Developing
An opinion article; the case examples are anecdotal and unverified.
In 60 seconds
- Reeve says marketing organisations reward employees for building AI tools, though most marketers will use tools built by others.
- A Dataiku survey of 685 CIOs (July) found 84% agree employees create AI agents and applications faster than IT can govern them, and 72% say they cannot reliably verify that those agents deliver intended results.
- She proposes four roles: builders, maintainers, monitors and users, the last being the largest group.
- Examples of post-launch failures: an agent not switched off after a promotion ends, a model update changing brand voice and a renamed CRM field breaking lead routing.
Impact Layer
- What changed
- An operations practitioner is arguing for explicit ownership of AI tools after they are built.
- Why it matters
- Ungoverned agents cause silent failures, and only some people want or need to build.
- Who is affected
- Marketing operations leaders and managers who set AI performance expectations.
- Martech stack area
- AI
- Stack impact
- Name a maintainer for every AI automation at launch, and check what your performance reviews reward.
- Business impact
- Low An opinion article by an author who promotes her own model. The Dataiku figures are cited second-hand and the case examples are anecdotes.
- Related technologies
- What to watch
- Whether companies formalise maintainer roles.
Melissa Reeve, founder of Hyperadaptive Solutions, argues in a MarTech opinion article published October 9, 2026 that marketing organisations reward employees for building AI tools although most marketers will use tools built by others.
The evidence she cites
- In April 2025 Shopify CEO Tobi Lutke told employees reflexive AI use is a baseline expectation and would factor into performance reviews; Accenture has linked AI use to promotion eligibility, per a cited IT Brew article.
- A Dataiku survey of 685 CIOs (July): 84% agree employees create AI agents and applications faster than IT can govern them, and 72% say they cannot reliably verify that those agents deliver intended results.
- She invokes Goodhart’s law: when a measure becomes a target, it stops being a useful measure.
Four roles
- Builders start new automations.
- Maintainers keep systems running and fix problems.
- Monitors watch system behavior and catch drift.
- Users apply AI to their work without building tools, the largest group.
Our read
The examples are anecdotes, but the failure modes are real: an agent left on after a promotion, a model change that shifts voice, a renamed field that breaks routing. Assign a named maintainer at launch and review what your reviews reward.
Sources
- MarTech: Not everyone in marketing should be an AI builder (opens in a new tab)
Source of all claims; the author is a contributor and the article is opinion.