Execution is a given. Explanation isn’t. Martech Futurist | August 13, 2026

Your AI spend bought execution. The thing your buyers now pay a premium for is explanation, and almost nobody has funded it.

Wednesday's edition put a number on the run cost: Gartner forecasts inference workload spending passing training spending in 2026, which moves AI from a capital decision to an operating line. This edition follows the money one step further. That operating line buys execution. Four items published between August 11 and 12 show what it does not buy, and what people are now paying a premium to get instead.

Stop reporting adoption rates to your board

Gartner polled 743 audit professionals and found 93% of audit leaders report some level of AI use against 38% who have an AI strategy. The distribution inside that 93% carries the real finding. Sixty percent use AI to draft audit issues, ratings, or reports. Forty-one percent use it to review drafts. Thirty-five percent use it to prepare stakeholder communications. Thirty percent apply it to audit testing. Twelve percent apply it to quality assurance reviews.

Read those five numbers in order and you can see exactly where the capacity landed. The assurance function put AI on the paperwork and left it off the assurance.

James Bourke, Director Analyst in Gartner's Risk and Audit Practice, states the consequence plainly: high adoption rates in audit functions are producing moderate productivity improvements without transforming audit processes or delivering better strategic insights.

Marketing has the same shape. Elizabeth Maxson, Chief Marketing Officer at Contentful, put the marketing version of this on the record when I interviewed her on The Agile Brand podcast: "We have found that 96% of CMOs are prioritizing AI adoption. However, only 65% are actually putting meaningful investment behind it. And I think that's really the gap there… a lot of marketers are still stuck in this experimentation phase. And they need to start thinking about how are they going to deeply embed AI into their workflows."

An adoption percentage tells your board how many people opened the tool. It says nothing about which decisions got better. If you present one number to your CFO this quarter, present the share of your AI spend that runs against a decision you can name and re-derive.

Budget for explanation before you budget for more output

Jamie Finstein documented the buy-side version of this in MarTech on August 12. IAB's 2026 Digital Video Ad Spend report shows that buyers trust the most transparent buying methods, direct I/O, programmatic guaranteed, and self-serve platforms, at 57%. Confidence falls below 50% for private marketplaces and to roughly a third for open exchange buying. Layer continuously shifting model weights onto that ecosystem and the logic a buyer could once partially infer from outcomes stops being inferable at all.

Publishers lose the same visibility from the sell side. They watch inventory get bought, priced, and packaged by AI-driven systems with no line of sight into why demand landed where it did, which makes pricing and fill-rate forecasting a guess.

Finstein's structural point is the one to carry into your next vendor conversation. Advertisers and media owners are placing more value on platforms whose operators can show why an outcome occurred, and that capability now competes with efficiency and performance as a reason to buy.

Here is the math, using a mid-market B2B budget. Take $1.2M in annual programmatic video spend. The 2020 ISBA and PwC supply chain study found roughly 15% of ad spend untraceable to any line item, which puts $180,000 of that budget somewhere you cannot name. Now assume AI decisioning erodes another slice of what you could previously reconstruct from outcome patterns, and call it conservatively 5%, or $60,000. You are at $240,000 of spend you can report but cannot explain, which is 20% of the budget. A measurement and log-analysis capability, whether you build it or buy it, that costs 3% of media spend runs $36,000. It has to recover 15% of that unexplained quarter-million to break even. The Open Measurement SDK, SupplyChain v1.1, and the MRC's auction transparency standards exist precisely so you can clear that bar. Most teams have never run this calculation, which is why the line stays unfunded.

John Durocher, Chief Customer Officer at Calix, framed the governing discipline when I interviewed him on The Agile Brand podcast: "I encourage people to experiment — think about what business problems we want to solve and experiment — but we've actually put governance around it because you can experiment, and if it doesn't get to scale, it's not that useful." Experiments you cannot explain do not scale, because nobody will sign off on repeating them.

This is the Trust and Verification layer of my AI capability framework doing its work under Orchestration. The question there is whether the agent did what it reports. In programmatic, the same question arrives as whether the platform did what it charged you for.

Name who owns the direction, and put a date on it

Two Harvard Business Review pieces published on consecutive days locate the constraint above the tooling.

Morgan Blangeois of Université Clermont Auvergne and Thomas Roulet of Cambridge Judge Business School spent three years running 23 interviews and three deliberative workshops with leadership teams at 11 European IT services firms. They expected employee resistance. They found the brake sat with the leadership team. Executives privately acknowledged that AI requires changes to pricing, staffing, and business models, then abandoned those positions once the group convened. The authors call it AI leadership drift, and their prescription is procedural: ground the discussion in your own data, study how comparable firms are responding, attach every AI initiative to a strategic question with a review date, and assign one person responsibility for keeping the uncomfortable items on the agenda.

That fourth instruction is the one most marketing organizations skip. Somebody has to own the agenda item that nobody wants to discuss.

