Monthly Digital Insights – October 2026

This month, three themes kept resurfacing across industry discussions, conferences and vendor announcements: AI decisioning, customer data unification and privacy-resilient measurement.

Give your AI models better confidence scores

Most organisations experimenting with AI are focused on selecting the best model, crafting better prompts or reducing costs. The new Jev model from TypeSafe AI raises a new question: when should AI make a decision on its own, and when should it ask for help?

Jev is an AI model developed by TypeSafe AI that makes fast choices instead of writing sentences or paragraphs. This new model opens the door to an interesting concept: structured AI outputs combined with confidence scoring.

Instead of sending every decision to a large language model, organisations could route straightforward decisions automatically and escalate uncertain cases for human review or a more advanced AI model.

Consider customer churn prevention. A model might predict that a customer is likely to leave, but how confident is it in that prediction? By setting confidence thresholds, you can automatically trigger retention campaigns for high-confidence predictions while referring uncertain cases to a more advanced AI model or a human expert.

While some organisations already use predictive scores, confidence-driven AI introduces an additional governance layer. According to TypeSafe AI, Jev offers much lower costs and higher speed while improving control and explainability.

Why this matters

Confidence-based decision-making could help you reduce processing costs, speed up decisions and limit the risk of incorrect actions. But confidence scores alone aren't enough. You also need to assess how reliable those scores are and monitor business outcomes, false positives and escalation rates.

In summary, don't just ask whether AI is correct. Instead, ask whether AI knows when it might be wrong.

Is Google Analytics 4 slowly becoming a CDP?

Recent developments presented at the Digital Analytics Summit suggest that Google Analytics 4 (GA4) is gradually moving beyond traditional web analytics and towards broader customer data capabilities.

Yes, GA4 is increasingly supporting capabilities that were traditionally associated with Customer Data Platforms (CDPs):

• Audience activation
• Data unification with non-Google data sources
• Business data enrichment
• AI-powered insights with Ask Advisor
• Cross-channel analysis

For many organisations, especially those early in their customer data maturity journey, this may be enough.

However, specialised CDPs still provide capabilities that go beyond what we've seen in GA4:

• Advanced identity resolution
• Customer profile governance
• Multi-channel activation
• Consent orchestration
• Enterprise-grade profile management

For mature organisations already operating platforms such as Tealium or Hightouch, the incremental value can be limited.

Why this matters

Before investing in a new CDP, it's worth assessing what you can already achieve with GA4. Depending on your customer data maturity and business needs, you might be able to get more value from your existing setup.

The key is understanding which customer data capabilities you actually need and where your current technology falls short.

Tracking in 2026: resilient measurement without losing trust

Simo Ahava's presentation at the Digital Analytics Summit was a real eye-opener for data professionals: are browser restrictions, privacy regulations and ad blockers preventing organisations from collecting valuable insights, or are they protecting users' anonymity and interests?

The future of measurement is not a choice between privacy and analytics. Instead, it focuses on building systems that achieve:

• Reliable measurement
• Responsible data usage
• Clear consent choices
• Transparent customer communication

This is a great reminder that server-side implementations are not a miracle solution and still need to account for consent requirements and transparent data practices.

Why this matters

Pardon my French, but I'd like to quote Simo's rule: "Don't be a dick."

Underneath the joke lies an important principle: just because technology allows us to collect, connect, enrich and activate more data doesn't mean we always should.

Organisations that focus solely on maximising data collection risk losing exactly what they need most: customer trust.

Key takeaway

The common thread across all three topics is trust:

• Trust that AI decisions are reliable enough to automate.
• Trust that customer profiles are accurate enough to activate.
• Trust that measurement remains effective while respecting privacy.

As analytics continues to evolve into decisioning, that trust may become the most valuable asset in your entire data stack.

Want to keep the conversation going?

On 12 November, we are hosting the DDMA Digital Analytics Meetup | From Dashboard to Dialogue.

This is a free event in Amsterdam for data specialists who are curious about how AI and conversational interfaces are changing the way organisations interact with data.

Receive data insights, practical cases and a look behind the scenes once a month?

Sign up to our mailing list and stay 'up to data':