Bottom-Up Use Cases, Top-Down Space
Let workers find AI use cases while leaders supply safety, time, and support
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 96%
This leadership model divides AI adoption responsibilities between two organizational levels. Workers closest to recurring tasks identify promising use cases because they understand the friction, exceptions, and desired outputs inside the work. Leaders do not attempt to prescribe every automation; instead, they create the conditions for discovery by providing training, protected time, approved tools, encouragement, and safe boundaries. Teams then experiment, share what they learn, and surface prototypes that deserve investment. A shared channel, internal wiki, demonstration session, or structured event helps useful discoveries spread beyond their originators. The mechanism combines local operational insight with organizational support, reducing both aimless experimentation and detached executive mandates while building a culture in which AI feels constructive rather than threatening.
Origin
Extracted from Marketing Against the Grain during a discussion of whether AI transformation should be led from the top or emerge from employees.
Core principles
- 01People closest to the work can see the strongest use cases
- 02Leaders enable adoption by creating space rather than prescribing every solution
- 03Training and psychological safety increase useful experimentation
- 04Sharing discoveries compounds learning across the organization
- 05AI adoption works through complementary bottom-up and top-down action
How to run it
- 1
Set the direction
Explain the outcomes the business hopes AI can improve and why employee participation matters. Frame the initiative around better work and customer value rather than indiscriminate replacement.
Pro tip Use a few bounded themes instead of dictating exact solutions.
Watch out Fear-based messaging will suppress honest participation and experimentation.
- 2
Create safe space
Give employees protected time, approved tools, training, and clear data-handling boundaries. Make experimentation legitimate work rather than an extracurricular expectation.
Pro tip Start with a small recurring block of time if a larger event is impractical.
Watch out Telling people to innovate without freeing time creates performative adoption.
- 3
Source use cases from the work
Ask employees to identify repetitive tasks, bottlenecks, and unmet needs in the processes they know firsthand. Let them propose how AI could assist or automate those tasks.
Pro tip Have teams document the current process before proposing the AI intervention.
Watch out Senior leaders may overlook critical edge cases when prescribing workflows from a distance.
- 4
Experiment and demonstrate
Build lightweight prototypes and show the results to colleagues. Compare the new workflow with the current process rather than judging novelty alone.
Pro tip Require each demonstration to state the user, task, benefit, and remaining risk.
Watch out Do not mistake an impressive demo for a production-ready system.
- 5
Share learning continuously
Maintain a common channel or knowledge page for useful resources, experiments, and lessons. Encourage teams to reuse and adapt proven ideas.
Pro tip Summarize discoveries in a consistent template so they remain searchable.
Watch out Scattered private experiments prevent organizational learning.
- 6
Scale selected use cases
Choose the most valuable and manageable experiments for formal implementation. Assign ownership, evaluation criteria, governance, and ongoing support.
Pro tip Prioritize frequent tasks where modest improvements compound.
Watch out Scaling every prototype will waste resources and increase risk.
In the wild
Customer-support employees notice that vague reports slow the product team. Leadership provides experimentation time and approved AI tools, allowing the employees to prototype a workflow that expands each report into expected behavior, actual behavior, and estimated impact.
→ A use case discovered in the work improves the handoff between support and product operations.
A company creates a Slack channel where employees post useful articles, videos, prompts, and experiment results. Leaders participate, protect learning time, and identify ideas that warrant further investment.
→ Distributed discoveries become shared organizational knowledge and accelerate learning.
Common mistakes
Mandating use cases from afar
Leaders who are not doing the work may select tasks that look attractive but do not solve the team's real operational problems.
Offering encouragement without resources
Employees need time, tools, training, and clear boundaries; verbal enthusiasm alone does not create usable capacity.
Keeping experiments isolated
Without a forum for demonstrations and shared learning, teams repeatedly solve the same problems and successful ideas fail to spread.
Is it for you?
Best for
It is best for organizations whose employees understand operational problems better than senior leadership does.
Not ideal for
It is not ideal for high-risk deployments that require centralized design and approval from the outset.
From the transcript
“I think that the use cases bottom up for sure.”
“I think the top down is more like creating space, giving training, encouragement to people.”
“Just create the space. Like I think if we're just still down to one thing, it's create the space.”
From the episode
AI Expert On How To Use Ai To Save Time & Grow Your Business (#149)