2026 AI Benchmark Report
AI adoption is accelerating. AI governance isn't.
Executive summary
AI is now well beyond the experimentation stage in Marketing Operations.
​
Across the organisations represented in this benchmark report, approximately 90% of respondents expect to expand their use of AI materially over the next 12 months. AI is already being used for content creation, reporting, data enrichment, personalisation, segmentation, routing, journey orchestration and lead prioritisation.
​
But there is a problem.
​
The infrastructure surrounding that AI adoption is considerably less mature.
​
Only around 30% of the respondents have an implemented written AI policy specifically for Marketing Operations. With roughly 80% having no clear owner when AI causes harm. 60% don't know how data sent to AI tools is protected. And 40% either don't have, or don't know whether they have, rules governing the retention of AI prompts and outputs containing business or customer data.
​
Meanwhile, data quality is emerging as the biggest obstacle to scaling AI safely.
​
This creates an uncomfortable paradox.
​
Marketing teams are trying to put more intelligence into their operations while many are still struggling to establish the data quality, documentation, ownership and controls that intelligence depends upon.
​
The lesson from this benchmark isn't that Marketing Operations should slow down on AI.
​
Quite the opposite.
​
The organisations that want to move fastest will need to build the operational foundations that allow AI to move safely.
The benchmark at a glance
The benchmark is based on completed responses from Marketing Operations professionals and adjacent roles across North America, EMEA and other regions, with submissions being recorded in the first 6 months of 2026.
​
The respondents represent organisations ranging from fewer than 500 employees to more than 20,000, with the largest group sitting in the 1,000–4,999 employee range.
​
The organisations represented span technology and SaaS, media and publishing, healthcare and life sciences, professional services and other industries.
​
This report's value is in identifying patterns in how teams are actually using AI, where they are putting controls around it, and where the operational gaps are appearing.
AI adoption is no longer the question
The most definitive finding in the benchmark is also the simplest.
​
Nine out of ten organisations expect to expand their use of AI materially over the next 12 months.
​
With the remaining 10% expect to reduce or restrict AI use.
​
That tells us something important about the maturity of the conversation.​ AI adoption is no longer an innovation project that Marketing Operations can choose to ignore. It has become part of the operating model.​ The question has shifted from:
​
"How can we use AI?"
​
to:
​
"How do we use AI at scale without losing control of our operations?"
​
That distinction matters.
​
30% of respondents describe AI adoption as standard across Marketing Operations, while another 30% say it is standard in pockets of the team. 30% describe usage as individual or optional, and 10% say AI has become standard across the wider GTM organisation.
​
The direction of travel is clear: AI is moving from an individual productivity tool to operational capability.
AI is spreading across the stack
AI usuage isn't just confined to content creation.
​
Content drafting and editing is the most common application, appearing in roughly 90% of the responses. But teams are also using AI for reporting and insights, data enrichment and cleansing, personalisation, segmentation, routing, journey orchestration and lead scoring.
​
This is significant because the risk profile changes as AI moves closer to operational decision-making.
​
Using AI to draft an email is one thing. Using AI to influence who receives that email, what journey they enter, how they are scored or what data is added to their record is another.
​
The closer AI gets to the operational machinery of Marketing Operations, the more important governance, traceability and QA become.
​
And that is where the benchmark starts to reveal some cracks.
The Governance Gap
Worryingly:
Only 3 in 10 respondents have an implemented written AI policy for Marketing Operations.
​
Half have no policy at all. 10% a draft that hasn't been finalised, while the rest aren’t sure.
​
The bigger issue, however, is accountability.
​
Eight out of ten respondents say they have no clear owner when AI causes harm.
​
That's arguably more concerning than the lack of a policy. A policy tells people what they are supposed to do. An accountable owner determines what happens when something goes wrong.
​
Without that ownership, AI governance can easily become everyone's responsibility and therefore nobody's responsibility.
This benchmark report suggests this is already happening.
​​
Security controls look stronger than operational controls
Security controls look stronger than operational controls
There are some encouraging signs.
​
100% of respondents said they always require vendor DPAs or a security review before AI tools are used.
​
70% of respondents also say they restrict AI usage to an approved list of tools, with the remaining 30% currently implementing that process.
​
70% also say AI-generated assets or AI-driven changes are always QA tested before production.
​
These are relatively mature controls.
​
But look one layer deeper and the picture becomes less reassuring.
​
Only 20% of respondents say data sent to AI tools is automatically redacted or anonymised.
​
Another 20% rely on manual redaction. The remaining 60% either don't know or consider the question not applicable.
