100% of agencies use AI. Only 9% have turned it into an edge. Get the ground reality on AI, pricing and demand from 55 CRO and experimentation agencies.
Research by
Craig Sullivan
13+ years of experimentation, UX research and analytics work for Google, Amazon, Spotify, Lego and Red Bull. 500+ clients.
Co-produced with
Sponsored by
Free. The PDF opens right away and a copy lands in your inbox.
Loading the download form.
What we found in 2026
9%
have embedded AI in delivery deeply enough to gain an edge. Adoption is 100%.
50%
believe AI is commoditizing what they sell and that they hold a human advantage it can't replicate.
81%
raised or held their rates this year. Just 15% cut them.
39%
name winning new business as their top risk. AI commoditizing delivery sits at 24%.
Sneak peek
We tested five measures of AI intensity against price changes. None tracked price. Of the 25 agencies that raised prices, 72% did it by moving to higher-value work.
AI intensity vs price direction
Kendall τ-b with price direction, not significant. Source: Agency & Consultancy Research 2026.
2025 vs 2026
Five questions appear in both surveys.
What the 2025 report got right in its predictions
Net agree, 2025 vs 2026
Percentage-point gaps, no significance testing. Source: Agency & Consultancy Research 2025 and 2026.
Free. Opens right away, plus a copy in your inbox.
What's inside
01
Everyone has adopted AI. Only 9% run it deep enough in delivery to see real gains. Where the other 91% stall on the way from adoption to advantage.
02
Why agencies feel commoditized and defensible at once, and how it plays out in price.
03
The risks agencies actually rank, and why they sit hardest on smaller, client-dependent shops.
04
Four moves separating the agencies pulling ahead in 2026.
05
The strongest relationship in the study. Hint: it isn't AI.
Agency leaders on in-housing, pricing, AI in delivery and where they are placing their bets, in their own words.
Free with the report
Each sticker is a finding from the report, designed as a print-ready file you can print yourself. Tick the box when you download the report and we'll send you the pack.
Pre-launch reviews
“Brilliant report. Clarifies the joint tensions we're seeing at Launch between racing to embed AI, generate pipeline demand, also prevent client churn. … Really well written, concise, backs up every claim …”
“This report is a genuinely useful benchmark for anyone leading a CRO function. … Seeing year on year changes adds a new level of insights we couldn't get from any other report.”
“Amid the flood of inflated AI claims we encounter every day, this report feels like a breath of fresh air.”
“This report helped me make sense of … the gap between how confidently we talk about using AI and how we're actually feeling about what it means for our careers.”
“A refreshing read in a world where apparently every second LinkedIn post has already replaced us with an agent.”
“For agency and CRO leaders, it's a compelling read that helps you question your business model and identify where you can create lasting value.”
About the author
Craig has spent over 13 years blending UX research, analytics and A/B testing for companies like Google, Amazon, Spotify, Lego and Red Bull. He has helped over 500 clients unlock growth inside their products and services.
Co-produced with
Sponsored by
Who is this report for?
Founders and leaders at CRO, experimentation and full-service agencies and consultancies. If you sell experimentation to clients, it’s for you.
Is it free?
Yes. Fill in the form and the PDF opens right away. We also email you a copy.
How was the research done?
A survey of 55 CRO and experimentation agencies and consultancies, run February to May 2026, plus conversations with agency founders and leaders. It’s a self-selected sample, so findings reflect this community rather than every agency. Relationships are measured with Kendall’s τ-b and bootstrapped 95% confidence intervals. The full methodology is on page 39.
How does it compare to the 2025 report?
Five questions appear in both surveys, and the report compares them. The samples differ, so the comparison shows direction, not a trend. Read the 2025 report.
Did AI write this report?
No. Craig did. Claude helped clean the survey data, scaffold the analysis, code open-text answers and handle personal data, with human review. The analysis and the writing are Craig’s.
Who can get the sticker pack?
Anyone who downloads the report can request the five print-ready sticker designs on the next screen. These are digital files, not physical stickers.
The floor is rising for everyone. The edge belongs to whoever decides, deliberately, what to do with what AI hands back.