As president, COO and co-founder of a scaled remote company with more than 200 team members, I spend a lot of time navigating the space between AI theory and AI operations. Through Coalition’s consulting and strategy work, I also see how several hundred brands are trying to apply AI across marketing, e-commerce and internal productivity.
Taken together, that gives me a fairly direct view of what AI adoption looks like inside real companies, both from the executive level and from the client side.
Most organizations are not adopting AI through a unified strategy. Middle managers are being asked to build hybrid teams where people and AI systems work together effectively, but they are doing it in the middle of a hype cycle, with limited organizational clarity around where AI should be deployed, how it should be evaluated, which risks matter most, and how success should be measured.
At the same time, AI capabilities are changing fast enough that many companies still lack the information needed to build frameworks that will hold up for more than a quarter or two.
Middle managers are still responsible for output, deadlines, quality control and team performance. At the same time, they are being told to integrate systems that are still poorly understood into workflows that were designed before those systems existed.
A manager may have spent years running a stable team with clear roles and predictable output, but AI changes that environment by adding variability at every step of the way. Even when the tool is useful, it may fundamentally change the “who, what, when, where and why” that had been so reliable before.
And it’s not just evolving processes that create burdens.
The pressure is coming from multiple directions
Leadership wants gains in speed and efficiency (i.e., profitability). Competitors market around their AI utilization. Vendors are promising transformation every other Tuesday. Meanwhile, the manager still has to decide whether a tool belongs in an actual workflow, whether the team knows how to use it, and whether the result will hold up under client or executive scrutiny.
Broader workforce data reflects the same issue. Executive enthusiasm for AI investment tends to run ahead of middle management confidence in implementation, especially around measurement, governance and accountability.
See also: 3 ways HR can take care of middle managers before they break
The uneven adoption problem
Inside most organizations, AI adoption is also not coordinated. It keeps showing up in pockets. One team experiments heavily. Another ignores it. A third starts using tools without approval, documentation or any real framework for evaluating outcomes. That is why AI adoption needs repeatable workflows. Teams need to know where experimentation is encouraged, what requires approval, what can touch company or client data and how useful findings are supposed to be shared upward. The informal enthusiasm of corporate AI culture is not a process. It is a recipe for expensive mistakes and a failed commitment to AI later.
The distraction problem in AI decision-making
One of the first things middle managers need to learn is how to separate reality from all of the hype. This is harder than it sounds because the AI market is built to make everything look urgent.
Managers do not need to track every model release, product announcement, benchmark or founder prediction. In practice, trying to follow all of it usually makes people worse at the job. Instead of evaluating tools against a business need, they start evaluating business needs against the week’s AI headlines.
A lot of this is driven by the economics of the sector.
Many AI companies are under pressure to justify valuations and keep investors engaged. Frequent announcements help with that. They do not always help an operations manager decide whether a tool is reliable enough to use with a client deliverable on Thursday afternoon.
Managers who consume this cycle too closely tend to become less decisive.
How uncertainty spreads through teams
Teams then pick up on the managerial uncertainty.
Team members are told to experiment, but also warned not to make mistakes. They are given access to a tool, but no guidance on where it fits. They are encouraged to move faster, but nothing is defined about review, quality thresholds or acceptable risk.
Some people end up overusing the tool and create clean-up work for everyone else. Some avoid it entirely and fall behind on the new expectation. Others wait quietly for the initiative to lose momentum, which is often a fairly rational read of corporate behavior.
A lot of this ends up feeling familiar to those who have been involved in software deployments at large organizations. We’ve all seen teams buy expensive licenses based on broad expectations. Implementation is rushed. Use cases are vague. Nobody defines success in measurable terms. A month later the tool is barely being used, and the company decides [software X or AI tool Z] was overhyped, when, in reality, the rollout was sloppy.
The long tail of failed AI adoption
Bad rollouts do not just waste money. They make future adoption harder.
When a team has an early experience with AI that feels chaotic, underwhelming or poorly managed, that memory lives long. The next time leadership wants to introduce a tool, the team brings the earlier failure with them. This creates friction across the organization. Even as models improve and more practical workflows become available, the company evaluates new opportunities through the lens of a previous mess.
This disconnect is significant because organizations do not benefit from what is technically possible. They benefit from what is adopted and well-managed. There is a large graveyard between those things.
What effective middle managers do differently
The middle manager’s job in this environment is to create stability.
Doing so requires enough practical understanding to know where AI tends to work and where it doesn’t.
That means beginning with a meaningful grasp on the basics of AI, how they’re trained and how they’re deployed. More than a passing knowledge of mainstream tools and use cases also helps.
Better managers go on to look for tasks that are repetitive, structured, time-consuming and easy to evaluate. Drafting, summarization, categorization, research support, pattern-based analysis and first pass content work are often useful areas to test. Tasks with high error costs (cough-cough: liability), vague standards or heavy judgment requirements (for people) should be less attractive unless the review process is extremely tight.
They also test in controlled conditions. They define the use case. They set a baseline. They measure time saved, quality impact and rework. Then they decide whether the tool belongs in a real workflow. That is much less exciting than making broad statements about transformation, but it tends to produce better results, which remains a mildly useful feature in business.
In our experience, this works best when it is supported by clear SOPs. Team members need to have the same “W questions” we referenced earlier answerable in a new AI test.
Middle managers can help this process by defining specific time periods for their whole team to test and familiarize themselves with AI in particular applications. Work on using AI in note taking for a month, then in task management. After? Perhaps quality assurance and review.
Stepping stones like this help secure value and keep everyone from getting lost in too many ambiguous objectives.
Building a repeatable AI adoption model
Once that structure is in place, AI adoption gets much more durable. Teams understand why a tool is being used, what problem it is supposed to solve and what success looks like.
This also makes it easier to evaluate new tools as they emerge. The company already has a framework for testing, approval, documentation and rollout. A stable evaluation process saves you from rebuilding internal logic every time a vendor updates a homepage.
People tend to produce better tests when they know the rules, the objective and who is reviewing the outcome. Chaos has never been an especially strong management system, despite its popularity.
A durable role in a changing environment
This is an ongoing management issue, not a temporary adjustment period. AI is going to remain part of how companies evaluate labor, process design and productivity. That puts middle managers in a central role whether they asked for it or not.
Middle managers who control the onboarding process for their AI-human teams will prove invaluable to companies, even in companies that are skewing toward the AI teams and away from the human ones.
Credit: Source link