​
Similarly, only 20% of respondents have rules covering retention of AI prompts and outputs containing business or customer data. 40% say they don't have those rules and the rest aren't sure.
​
This suggests that organisations are becoming better at controlling which AI tools enter the business than controlling what happens to information once it enters those tools.
​
That is an important distinction.
There are some encouraging signs.
​
100% of respondents said they always require vendor DPAs or a security review before AI tools are used.
​
70% of respondents also say they restrict AI usage to an approved list of tools, with the remaining 30% currently implementing that process.
​
70% also say AI-generated assets or AI-driven changes are always QA tested before production.
​
These are relatively mature controls.
​
But look one layer deeper and the picture becomes less reassuring.
​
Only 20% of respondents say data sent to AI tools is automatically redacted or anonymised.
​
Another 20% rely on manual redaction. The remaining 60% either don't know or consider the question not applicable.
​
Similarly, only 20% of respondents have rules covering retention of AI prompts and outputs containing business or customer data. 40% say they don't have those rules and the rest aren't sure.
​
This suggests that organisations are becoming better at controlling which AI tools enter the business than controlling what happens to information once it enters those tools.
​
That is an important distinction.
The data problem underneath the AI problem
When respondents were asked about the biggest blocker to safe AI scaling, the answer was clear.
​
Data quality was the number one blocker, cited by 40% of respondents.
​
Legal and compliance uncertainty came next at 20%.
​
That finding is reinforced by the operational data.
​
70% of respondents describe key-field completeness across their core records as mixed.
​
With the same percentage also saying duplicate or identity issues affect reporting at least sometimes, with 30% admitting the problem occurs frequently.
​
This creates a fundamental problem.
​
AI can process bad data much faster than a human can.​ It doesn't magically turn incomplete records, inconsistent lifecycle definitions and duplicate identities into trustworthy intelligence.​ It simply gives those problems a faster delivery mechanism.
​
For Marketing Operations teams, this may be one of the most important lessons from the benchmark:
​
AI readiness is increasingly constrained by data readiness.
Speed is winning. Trust is lagging.
60% of respondents identify speed or throughput as the biggest benefit AI has delivered so far.
​
That's hardly surprising. AI is very good at accelerating work that already exists.
​
But only 10% identifies better targeting as the biggest benefit, while another 10% identifies cost reduction, whilst 20% say AI hasn't delivered a significant benefit yet.
​
This creates an interesting gap between productivity and strategic impact.
​
AI is already making Marketing Operations teams faster. This benchmark report however, provides less evidence that it is making them fundamentally better.
​
That may change as adoption moves deeper into operational workflows. But to get there, organisations will need to move beyond using AI as a productivity layer and start treating it as an operational capability that requires design, controls and measurement.
QA is one of the bright spots
There are signs that Marketing Operations teams understand this risk.
​
70% of respondents say AI-driven changes and AI-generated assets are always QA tested before production.
​
50% also say changes to journeys, scoring or routing always require documented approval.
​
40% have formal incident-management processes involving tickets, SLAs and post-mortems.
​
And 40% have robust versioning in place for critical workflows.
​
These are exactly the kinds of controls that become important as AI moves closer to production. However, maturity is inconsistent.
​
40% of respondents only require documented approval sometimes.
​
30% only sometimes QA test AI-generated work.
​
20% don't have rollback/versioning capabilities and another 20% aren't sure.
​
And 50% of respondents aren't sure whether a formal incident-management process exists.
​
The message is clear:
​
The controls exist. The consistency doesn't.
The traceability problem
One of the most revealing questions in the benchmark asked whether teams could trace why an individual received a particular message from their data, through logic, to an asset send.
​
Only 40% of respondents said they could always track from end to end.
​
30% said often, 20% said sometimes, and 10% said rarely.
​
That matters enormously in an AI-enabled environment.
​
If an AI system starts influencing segmentation, scoring, routing or journey decisions, the ability to reconstruct the chain of events becomes increasingly important. Without traceability, organisations may be able to automate a decision without being able to explain it.
​
"the AI did it" is unlikely to be a particularly satisfying answer for a customer, compliance team or executive when something goes wrong.
Documentation remains stubbornly average
Documentation is another weak point.
​
60% of respondents rate the quality of documentation for their key automations at 3 out of 5.
​
Only 30% rate it 4 or 5. 10% rate it just 1.
​
Lifecycle-stage consistency tells a similar story.
​
Only 30% of respondents rate their confidence at 4 out of 5, while 60% rate it 1–3.
​
This matters because AI increases the value of well-documented operations. An AI system can only make sensible decisions within the context it is given. If nobody can clearly explain what a lifecycle stage means, why a journey works the way it does, or which data fields can be trusted, adding AI doesn't solve the ambiguity - it just automates it.
