What Does It Take to Get a Marketing Operations Team to Actually Use AI Tools?

Enterprise marketing teams have become reasonably good at selecting and deploying AI tools. The harder problem - getting the team to actually use them, consistently, as part of daily work - receives far less attention and produces far more failure.

The pattern is consistent: a tool is selected through a structured evaluation process, deployed on schedule, and introduced to the team with a product walkthrough. Usage spikes in the first two weeks, then declines steadily. By day 90, a small number of early adopters are using the tool regularly, the majority have reverted to prior methods, and the organization is paying for a capability it is not realizing.

This is not a technology problem. The tool works. It is a change management problem, and it has a practical solution that does not require a large program or external consultants to execute.

The Prosci ADKAR model identifies five conditions required for individual behavior change: Awareness, Desire, Knowledge, Ability, and Reinforcement. Most AI tool deployments in marketing operations address Knowledge through product training and stop there. Desire, Ability, and Reinforcement - the conditions that determine whether a behavior becomes a habit - are left to chance. Rarovera's consulting work with enterprise marketing and operations teams consistently surfaces this same gap: the deployment plan is detailed; the adoption plan is a slide.

AI adoption in marketing operations fails at the workflow integration layer more often than at the technology layer. The tool is capable; the team does not change how they work to use it.

Three specific failure patterns account for most cases:

• The tool is added alongside the workflow, not embedded in it. If using the AI tool requires a team member to open a separate application, re-enter context that already exists in their primary system, or complete an additional review step that was not part of the prior process, most people will revert within weeks. This is the ADKAR Ability gap in practice: the team knows how the tool works but cannot use it without disrupting the flow of real work. Friction compounds. A tool that saves time in the task but costs time in setup will not be used by a team under deadline pressure.
• The value proposition is stated at the organizational level, not the individual task level. 'This tool will improve our content velocity' is not a reason for a content operations manager to change their daily behavior. 'This tool will cut the time you spend briefing an agency' is. The individual task-level answer to 'what is in it for me' is the Desire lever in ADKAR; the organizational benefit is the business case, and those are different conversations requiring different communication.
• Adoption accountability is diffuse. The NIST AI RMF 1.0 Govern function and ISO/IEC 42001:2023 both require that AI tools in production use have named internal owners and documented accountability structures. When no named person is accountable for whether the team is using the tool at 60 days, no one is watching the early warning signs. The vendor is not accountable for adoption. IT is not accountable for adoption. The marketing operations leader who approved the purchase is accountable, and they need to delegate that accountability explicitly to an internal champion before launch, not after the 90-day review reveals a problem.

Three interventions, applied before launch rather than after adoption stalls, consistently improve sustained AI tool use in marketing operations teams. Each addresses a specific gap in the ADKAR change sequence. These are not theoretical; they are the interventions Rarovera applies when enterprise marketing teams ask why a tool they selected carefully is not being used six months later.

1. Embed the tool in the workflow before training anyone on it. Work with the internal champion and one or two representative team members to map exactly where in the existing workflow the tool will be used. Identify the specific moment in the task sequence where the tool replaces or augments a current step. Configure the tool to minimize context-switching. Only then design training around that embedded workflow. Training people on a tool that is not yet integrated into their work addresses Knowledge but not Ability. Ability requires practice in the actual work context, not a demo environment. This sequencing distinction is the single most common fix Rarovera applies to stalled AI adoption programs.

2. Build task-specific, role-specific training sessions. A content operations team and a campaign operations team use AI tools differently. Train them separately, on the tasks they actually perform, using real work samples from their own queue. Generic product walkthroughs produce initial engagement and rapid drop-off because they address Knowledge without creating the task-level Ability the team needs. Task-specific training produces slower initial uptake and sustained use, because team members leave the session knowing exactly what to do the next morning. Follow each session with a structured practice period on real work before the next session, so Ability is built in the context where Reinforcement will later operate.

3. Name a champion and give them a clear accountability mandate. The champion's job is not to be an internal help desk. It is to monitor active usage, surface structural friction before it becomes abandonment, and maintain the connection between the tool's capabilities and the team's evolving task load. The champion provides the Reinforcement stage of ADKAR: the ongoing support and correction that converts a new behavior into a durable habit. The champion should report adoption metrics to the marketing operations leader at 30, 60, and 90 days, with a clear escalation trigger: if active usage rate drops materially below the pre-launch target at day 60, that is a structural problem requiring a process or configuration change, not more training.

Measure active usage rate, not login rate. Active usage means the tool was used to complete a defined task that produces a measurable output. Login rate measures access; it says nothing about behavior change. Set the active usage target before launch, baseline it in week one, and measure at 30, 60, and 90 days.

A decline between day 30 and day 60 is the most important signal in the adoption curve. It indicates that initial curiosity has worn off and the tool has not yet become habitual. That is the intervention window. Waiting for the 90-day review to address a day-60 decline means losing the window when the behavior is still malleable.

ISO/IEC 42001:2023 section 9 requires ongoing monitoring and measurement of AI management system performance. An active usage metric, tracked by the internal champion and reported to the marketing operations leader, satisfies that requirement at the team level and connects the adoption workstream to the broader governance obligation the organization already carries.

What this article does not cover: This article addresses adoption change management for AI tools already selected and deployed. It does not cover the evaluation and selection process for new AI tools, post-deployment governance and audit requirements, or the technical integration architecture that makes workflow embedding possible. Those are distinct workstreams. Rarovera works with enterprise marketing and operations teams across all three stages; this article focuses on where adoption most often breaks down after a technically successful deployment.

Disclosed interest: Rarovera advises enterprise marketing and operations teams on AI adoption, marketing operations design, and platform strategy. The approach described here reflects the kind of work we do with clients. We have a commercial interest in organizations finding it useful.