Closing the Proof Gap in AI-Driven Marketing Decisions

Summary

A cluster of 2026 research from Iterable, BCG, Grant Thornton, McKinsey, and PwC points to the same underlying issue across marketing organizations: the barrier to using AI more fully isn't the technology, it's ownership. Most teams have the tools to move faster, personalize more precisely, and act on data in real time. What's missing is a clear answer to who's accountable when a decision goes live. Closing that gap, one named owner per decision, is what separates organizations moving quickly with AI from those stuck re-approving the same choices on a loop.

Key Highlights

• The proof gap is real and widespread. 78% of executives can't confidently say they'd pass an AI governance audit within 90 days (Grant Thornton).

• Approval bottlenecks are structural, not cognitive. 54% of teams need two to three separate approvals to change a live customer journey, and 59% take two to four weeks to act on insights they already have (Iterable).

• Unclear ownership, not excess control, is the top barrier. When marketers were asked why they hesitate to give AI systems more autonomy, unclear ownership when something goes wrong ranked above brand risk, trust in decision-making, or leadership readiness (Iterable).

• Leadership and practitioners are seeing two different companies. 96% of CMOs say AI is driving full transformation of their function. Only about a third have actually built that (BCG).

• Most personalization can't explain itself. Only 2 in 5 marketers can say why a specific customer got a specific message, and 64% admit their personalization is more about optics than impact (Iterable).

• Clear ownership correlates with AI maturity. Organizations with a clearly accountable AI owner score meaningfully higher on maturity, 2.6 versus 1.8 (McKinsey).

• The fix pays off in revenue, not just speed. Fully integrated AI organizations are almost four times more likely to report revenue growth than those still piloting, 58% versus 15% (Grant Thornton).

• The fix is achievable without new bureaucracy. Naming one person per recurring decision type, not a governance department or an audit process, is enough to start closing the gap, even for mid-market teams.

---

Where Marketing is Heading

Marketing is heading toward a version of itself where AI systems run constant experimentation and adjust channel, timing, and frequency on their own, without a review cycle. That's not a distant forecast. One 2026 survey of marketers describes it as already taking shape. It's also exactly the kind of system where an unnamed decision gets expensive fast, because there's no human in the loop to catch it before it compounds across thousands of customers.

The organizations getting comfortable handing more autonomy to these systems aren't the ones with the most advanced AI. They're the ones who did the ownership work first. Naming who's accountable for a decision isn't a brake on moving toward more automation. It's what makes moving toward it safe enough to actually do.

And most teams haven't done that work yet, even with a human still fully in the loop. More than 70% of marketers won't touch a live program because they can't predict what happens downstream. Not because the idea's bad. Not because the data's unclear.

Because if it goes sideways, nobody's sure whose name is on it. That's from Iterable's 2026 Customer Engagement Report, and it lines up with what a few other firms found this year when they went looking at the same problem from different angles.

Marketing teams have more tools than they've ever had. Faster reporting, better segmentation, AI that generates and tests variants in minutes instead of weeks. Most of that capability sits unused, not because it doesn't work, but because nobody's decided who's allowed to say yes.

Grant Thornton asked business executives a blunter version of the same question this spring: could you pass an independent AI governance audit in the next 90 days? 78% said no, not with confidence. They called it the proof gap. Organizations can't show how a decision got made or who owns what happens next. And the ones who've closed the AI proof gap aren't moving slower because of the extra structure. They're moving faster, because they're not stopping to ask "wait, who approved this" every time something needs to change.

78% of business executives lack confidence they could pass an independent AI governance audit within 90 days. — Grant Thornton

One Owner, One Decision to Close the AI Proof Gap

The numbers on this are pretty specific. 54% of teams need two to three separate approvals just to touch a customer journey. 59% say it takes two to four weeks to act on something they already know from their own data. It's tempting to read that as "too many meetings," but that's not really it. The thinking isn't slow. The ownership was never assigned, so every decision defaults to committee because committee is what's left when nobody owns the call.

54% of teams need two to three separate approvals to make a single change to a live customer journey, and 59% say it takes two to four weeks to act on insights they already have. — Iterable

PwC found something similar this year but from the inside of governance programs specifically: the organizations further along have stopped routing everything through a shared committee and started splitting up who's actually responsible for what. Committees are fine for the early stuff, deciding scope, setting risk appetite. Past that they turn into the bottleneck.

