The marketing AI budget nobody can audit, and how AI changes that

Why most marketing AI stalls between speed and control, and what it takes to give CMOs, CFOs and ecommerce teams an answer they can all trust.

In short:

• Most companies are spending more on AI but few are getting value from it. Deloitte’s 2026 research calls this the gap between access and activation.

• Marketing is where this stalls most, because the data sits in GA4, Google Ads, Meta and the store, and those sources rarely agree.

• The CMO wants speed, the CFO wants proof, and the ecommerce manager is judged on both at once.

• IDIRA.chat closes the gap with one design choice: the same cross-platform check that makes an answer fast for the marketer makes it auditable for finance.

In most companies, two people look at the marketing budget and see two different problems.

The Chief Marketing Officer sees a question of speed. Money moves every day across Google Ads, Meta and other channels, and the decisions cannot wait for next month’s report.

The Chief Financial Officer sees a question of proof. Marketing is often the one large line of spend that nobody can fully audit. The numbers come from different tools, they do not always agree, and “trust me, it is working” is not a sentence a finance leader likes to hear.

And in many businesses, especially in ecommerce, one person feels both pressures at the same time. The Ecommerce Manager is judged on growth and on profit at once, and lives in the daily numbers where the two collide: profit on ad spend, customer acquisition cost, and lifetime value, all moving as budget goes out across the store and the ad platforms.

Most marketing technology asks these people to choose. You can have speed, or you can have control. This article is about why that trade-off still exists, why artificial intelligence has not yet removed it for most companies, and what it takes to give all three an answer they can rely on.

Why does so much marketing AI fail to deliver value?

Because most organisations have reached AI tools without turning them into results. Deloitte calls this the gap between access and activation.

Deloitte’s State of AI in the Enterprise 2026 report surveyed more than 3,200 business and technology leaders. On the surface, the news is good: 84% of organisations increased their AI investment, and 78% of leaders report greater confidence in the technology.

The detail is more sobering. Only about one in five organisations is currently growing revenue from AI, while roughly three in four still only hope to. The report also names a familiar pattern it calls the proof-of-concept trap: a pilot runs well in a controlled setting, then never makes it into daily use. Teams keep funding new experiments because they are cheap and low risk, rather than doing the harder work of scaling the ones that already work.

For a CFO, this is the real worry about AI. Not that it will not work, but that it will quietly absorb budget without ever changing a number on the page.

Why is marketing the hardest place to get value from AI?

Because the data lives in too many places and rarely agrees with itself.

GA4 tells one story. Google Ads tells another. Meta reports its own version. A spreadsheet the team maintains by hand is already out of date by the time anyone opens it. So the marketing leader spends half the week pulling and reconciling data instead of acting on it, and still cannot prove exactly where the budget is being wasted.

By the time a clean monthly report arrives, the campaign has run, and the money is spent. The moment to act has passed. This is the daily reality behind the CMO’s demand for speed. It is not impatience. It is the cost of waiting.

What does a CMO actually need from marketing AI?

A plain question answered in plain language, in seconds, from the tools the team already uses.

The marketing leader does not want another dashboard to configure, another login, or a months-long implementation. They want to ask “What is driving our best leads this month?” or “Why did conversions drop last week?” and get a clear answer straight away. No SQL. No data team in the loop for every question. No new workflow to maintain.

What does a CFO need before approving marketing AI?

A return they can justify, an answer they can audit, compliance they can demonstrate, and confidence the system will not invent numbers.

The Deloitte data shows these are the right questions to ask. When leaders were asked which AI risks worry them most, the top answers were all about trust and control: data privacy and security (73%), legal and regulatory compliance (50%), governance and oversight (46%), and model quality and explainability (46%).

There is a second theme that finance leaders increasingly raise: where the AI is built. Deloitte found that 77% of companies now factor an AI solution’s country of origin into their buying decisions. Across Europe, the Middle East and Africa, around a third of organisations are uneasy about depending on foreign-owned AI for the core of their stack. For a European business, this is no longer a technical footnote. It is a board-level question about control and resilience.

What about the Ecommerce Manager caught in the middle?

