Notes, articles, talks.

Adaptive Personalisation. The more ambitious version. I looked at how a global automotive brand and its agency partner approached personalisation. The organising idea was striking: no two digital experiences should be the same. Express interest in something, and the next time you land on the site, the entire look and feel has reshaped around your inputs. One user's view of the brand looks nothing like the next - different cars, different content, aligned to lifestyle and habits. What the case actually signals about where the space is heading: → Brands are pulling data, marketing and e-commerce under a single partner instead of stitching together specialist vendors. → Making segmentation, integration and real-time delivery work together across regions — each with its own data and rules — is the real challenge. → The technology is commoditising but the discipline to measure whether tailoring actually pays is not. And that's the gap! → ROI is becoming the organising principle. The brands pulling ahead treat measurement as a first-class pillar, not a reporting line bolted on at the end. The takeaway for anyone weighing their own strategy: mature personalisation is less about which tool you buy, and more about the operating discipline to make it pay.

I keep meeting brilliant AI strategists who can't ship. Not because they're not smart. Because strategy and “vibe-coded” prototypes are cheap now — and the thing that's actually scarce is the technical weight to turn a prompt into something a bank would trust in production. Every firm I talk to has horizon scanners. Almost none have enough builders. And in financial services, that gap is starting to hurt. Talking fluently about AI used to be a differentiator. Now it's the price of entry. The firms pulling ahead aren't scanning harder — they're building better. A few things I keep seeing separate the two: → A prototype built in two hours is a great demo and a fragile product. Rapid generation isn't robust engineering. → “PoC purgatory” is real — constant innovation, nothing reaching the balance sheet. Shadow AI, spaghetti code, and security gaps all hide in the demo and surface in production. → Ground-truth beats a chatbot's guess. When gatekeepers blocked a customer study recently, I paid for real interviews rather than pivoting to a synthetic summary. → Model routing matters. Don't burn a high-reasoning model on an order-status lookup — the cost gap between tiers is roughly 10x. The through-line: technical credibility isn't knowing what's next. It's the ability to build it, well, in record time. I wrote the full thinking up as an editorial — the builder problem, and how a hub-and-satellite model lets business teams innovate without losing control.

I had a conversation recently with someone running marketing technology inside a global business that's scaling fast. I expected the usual worries — budget, attribution, pipeline. Instead they said the riskiest thing they could do was get comfortable with the tool they chose last year. That stuck with me. Because it flips how most teams think about AI. The instinct is to pick the right platform, integrate it, and move on. But the teams actually pulling ahead aren't the ones that picked well once — they're the ones that built a habit of re-picking. Honestly, often, without sentiment. Three things I keep seeing from the ones getting it right: → They treat their stack as provisional — reassessed in months, not years. → They've reopened build-vs-buy on every decision, because a working app can now be built in days (but still has to be maintained for years). → They refuse to let "hours saved" be the ceiling — the harder, more valuable question is where AI becomes a growth lever, not just a cost cut. I wrote the full thinking up as a short piece — what works, where it gets uncomfortable, and why the edge isn't a tool you install once. Link's in the comments. And if this is the kind of thinking you want more of — connect. I share this stuff regularly. What's the tool you got too comfortable with? I'll go first in the replies.

Many AI projects begin with the question, ”Where can we plug this in to save money?” Unfortunately, many of these projects fail for the same reason. The successful ones flip the question. Instead of ”How do we make the current thing cheaper?” they ask, ”What would we build if we started from what’s now possible?” This is the difference between augmenting a business and reimagining one. The former has a ceiling, while the latter doesn’t.

Hear what Jagdev Panesar a Senior Product Owner recently in Decisioning has to share about overcoming cross-department resistance. 1. Who needs to be heard? 2. What strategies to follow? 3. How will they be effective?