneck wasn't the technology. It was us."

Sit with that for a second. A design-software company that's AI-native, generously resourced, and staffed with motivated people who actively wanted the tools discovered that the hard part of introducing AI into their org wasn’t actually the AI. It was getting humans to change how they work. Changing habits. Changing workflows. His verdict? Most organisations don’t have a technology gap any more. They have a behaviour gap.

I gotta say, I think he’s right. And if it’s true at Canva, it’s even truer at your organisation, where the budget was tighter, the tools blunter, and nobody cleared a week off anyone’s calendar.

This is the real shape of AI adoption challenges in 2026. Not “which model should we buy”, but “Ah crap. We bought it, and nobody’s using it”. I'm going to explore why that happens, what’s actually broken underneath it, and also show you the one discipline that fixes it all. This discipline is so overlooked that even Canva flew straight past without a second thought.

AI adoption challenges start where the technology stops.

There’s a paradox at the centre of almost every AI implementation programme right now. The investment and budget burn is real, and we know the tools work. And yet almost nothing is really changing.

Enterprise spending on generative AI hit roughly $37bn in 2025, more than triple the year before, according to Menlo Ventures. Nearly nine in ten organisations now use AI in at least one business function. And still: MIT’s NANDA initiative studied 300 public AI deployments and found that 95% delivered no measurable impact on the bottom line. In 2025 alone, 42% of companies had abandoned most of their AI initiatives, up from 17% just a year earlier.

So we've got record investment. Near-universal access. And almost no return. What's going on?

If this were a technology problem, newer, more modern technology would have solved it by now. That hasn't happened because the failure isn’t in the tool. It sits in the gap between buying the tool and getting people to genuinely use it. This gap is where AI adoption challenges actually live, and you can’t close it with a better licence tier or using the latest Anthropic release.

What is AI adoption? Access, adoption and embedding aren’t the same thing.

Part of the problem is that we use one word — “adoption” — for three very different things, and quietly congratulate ourselves for the wrong one.

• Access is giving people the tool: a licence, a login, a policy document. Most organisations are brilliant at this and mistake it for the finish line.
• Adoption is people actually using the tool, by choice, in their real work. This is where the overwhelming majority of programmes stall.
• Embedding is the tool becoming the default. It's the new normal that survives after the novelty, the mandate and the launch-day noise have all faded. It’s the only stage that ever shows up in the numbers.

“Deploying isn’t the same as enabling.” — Rob Giglio, Canva

You can deploy to 5,000 people in an afternoon. Enabling them — and then embedding the behaviour so it sticks — is a different discipline altogether. It’s why our service is called AI Adoption & Embedding, not “AI rollout”. The rollout is the easy 10%. Everything that determines whether it works or not is the other 90%.

Why AI adoption fails: a rollout is not a launch.

Canva didn’t have an adoption problem. They had a behavioural change problem, and they did nothing to tackle it beyond providing access to AI.

When a consumer company launches a product to customers, it doesn’t just fire off one email and hope for the best. It works out who the audience is, what they care about, and what would actually make them change behaviour. It builds anticipation before launch. It makes a case to audiences, repeatedly. It proves value, handles their objections, and repeats the message until it lands. In a nutshell, they strategically market to their potential customers.

And yet, when that same company “launches” a tool to its own employees, it sends a single all-staff email and books a training session. That isn’t a launch. It’s an announcement, and announcements get archived in the time it takes to say: "There's nothing in that email for me."

The proof that this is a marketing failure, rather than a motivation one, is sitting in your own data. Your people are not anti-AI. Over 80% of workers already use AI tools at work, and 78% bring their own — unapproved, unprompted, on their own time, according to research reported by CIO. Their behaviours suggest they will happily adopt AI. They just won’t adopt the AI you chose, because nobody sold them on it the way ChatGPT sold itself. Shadow AI isn’t really a security problem. It’s a review of your internal marketing, and the review is damning.

The 5 barriers to AI adoption.

(And why every single one is actually a marketing problem).

If you were to strip back any stalled AI programme, I promise that you're going to find the same five failures underneath. Not one of them is technical. And all five are things a marketer would never dream of skipping.

