The distance between perceiving and deciding. Change Futurist | September 14, 2026

Ask a marketing leader to define judgment and you will get adjectives. Curious. Commercial. Strategic. Adjectives survive a conference panel. They do not survive a performance review, and they do not survive a CFO.

Four publications inside four days each supply a piece of the definition. Adrian Leow at Gartner put a boundary on what the machine covers. Five senior marketers writing in Marketing Week named the behaviors they hire for. Media agencies, reported through MarketingProfs, attached a human name to every agent-produced deliverable. And Jonathan Knowles and Chris Burggraeve explained why marketing keeps getting graded on the wrong instrument no matter how good the judgment is.

The August 31 edition argued that AI programs need a named owner. September 3 gave that owner a measurement instrument. September 7 added the cost denominator and the routing rule. September 9 found the manager layer applying that rule trained on channels. This one writes the job description for the thing that layer is missing.

I have sat in calibration meetings where two directors rated the same person four points apart, and neither could name what they were rating.

Pricing the distance

Leow, VP Analyst at Gartner, presented a strategic guide to AI agent deployment models on the opening day of IT Symposium/Xpo APAC on the Gold Coast, September 14. Three of his takeaways matter for marketing planning.

He described current LLM-based AI agents as efficient at perceiving information and short on adaptability and contextualized decision making. He called out agent washing, where vendors rebrand existing capabilities without true agentic functionality. And he gave IT leaders a sequence: prioritize high-frequency, low-complexity use cases, apply guardrails, upskill the workforce.

Read the middle finding as a specification. It describes a division of labor. In the capability framework I use, Insights produces understanding and Goals sit above the pillars as the aim a human brings to the system. Leow draws the same line from the vendor side. The machine covers perception. Your people own the contextual decision that perception feeds.

That line is the job. Once you can draw it on a specific workflow, you can staff it, train for it, and grade it.

Most marketing organizations have never drawn it. They have an org chart and a tool list, and the space between the two absorbs every decision nobody assigned. The wiki entry on decision rights covers the mechanics of doing this on paper before the agents force the question.

Write down the behaviors your senior marketers already named

Marketing Week published responses from Festival of Marketing speakers on September 14, asking what approaches set marketers up to succeed. The answers converge on the same layer Leow leaves to humans.

Lauren Berkemeyer, CMO of YuLife, advises her team to understand how the whole machine works in a business and to stop waiting to be asked. Georgina Bramall, CMO at Giffgaff, puts the emphasis on unlearning as you go and relearning what works in today’s market. Beth O’Malley, CRM and email specialist and founder of astral, tells marketers to learn to sell and learn how the brain works, putting psychology ahead of instinct. Stefanie Nastou, CMO of AMS, describes the combination she has used to build trust with executive teams: define a clear measurable goal, break it into steps, and stay accountable when things go wrong. Fractional CMO Holly Rix gives the sharpest single test. Ask whether another brand could run the work you are about to publish. If yes, go again.

Five behaviors, and every one of them converts into something you can observe. That is what makes them useful. Curiosity is an adjective. Asking a second question about the P&L before you approve a brief is an event with a date on it.

Book 2 splits this cleanly. Intelligence tells you what is and what is likely. Deciding what the organization should do is normative work, and normative work stays with humans even when the analysis runs itself. Rix’s test is normative work compressed into one sentence: this output is legal, on brand, and interchangeable, so we will not ship it.

It reminds me of what Sid Banerjee, Chief Strategy Officer at Medallia, said when I interviewed him on The Agile Brand podcast: “If I’m a store manager, I’m not going to look at a dashboard to figure out how I’m doing. I’m going to ask my CX assistant, what do I need to focus on today?” The assistant answers the descriptive question. The manager still decides which of the three things surfaced actually gets the shift’s attention, and that decision carries the cost of being wrong.

Put a human name on every agent-produced deliverable

The MarketingProfs AI Update of September 11 carried a Digiday finding that reads like an operating manual. Media agencies including Rise, Brainlabs, Dept, and PMG have built monitoring and auditing systems for AI agents after discovering that automated planning and buying generates errors, drifts beyond instructions, and burns excessive tokens. They track agent decisions, keep audit logs, impose token and cost limits, and control which models employees can use. Gartner estimates that 60% of organizations using AI will hit cost overruns from inadequate usage tracking.