You'll recognize this from the buyer's side too, if you sell into mid-market or enterprise. A prospect who needs sign-off from three stakeholders before approving anything isn't being difficult. They're stuck in the same structure. Somebody in the building has the authority to decide. Nobody's been handed it clearly enough to use it. The RFP that resets pricing instead of actually going anywhere, the "let me loop in my partner," the quarter that quietly becomes two quarters. Different department, same mechanism.

Clear Authority to Decide

A lot of teams reach for words like bureaucracy or command and control to explain why decisions stall. The data points somewhere a little different. The top reason marketers give for not handing AI systems more autonomy is unclear ownership when something goes wrong, cited more than any other reason. That's less about too much authority sitting at the top, and more about nobody specific holding it.

Unclear ownership when something goes wrong is the top barrier marketers cite for giving AI systems more autonomy, ahead of brand risk, trust in decision-making, or leadership readiness. — Iterable

Somewhere in the move toward flatter teams and shared decision-making, a lot of organizations moved past the old model, where one person just decided, without building anything to take its place. What's left often defaults to committee, since that's the structure available once nobody's been handed clear ownership. Research on governance programs backs this up: the organizations furthest along didn't add more layers of sign-off.

They moved away from shared committees and handed individual decisions to named owners instead.

The fix, then, is less about cutting through structure and more about restoring one piece of it: a name attached to each decision.

Ownership Without New Bureaucracy

Most of the research on this problem comes out of large enterprises with dedicated risk teams, which can make it sound like the fix is out of reach for anyone smaller. It isn't. A mid-market business doesn't need a chief AI officer or a formal audit process to close this gap. It needs one person per decision type who can say yes or no without checking with three other people first.

That might mean the head of growth owns anything touching pricing or offers, while a marketing ops lead owns send timing and channel selection. The split matters less than the clarity. What kills momentum in a smaller org isn't the absence of a governance framework. It's the absence of a name. The moment someone can point to who owns a call, the two to four week lag starts shrinking, because the question stops being "who do I ask" and starts being "did I check with the person who owns this."

Automation in Practice

BCG surveyed 300 CMOs this year and found 96% of them said AI is driving full transformation of their function. Then they checked. About a third had actually built it. The rest were still assisting humans on individual tasks, one at a time.

96% of CMOs say AI is driving end-to-end transformation of their function. About a third have actually done the work. — BCG

That same research found close to the same split from a different direction. Marketing leaders were 36% more likely than the people doing the work to trust AI analytics as much as a human analyst, and over three times more likely to believe AI was already running execution on its own, when in reality two-thirds of that work is still AI-assisted with a human signing off on every step.

That's not really a trust issue. It's that leadership and the team closest to the work are looking at two different pictures of the same system, and neither one is wrong from where they're standing. The gap between what's claimed and what's built is, per BCG, kind of the whole story of this year.

Explainable Decisions

Only two in five marketers can say why a specific customer got a specific message. Nearly two-thirds admit the personalization they're shipping is built more for how it looks than what it does.

Only 2 in 5 marketers can explain why a specific customer received a specific message. 64% say their personalization is more about optics than impact. — Iterable

That's the actual cost of skipping the ownership question. It's not that nothing gets shipped. Plenty ships. It's that nobody can trace any of it back to a reason, which means nobody can defend it, fix it, or hand it to someone else when they leave. McKinsey found the same pattern at the org level this year: companies with a clearly accountable owner for their AI programs score meaningfully higher on maturity than the ones without one, 2.6 versus 1.8 on their scale. Ownership isn't a nice-to-have layered on top. It's the thing that makes the rest of it usable.

Closing the Gap

More oversight isn't the fix. More oversight is usually what caused the freeze in the first place. What closes the gap is putting one person's name on the outcome of a given decision ahead of time, then letting the system act inside the boundaries that person actually sets.

If you're building the system, that means deciding what success looks like and who's accountable before it goes live, not after something breaks. A blast radius you drew on purpose is a lot easier to stand behind than one you discover by accident.

If you're buying from a vendor, ask who owns the outcome on their end, not just who owns the relationship. The ones who can answer that in a sentence have usually already done the work internally, which tells you something about how they'll handle it when it's your account.

And if you're the one reporting up, the space between what leadership thinks is running on its own and what's actually still hand-approved is worth closing now, while it's a conversation instead of an incident.