1. Motivation: nobody sold them the “why”

People use what they want, not what they’re told to want. If you never answered the only question your people are actually asking: “what’s in it for me, specifically, in my job?”, then guess what? They've gone ahead and answered it themselves, and most likely guessed at “more work”. Training teaches the buttons to press, and the prompts to use. But it never makes the case for why they'd use it. You can run every lunch-and-learn going and still watch usage flatline, because a feature walkthrough has never once changed a human being’s behaviour.

2. Message: a launch email isn’t a launch

One message, sent to everyone, lands with no one. Finance, sales, HR and operations all use AI differently and care about different things, but the typical rollout blasts the same generic note at the entire lot. Internal communication that genuinely changes behaviour is segmented, repeated and relevant to the person reading it. Most AI “comms” is a single broadcast, which is precisely why it sank without trace.

3. Trust: fear and fatigue do the quiet killing

These two human forces throttle adoption before it even really starts. The first is fear: if people suspect the AI is there to replace them, they won’t lean in; they’ll keep their heads down and wait it out. Resistance to change is rarely stubbornness; it’s self-protection. The second is exhaustion. Gartner found that employees’ willingness to support organisational change collapsed from 74% in 2016 to just 38% by 2022, while the number of major changes the average worker faced rose from two a year to ten. Your people aren’t so much resistant to AI as worn out by it.

This is a new strain of the broader change fatigue every organisation is already carrying, but this one has teeth. Stack AI fatigue on top and even a brilliant tool reads as a threat to the last fraying thread of someone’s capacity (and sanity).

4. Proof: no champions, no visible wins

We adopt what we see our peers adopting. Without visible champions — real colleagues, not the IT team — and without specific, local wins to point at (“Priya in finance got a day a week back”), there’s no social proof that any of this is worth the effort. Canva understood this in its bones: their breakthrough came from a hackathon and an internal community of “AI Exemplars”, not a memo from on high. Abstract promises of productivity move nobody. Named, nearby results move everybody.

5. Measurement: you tracked logins, not behaviour

A licence-utilisation dashboard tells you who opened the tool. It tells you nothing about whether anyone changed how they work, which actually is the only thing that really matters. It’s the most common trap here: roughly nine in ten organisations have adopted AI, but only about a quarter can actually measure the return. Measure the wrong thing and you celebrate the wrong wins, missing the real failure until it gets expensive. Very expensive.

AI adoption and the long pattern of digital transformation failure.

If this all feels familiar, it should. We’ve watched this exact film before, with CRMs nobody updated, intranets nobody read, HRIS' people hate, and learning platforms nobody logged into. Digital transformation failure has worn a lot of costumes over the last twenty years, but sadly the plot never changes: buy the technology, skip the human change, blame the tool.

What’s different about AI is that the technology genuinely works this time. And that removes the last excuse. When the software was clunky, you could blame the software. You can’t blame Claude et al. The only variable left in the equation is whether you marketed the change to the people who had to live it.

From barriers to adoption: what 82% engagement actually takes.

None of these five barriers is removed by bringing in better software. Every single one is either a communication, motivation, trust or proof problem, which is to say, a marketing problem. Treat your people like an audience you have to win over, not users you can instruct, and I guarantee the numbers will move.

We've made it happen. When we worked with Capgemini and ran an AI and digital-skills programme like a proper internal launch — audience-segmented messaging, a real value proposition, sustained communication rather than a one-off announcement — we drove 82% engagement across a workforce of more than 340,000. Another client of ours went from 5% to 23% platform adoption in 30 days, with no new software at all. Just better marketing of what they already owned.

Canva got most of the way there. They diagnosed the behaviour gap, cleared the calendar, and built the conditions for people to experiment. Fair play to them, after all, this is how it's always been done right? But when it comes to changing behaviour, “give everyone a week off” doesn’t scale to an enterprise that can’t press pause, and it isn’t a repeatable system. The discipline that closes a behaviour gap reliably, at scale, again and again, is the one Giglio walked right up to and didn’t name: marketing. As he put it himself: “The AI was never the hard part. We are.”

Marketing is simply the craft of moving “us”. And when it comes to AI adoption, it's something we all need to do more of. If you want to know how to get started, I've created a guide that shows you the same methodology we've helped our work with Capgemini get impact, and win awards. Download it below.