The AI maturity divide
The benchmark's overall scoring provides another useful directional signal.
​
The average benchmark score across the respondents is 56.5 points.
​
Scores range from 34.8 to 75.6, suggesting a meaningful spread in operational maturity within this sample.
​
Half of all the respondents fall into the "AI Explorer" category, 40% into "AI Builder" and 10% into "AI Beginner".
​
Interestingly, the strongest directional relationship appears around governance.
​
Respondents with an implemented written AI policy achieved an average score of 68.7, compared with 54.9 among those without an implemented policy.
​
Likewise, respondents with clearer accountability tended to score higher.
​
This should not be interpreted as proof that having a policy causes higher AI maturity, but it is a compelling signal.
​
The organisations further along the AI maturity curve appear to be doing more than adopting AI tools - they are putting structure around that adoption.
What the benchmark tells us
Taken together, the findings point to five characteristics of the next stage of AI adoption in Marketing Operations.
​
AI adoption will accelerate
​
The overwhelming majority of respondents intend to expand AI use. Waiting for the market to settle before developing an AI operating model is unlikely to be a viable strategy.
​
Governance will become an operational discipline
​
Tool approval and security reviews are already relatively common. The next challenge is deeper governance: accountability, logging, retention, traceability, monitoring, rollback and incident management.
​
Data quality will determine the ceiling
​
Teams can add more AI tools. They cannot add more trustworthy data by buying another AI subscription. Data quality, identity management, lifecycle definitions and field completeness are becoming prerequisites for useful AI.
​
AI will move closer to production decisions
​
The current use cases already extend beyond content creation. As AI begins influencing scoring, segmentation, routing and orchestration, the distinction between "AI tool" and "Marketing Operations infrastructure" will become increasingly blurred.
​
The winning teams will be the ones that can move quickly without losing control
​
The benchmark does not suggest that organisations should slow AI adoption.
It suggests they need to make their operating environment ready for it.
That means building the guardrails before the AI becomes responsible for more of the machinery.
The new AI readiness question
For Marketing Operations leaders, the most useful question may no longer be:
​
"How much AI are we using?"
​
But:
​
"How much of our Marketing Operations can AI safely influence?"
​
Those are very different measures.
​
An organisation using ten AI tools without ownership, data controls or traceability may be less AI mature than an organisation using three tools within a well-governed operating model.
​
AI maturity isn't measured by the number of copilots, agents or subscriptions.
​
It is measured by the organisation's ability to put AI into production, understand what it is doing, detect when it goes wrong, and recover when it does.
​
That is the benchmark that now matters…
What Marketing Operations teams should do next
Based on the findings from this benchmark report, organisations looking to scale AI should focus on five practical areas.
​
Establish ownership.
​
Define who is accountable for AI decisions, incidents and risk. "The business" isn't an owner.
​
Build an AI operating policy.
​
Document which tools can be used, what data can be shared, which use cases require approval and what happens when AI is used in high-risk workflows.
​
Fix the data foundation.
​
Prioritise identity resolution, field completeness, lifecycle definitions and data governance before pushing AI deeper into operational decision-making.
​
Make AI traceable.
​
Where AI influences a customer-facing decision, organisations should be able to understand what data was used, what logic was applied and what happened as a result.
​
Move QA from an activity to a system.
​
Testing, monitoring, rollback, versioning and incident management need to become part of the operating model rather than heroic acts performed when something inevitably breaks.
Conclusion
The AI revolution in Marketing Operations isn't waiting for governance to catch up. It's already happening.
​
Nine in ten organisations in this benchmark plan to increase their use of AI over the next year. The question is whether their operational foundations can support that acceleration.
​
Right now, the answer is mixed.
​
The good news is that teams are already putting important controls in place. Vendor security reviews are widespread. Approved-tool lists are becoming normal. QA is happening. Some organisations have robust versioning and formal incident management.
​
But ownership, data quality, documentation, traceability and deeper governance remain inconsistent.
​
And that creates the central finding of this benchmark:
​
The next competitive advantage in AI won't come from adopting AI faster. It will come from being able to operationalise it safely, intelligently and at scale.
​
You can have AI in your content workflow, your CRM, your reporting and your campaigns and still have fundamental gaps in data quality, governance, documentation and accountability.
​
The organisations that get ahead won't simply be the ones adopting more AI. They'll be the ones building the operational maturity to support it.
​
Because having more AI doesn't necessarily make you more AI mature.
​
Knowing how to use it, govern it, measure it and improve it does.