Here's where to start, whatever seat you're in: pick the three decisions that get remade most often inside your customer journeys, whether that's discount thresholds, send frequency, or channel selection, and write down who owns each one today. If the honest answer is "it depends" or "a few of us," that's the gap. Naming it is most of the work.

None of this is waiting on better technology. Grant Thornton's data makes that pretty clear: fully integrated organizations are almost four times more likely to report revenue growth than the ones still piloting, and the difference wasn't the tools. What's missing usually isn't capability. It's a name next to the decision. The businesses closing that gap now are the ones who'll be ready to hand more of it to the system later, on purpose, instead of discovering how much they'd already handed over after something goes wrong.

Organizations with fully integrated AI are almost four times more likely to report revenue growth than those still piloting, 58% versus 15%. — Grant Thornton

---

Frequently Asked Questions:

What is the "proof gap" in AI marketing?

It's a term from Grant Thornton's 2026 AI Impact Survey. It describes the moment when a business can't show how an AI-driven decision got made or who's accountable for what happens next. In that survey, 78% of executives said they couldn't confidently pass an independent AI governance audit within 90 days. The gap isn't in the technology. It's in the paper trail behind the decision.

Why do marketing teams hesitate to change live AI-driven programs?

Mostly because ownership was never assigned. In one 2026 survey of marketers, more than 70% said they avoid modifying live programs because they can't predict the downstream effects, and unclear ownership when something goes wrong was the single most cited reason. It's not that the idea is bad or the data is unclear. It's that nobody's sure whose name is on the outcome.

Is this a technology problem or an organizational one?

Organizational. The same research found 54% of teams need two to three separate approvals just to change a customer journey, and 59% say it takes two to four weeks to act on insights they already have. That's not a limitation of the AI tools. It's a structure where every decision defaults to committee because nobody's been handed clear authority to make the call alone.

Does this only affect large enterprises with dedicated governance teams?

No. Most of the published research comes out of large organizations, which can make the fix sound out of reach for smaller teams. It isn't. A mid-market business doesn't need a formal audit process or a chief AI officer to close this gap. It needs one named person per decision type, someone who can approve a change without looping in three other people first.

Is this just bureaucracy slowing things down?

Not exactly. It looks like bureaucracy from the outside, lots of sign-offs, long timelines, but the underlying cause is closer to the opposite. A lot of organizations moved away from a single decision-maker toward flatter, more shared decision-making, without building a clear ownership structure to replace it. The committee isn't holding too much power. It's what's left standing when nobody specific has been given the authority to decide.

How big is the gap between what leadership believes and what's actually happening?

Significant. BCG surveyed 300 CMOs in 2026 and found 96% said AI was driving full transformation of their marketing function. When they checked, only about a third had actually built that. The rest were still using AI to assist individual tasks, with a human approving each step. Leaders were also 36% more likely than practitioners to trust AI analytics at the same level as a human analyst, and over three times more likely to believe execution was already autonomous.

What does "explainable" AI decision-making actually mean in practice?

It means someone in the organization can say why a specific customer received a specific message, not just that the message went out. Only two in five marketers in the 2026 Iterable survey said they could explain that reasoning, and 64% admitted their personalization is built more for how it looks than what it actually does. McKinsey found the same pattern organizationally: companies with a clearly accountable AI owner scored meaningfully higher on maturity, 2.6 versus 1.8 on their scale, than those without one.

Does closing the ownership gap actually affect revenue?

Yes, according to Grant Thornton's data. Organizations with fully integrated AI were almost four times more likely to report revenue growth than those still in pilot mode, 58% versus 15%. The difference wasn't better tools. It was clarity about who owns the decision to use them.

Where should a business start if it wants to close this gap?

Pick the three decisions that get remade most often inside your customer journeys, things like discount thresholds, send frequency, or channel selection, and write down who owns each one today. If the honest answer is "it depends" or "a few of us," that's the gap. Naming a single owner for each of those three decisions is most of the work.

---

Sources:

• Iterable, 2026 Customer Engagement Report (survey of 1,084 consumers and 505 marketing professionals, fielded March 2026 by Propeller Insights).

• PwC, "PwC's 2025 Responsible AI Survey: From Policy to Practice."

• BCG, "Moving the Agentic Marketing Transformation from Illusion to Reality" (2026 global survey of 300 CMOs).

• Grant Thornton, 2026 AI Impact Survey. McKinsey, State of AI Trust in 2026. PwC, 2025 Responsible AI Survey.