Kickframe Newsletter

Media VP Kelvin Mak on his AI workflow, plus why big companies should approach AI like small businesses

A few months ago, APG Canada ran a fun event called the AI Battle. Two teams, one from a creative agency and one from a media agency, each got a marketing brief and two hours to respond. Both could use AI. What they produced in that window was impressive, and it was interesting to see how these two flavours of strategists put AI to work differently.

After the event, I met one of the participants, Kelvin Mak, VP of Innovation and Intelligence at UM. He shared a lot of smart thinking about how he uses AI in his work. So, naturally, I badgered him into showing up in my newsletter to share some of it with you.

Big Business Should Act Small

Derek Thompson has a new post, “It’s the Best Time Ever to Become a Millionaire by Working Alone”. He explores how AI is making it possible for individuals (like me) to start or grow a business. I’m still waiting for my millions to roll in, but he’s hitting on something that I’m seeing in my work.

The more meetings I have with organizations, the wider the gulf looks between their situation with AI and mine. When I talk with other independent consultants, we regularly share how helpful these tools are, giving us capabilities and support we wouldn’t have otherwise. So why don’t I see the same enthusiasm with teams inside larger organizations?

The reasons are obvious, but I think they’re a big deal. Large legacy systems, loads of disconnected data, regulations and legal requirements, tight IT governance, and staff who may be wary of adoption. Plus, everyone is busy. They don’t want to spend their evenings and weekends learning and experimenting with AI on their personal computers.

A small business like mine is far freer to experiment and pivot to what works. I don’t carry the same pressures around internal optics or the fear of a manager eliminating my role. For parts of my work, a tool can be “good enough” without the same downside risk. I have the luxury of experimenting and using AI in my job, and the motivation to do it because I see the benefits firsthand. (Being a small business owner has its downsides too, but I’ll save that for another post!)

Maybe this is short term. Small businesses are better at capturing quick wins and trialing new tools, but larger companies have the capital, proprietary data, and expertise. Until they catch up, I think there’s real risk for large teams in putting off AI use until “the business figures it out.” Business leaders should find ways for their teams to meaningfully experiment and use AI in their actual jobs: build a new capability, reimagine the workflow for a priority area, or set up an incubator for product trials. Grow by thinking like a small business owner.

AI This Week, in 5 Charts

The Work AI Index 2026 (Work AI / Glean):

As we all spend more time using AI, we want too make sure that time isproductive. But hours are going into work nobody planned for: providing context, supervising output, debugging mistakes. It even has a name, ‘botsitting’. AI may be a productivity tool, but we still need to use the tool productively.

Strategy Matters More Than Tools (BCG):

More than two-thirds of regular AI users report higher job satisfaction since adopting the technology, but 41% of all respondents also report increased mental strain from using it. The Glean study above cites this same cognitive load as a reason people accept inferior AI-generated output (aka botshitting).

AI Economic Index, June 2026 (Anthropic):

Plenty of research covers how people use AI. Less covers when. This look at Claude usage shows how AI reflects the rhythms of daily life: recipe requests peak around 6pm, sleep advice around 5am.

Digital Consumer Trends – UK (Deloitte):

Frequent use, but still basic use. The most popular work applications are research and writing emails, 60% of respondents haven’t heard of agentic AI, and over a quarter of AI users still don’t realize AI can generate inaccuracies.

AI Models & Values (Economist):

What worldview does your AI model hold? This chart visualizes how different AI models perform based on World Values Survey. The models cluster in the quadrant populated by richer countries, a product of their training data and how companies conduct post-training.

Cool Beans

Last week I vibe-coded a new Kickframe website using Lovable. The old one needed an update, and I was curious how user-friendly the latest tools have become for a non-technical audience. The verdict: fairly simple, and pretty fun. To be clear, the tool doesn’t replace the thinking. You still need a clear view on web strategy, design, and content. But if you haven’t tried building something with one of these tools yet, give it a go, even if you never use the end product. As I share a lot in my training, using is learning. Use this kind of tool to learn how it might benefit you and your team.