The plain question
Should AI adoption be led from the top, or should it grow from the teams doing the work?
The practical answer is: both are needed, but they play different roles.
Bottom-up adoption finds the real workflow pain. Top-down leadership gives permission, priority, resources, and rules.
When either side is missing, AI adoption tends to stall.
What bottom-up adoption is good at
People close to the work usually know where time is wasted.
They know which questions repeat, which documents are hard to find, which handoffs are messy, and which tasks everyone quietly dislikes.
That makes bottom-up discovery valuable.
Examples:
- Support agents know which questions repeat every day.
- Operations staff know which spreadsheet handoffs break.
- Sales teams know where lead intake is inconsistent.
- HR teams know which policy questions keep coming back.
These observations are often better than abstract strategy because they are connected to real work.
Where bottom-up adoption fails
Bottom-up energy can still fail if it lacks structure.
Common problems:
- too many small tool experiments
- no shared security or data rules
- no budget owner
- no agreement on what counts as success
- no path from personal productivity to team workflow
The result is scattered AI usage. People experiment, but the company does not learn.
What top-down leadership is good at
Leadership can create the conditions for AI adoption to become operational.
Leaders can decide:
- which business problems matter most
- what risk boundaries must be respected
- who owns the pilot
- what budget is available
- what metrics should be reported
- when a pilot should expand or stop
This matters because AI workflow adoption is not only a technology decision. It changes responsibility, review, quality control, and sometimes customer experience.
Where top-down adoption fails
Top-down AI programs often fail when they stay too abstract.
Examples:
- “We need an AI strategy.”
- “Every department should use AI.”
- “Let’s transform operations with automation.”
These statements may be directionally correct, but they do not identify the first workflow.
Without a workflow, teams do not know what to change on Monday morning.
The better model: leadership frame, workflow proof
A better approach is to combine both.
Leadership sets the frame:
- why AI matters
- what risks are unacceptable
- which business areas matter first
- how pilots will be judged
Teams bring workflow candidates:
- repeated support questions
- internal knowledge lookup
- document-heavy admin work
- sales intake
- operational handoffs
Then the company chooses one or two workflows to test with clear ownership and metrics.
Example
Weak top-down approach:
“This year we will become an AI-first company.”
Weak bottom-up approach:
“Everyone can try whatever AI tools they like.”
Better combined approach:
“This quarter, we will test AI on two repeated workflows: support policy replies and internal SOP lookup. Each pilot must have a workflow owner, approved source material, human review, and one measurable outcome.”
That gives both direction and practicality.
Questions leaders should ask
Before launching an AI adoption program, ask:
- Which workflows are teams already asking to improve?
- Which ones are repeated enough to test?
- What data or source material can AI use safely?
- Who owns each workflow?
- What human review is required?
- What metric will decide whether we expand?
These questions connect leadership ambition to operational proof.
The Glasrocks view
Top-down without workflow proof becomes strategy theater.
Bottom-up without leadership frame becomes scattered experimentation.
The useful middle is a managed workflow pilot: leadership sets the boundaries, teams bring real work, and the pilot proves whether AI belongs in daily operations.
Start by testing one workflow with the AI Workflow Fit Assessment and use the Glasrocks Method to design the pilot path.