In an ecommerce business, the person who feels both pressures first is rarely the CMO or the CFO. It is the Ecommerce Manager.

They are measured on growth and on profit at the same time. Profit on ad spend, customer acquisition cost and lifetime value are not abstract metrics to them. They are the score, updated every day as budget converts or does not.

Their hardest problem is that the numbers come from different places and rarely agree. Meta reports one set of conversions. Google Ads reports another. GA4 shows a third. The store tells its own story. Deciding where to push the next thousand euros means trusting figures that quietly contradict each other.

This is the person who needs speed and control in the same answer, not as a trade-off. IDIRA.chat checks that the store, Google Ads, Meta and GA4 agree before it tells them anything, so the daily call on spend rests on numbers that line up.

How can one tool give speed and control at the same time?

Because the feature that makes an answer fast for the marketer is the same feature that makes it auditable for finance.

IDIRA.chat checks that Google Ads, Meta and GA4 tell the same story before it gives you an insight. When the platforms disagree, you see it first. When data is missing, it tells you what is missing rather than filling the gap with a guess. The assistant does not answer without real data from your connected platforms, so there are no invented figures.

For the CMO, that means an answer they can act on without re-checking it by hand. For the CFO, the very same validation produces something rare: a marketing answer that is traceable back to the data behind it. Every reply records which model produced it, when, and from which data, in a form ready for a security or finance team to review. The marketing budget stops being the one line of spend nobody can check.

On compliance, IDIRA treats the European requirements as one package rather than separate boxes to tick. Your data stays on infrastructure you control, with EU data residency. The platform is built to align with GDPR and NIS2 and is making ready for DORA. And it already meets the EU AI Act’s Article 50 transparency duties, ahead of the August 2026 enforcement deadline, so every answer identifies the AI and every action that changes data asks for human confirmation first.

Separately, and this matters to anyone who has experienced a single vendor outage, IDIRA.chat runs several AI models in a chain. If one provider fails, the next takes over automatically, so a single point of failure never stops the team mid-decision.

How do you move from AI access to AI activation?

By choosing tools built to run in daily work, not pilots you hope to scale one day.

Recall Deloitte’s central finding: the winners are not the companies with the most AI pilots, but the ones that move AI into the everyday way they work.

IDIRA.chat is built for that move. It works with the platforms you already use, with no migration and no new system to learn, and it answers a real question about your own data on the first day. It is not a general-purpose chatbot you have to coax, and not a business-intelligence tool built for data engineers. It is purpose-built for marketing decision-makers, pre-connected, validated, and structured by design.

That is what closes the gap between the people who decide how marketing money is spent. The marketer gets speed. The finance leader gets control. The ecommerce manager gets both in the one answer they check every day. And for once, they are all looking at the same one.

Frequently asked questions

What is IDIRA.chat? IDIRA.chat is a conversational marketing analytics platform for marketing decision-makers. It connects GA4, Google Ads and Meta Ads, checks that the platforms agree, and answers questions in plain language in seconds. When data is missing, it says so, so it never invents figures.

How does IDIRA.chat stop AI from inventing numbers? It cannot answer without real data from your connected platforms. It also checks that Google Ads, Meta and GA4 agree before giving an insight, and shows the actual date range of the data it used. Every answer is traceable back to its source.

Who is IDIRA.chat for? It is built for marketing decision-makers: CMOs and marketing directors who need speed, CFOs who need to audit marketing spend, and ecommerce managers tracking profit on ad spend, acquisition cost and lifetime value across the store and the ad platforms.

Is IDIRA.chat compliant with the EU AI Act? Yes. IDIRA.chat already meets the EU AI Act Article 50 transparency duties, ahead of the August 2026 enforcement deadline. It is also GDPR and NIS2 aligned and DORA-ready, with EU data residency and your data kept on infrastructure you control.

How is IDIRA.chat different from Power BI Copilot or ChatGPT? Business-intelligence copilots are built for data engineers and need semantic models. Generic AI works from live APIs and can assume beyond the data. IDIRA.chat is purpose-built for marketing, pre-connected, and validates across platforms before it answers.