J.P. Eggers, Sarah Ryan, and Asha Dinesh reach the same place from the product side. Watching an AI-enabled innovation challenge, they found teams generating the most value during problem definition and solution design, where colleagues with different vantage points challenge assumptions before anyone builds. As AI commoditizes execution, their conclusion is that advantage concentrates in deciding what is worth building and why.

There is a line between descriptive and normative work, and this is where it earns its keep. Intelligence tells you what is and what is likely. Deciding what your organization should do about it is a normative act, and it stays with a human who can be held to it. Book 2's second guiding principle says the same thing in one phrase: humans stay accountable for direction. An AI Governance Board exists to convert that accountability into a named seat with a written charter and defined decision rights.

Agile Brand Principle 7 asks for the humility to acknowledge room for improvement. Leadership drift is what happens when a team locates that humility privately and abandons it in the room.

Featured Insights

Audit teams adopted AI at the drafting layer and skipped the testing layer. Gartner | Gartner Survey Finds Audit Teams' AI Use is Common, But Most Teams are Lacking Strategic Adoption and Application | August 11, 2026

A webinar poll of 743 audit professionals conducted in 2026 shows 93% reporting some AI use against 38% with an AI strategy. Use concentrates in reporting: 60% draft issues, ratings, or reports, 41% review drafts, 35% prepare stakeholder communications. Audit testing sits at 30%, quality assurance reviews at 12%. Bourke's warning is that productivity and adoption metrics understate and can obscure AI's effect on audit outcomes and decision-making.

Run this distribution against your own function this week. List your AI use cases and mark each one as production or verification. If the production column is three times longer, you have built the same imbalance, and the numbers your team spends against are the ones at risk.

Leadership teams name the hard AI decision privately and drop it in the room.Harvard Business Review | "Leadership Drift" Is Stalling Your AI Strategy | August 11, 2026

Blangeois and Roulet report on three years of fieldwork across 23 interviews and three leadership workshops at 11 European IT services firms. Executives recognized in private that AI forces changes to pricing, staffing, and business models, then replaced those positions with reassuring narratives during group discussion. Their four countermeasures: use your own data over industry hype, study comparable firms, tie each initiative to a strategic question with a review date, and give one person the job of keeping uncomfortable issues on the leadership agenda.

Assign the agenda-keeper by name at your next leadership meeting and write the review dates into the calendar. Teams that skip both steps let the hard question quietly leave the agenda.

Advertisers are starting to pay for explanation as a product feature. MarTech | Programmatic video needs more visibility into AI-driven decisions | August 12, 2026

Finstein connects IAB's 2026 Digital Video Ad Spend data, where buyer trust reaches only 57% for the most transparent buying methods and about a third for open exchange, to the opacity that continuous model updates introduce. He identifies three durable positions for intermediaries: owning more of the workflow, owning signals competitors cannot replicate, and owning trust and transparency. Retail media networks illustrate the second through deterministic closed-loop transaction data. The third now competes with efficiency as a purchase criterion.‍ ‍

Add one question to every media platform RFP: show me the audit trail for a single decision, end to end. Vendors who can produce it will tell you fast. So will the ones who cannot.

Problem selection is where teams still beat the tooling. Harvard Business Review | AI Makes Building Easy. Choosing What to Build Is Harder. | August 12, 2026

Eggers, Ryan, and Dinesh observed an AI-enabled innovation challenge and found that value concentrated in problem definition and solution design. Fast prototypes served as instruments for learning, making debates concrete and exposing weak assumptions early. Their guidance to managers: protect time for defining problems, use prototypes to sharpen thinking, and assemble teams around complementary perspectives over technical specialties.

Put a fixed block on the calendar for problem definition before your next campaign build, and staff it with people who disagree with each other. Execution capacity is the cheapest input you have now.

Key Takeaways

1. Replace adoption metrics with decision metrics in your reporting. The share of your team using AI predicts nothing about outcomes. The share of AI spend attached to a nameable, re-derivable decision predicts a great deal.

2. Price explanation capacity against your unexplained spend. Run the arithmetic on what portion of your budget you can report but cannot trace, then compare it to what a measurement and audit-trail capability costs. In most media budgets the number clears easily.

3. Give direction a named owner and a review date. Leadership drift persists because no individual holds the uncomfortable agenda item. Assign that seat, put the review dates in the calendar, and ground the discussion in your own numbers.

4. Protect the problem-definition block. Execution has become the commodity input. Advantage concentrates in choosing what to build, which takes calendar time and colleagues who will argue with you.

Closing Note‍ ‍

I studied photography before digital cameras were common, and I spent a lot of hours in a darkroom. What that taught me was traceability. Exposure time, aperture, developer temperature, agitation interval. When a print came out wrong, I could name the variable and change one thing. That discipline is worth more to me now than just about any technique I learned.

Most marketing organizations this month can produce more output than they have ever produced and can explain less of it than they could three years ago. Your AI spend bought the execution. The explanation is a separate purchase, and the people selling it have noticed.