Dellon Stefanus , Head of Digital & Web Builder

TL;DR

• AI now accounts for 9% of marketing budgets, up from 7% in 2024
• Yet 67% of CMOs can't quantify ROI from AI investments
• The gap: adoption is racing ahead of accountability infrastructure
• This creates a dangerous window where bad AI investments look good

Marketers are moving fast. AI budgets jumped 29% in the last 18 months. Brands are hiring AI strategists, spinning up generative workflows, optimizing personalization stacks. The enthusiasm is real. The problem is measurement is broken.

AI Budget Share 2026

67%

CMOs Can't Measure ROI

$2.8B

AI Marketing Market 2026

340%

Budget Growth Since 2022

The Adoption Illusion

Speed Doesn't Equal Results

Everyone is moving to AI. 73% of marketers plan AI adoption in 2026. Brands are investing in LLM APIs, spinning up generative workflows, stitching AI into demand gen, personalization, customer service, content. The momentum feels inevitable. But momentum isn't ROI.

The trap is structural. Budgets are rising because AI adoption is rising, not because AI is proven to work. Brands throw money at tools because competitors are doing it, because vendors are pushing it, because the market narrative says it's required. That's fear-driven spending, not performance-driven spending.

Budget allocation across marketing channels — AI budgets are growing fastest, but measurement infrastructure hasn't kept pace

Why Measurement Fails

Attribution Is Already Dead

The real problem: attribution is fundamentally broken. When your campaign touches 12 channels and a customer journey spans 47 touchpoints, who gets credit for the conversion? When AI personalizes offers based on behavioral data, when retargeting fires across networks, when organic search co-mingles with paid - traditional attribution collapses.

Here's the game: If you can't measure it, you can't be held accountable for it. So AI adoption continues in this measurement void, looking good on spreadsheets while outcomes stay murky.

The Three Measurement Gaps

1. Multi-Touch Attribution is Fantasy

Linear, first-touch, last-touch, algorithmic - pick one. None of them are right. AI can't fix a broken model.

2. Incrementality is the Real Test, Rarely Run

Only 19% of brands run lift tests. Yet those are the only honest way to know if AI drove incremental sales or just accelerated orders that would've happened anyway.

3. Time-to-Impact is Hidden

AI personalization takes 3-6 months to generate measurable lift. Most CMOs evaluate quarterly. By then, they've already decided based on activity metrics, not outcomes.

CMO measurement confidence gaps — 67% of CMOs lack confidence in AI ROI measurement - a structural problem, not a tool problem

What Happens Next

The Reckoning Is Coming

By Q4 2026, boards will start asking harder questions. "We spent $8M on AI. Show me the incremental revenue." CMOs won't have a clean answer. Some will point to activity metrics (more personalized emails sent, faster content production). Some will retreat to vanity metrics (engagement lift in the test cohort). Some will quietly kill programs that showed no lift.

The survivors won't be the brands that adopted AI fastest. They'll be the ones that took time to build measurement infrastructure first. That's unsexy. It doesn't generate conference talks. But it's the only way to know if you're actually winning.

The Real Gap

You can't manage what you don't measure. Right now, marketing teams are managing AI adoption, not AI outcomes. That's a 12-18 month lag that will correct itself - painfully.

The 5-Move Fix

Move 1: Run Incrementality Tests First

Before scaling AI personalization, run a 4-week holdout test. Measure incremental lift. If it's 3% or less, don't scale. If it's 8%+, now you have a number to defend.

Move 2: Build Multi-Touch Attribution (Even If Imperfect)

Pick a framework. Could be algorithmic, time-decay, or Bayesian. Own the imperfection, document it, and be consistent. Imperfect measurement beats no measurement.

Move 3: Audit Your Baseline

What did sales/revenue look like before AI? Without a baseline, you can't measure lift. Most teams skip this. Don't.

Move 4: Extend Your Measurement Window

AI personalization is slow. Give it 6 months before you evaluate. If your board needs quarterly wins, run concurrent quick-hit campaigns alongside long-play AI bets.

Move 5: Separate Activity from Outcome Metrics

More emails sent ≠ revenue. Track both. But measure success on revenue, not activity. Activity metrics will always make AI look good. Outcome metrics won't lie.