One detail in that list does more work than the rest. The agencies tie every AI-generated deliverable to an accountable human, and they pick models on cost, task requirements, legal considerations, and data-location rules.

It reminds me of what Shawn Surber, Director of Solutions Engineering at Island, said when I interviewed him on The Agile Brand podcast: “An agent operating under your brand’s credentials is your brand’s action, full stop.” Accountability follows the credential. Somebody holds it whether or not you have written the name down.

Ernst & Young’s US division put a price on the same capability. The firm is spending $100 million this fiscal year on bonuses for employees who demonstrate adaptability, judgment, and innovation, with awards reaching $25,000 for individuals or teams making major impact. KPMG has pushed critical thinking into audit internship training. PwC has added empathy and creativity alongside AI skills. Those firms bill judgment by the hour, so they notice faster than the rest of us when its market value moves.

Here is the worked version. Take a lifecycle team producing 400 agent-assisted deliverables a month: segment definitions, subject lines, offer selections, journey branch rules. Today the team reports 400 deliverables and a 60% time saving, and nobody can grade anyone on that number. Now name an accountable human on each one. Say 320 fall into the high-frequency, low-complexity class Leow tells you to prioritize, and one person signs for that batch under a standing rule. The remaining 80 each carry a named owner, because each one commits the brand to a position on price, eligibility, or claim.

That gives you a review question worth asking. Of those 80, how many did the named owner change before release, and what did each change protect? A director who changed 22 and can walk you through every one is doing the job. A director who changed two is either exceptional or asleep, and the audit log tells you which within an hour.

Give judgment a number your CFO already respects

Jonathan Knowles and Chris Burggraeve published “Double CFO Syndrome” in Marketing Week on September 11, and it explains the failure mode waiting for every marketer who builds the capability above.

Their starting figures: more than 70% of CEOs come from finance, operations, or sales backgrounds, and only one in 10 Fortune 250 CEOs has direct marketing experience. The institutional result is a bias toward the inside of the business, where all you have is costs. The CMO ends up facing two CFOs.

Knowles and Burggraeve map the tension onto four quadrants a CEO has to satisfy at once: value for investors against value for customers, and current performance against future performance. The CFO owns the top left by job description. The CMO owns the bottom right by job description. Then they name the asymmetry that decides who wins the argument. Widespread consensus exists on how to measure the top left. Almost no consensus exists on how to measure the bottom right. So marketers keep getting pulled into ROI conversations about short-term financial efficiency while the thing they actually built sits in a quadrant with no agreed instrument.

Their proposed metric is Pricing Power, Warren Buffett’s criterion for a company that can raise prices without losing demand or share. Tim Ambler’s definition of brand equity sits underneath it: a reservoir of cash flow that a company has earned and not yet released to the income statement. The wiki entry on brand equity covers the measurement approaches that feed it.

Connect that to the judgment argument and the logic closes. The behaviors the Festival speakers named all produce effects in the bottom right quadrant. Rix’s distinctiveness test protects pricing power directly. Berkemeyer’s insistence on understanding the whole machine is what lets a marketer make the pricing power case in the CFO’s language. Nastou’s measurable goal and step breakdown is the instrumentation. Build the judgment layer and report it in ROAS, and you will have built something your own reporting cannot see.

Thomas McKinlay, Stefano Puntoni, and Serkan Saka published research in HBR on September 9 that supplies the adoption mechanism. Their finding is that employees give agents meaningful autonomy when organizations state an agent’s limitations explicitly, emphasize competence over friendliness, connect recommendations to the user’s broader goals, and preserve human control over consequential decisions. They call the target calibrated trust: people knowing where agents are capable, where they fail, and when they hold the decision. Organizations that skip it end up running sophisticated agents as glorified chatbots.

Calibrated trust and a judgment job description are the same document read from two directions. One tells the agent what it owns. The other tells the person what they own. Write either and you have most of the other.