Employees Aren't Using Your AI Tools: An Executive Adoption Playbook
Low AI usage is often a workflow and trust problem. Golpo turns one approved task into short, role-specific explanations employees can follow at the moment of use, with the real screen, the allowed input, and the escalation route.
Your AI rollout needs a visible example of good work. Golpo turns approved tasks into repeatable, role-specific explanations employees can follow at the moment of use. A license, a launch event, and a generic feature course leave the allowed input, the review decision, and the escalation route unshown.
Start from the approved task card, the policy, a script you already locked, or a manager’s recording. Golpo drafts the narration, plans the scenes, and lets you approve the script before rendering. You insert the real tool screen, keep one visual identity, and reuse the voice on the next role. The videos teach the workflow your policy already allows. They do not write the policy, authorize sensitive data, or collect adoption analytics.
Who it serves
Executives and enablement leads who can see licenses going unused, and managers who need people to complete one safe task before the next rollout meeting. The artifact is two short videos of the same policy: one for the person doing the task, one for the person reviewing it.
Diagnose the friction, then show the task
Low usage may be rational. People may not know which tasks are allowed, may distrust the output, may have to leave their work system to try the tool, or may expect a mistake to be punished. Interview them and watch the work before you call the behavior resistance.
| Friction | Evidence | What the video has to show |
|---|---|---|
| Relevance | “This does not help my work.” | That role’s real task, not a feature tour. |
| Ability | People cannot produce a safe result. | The allowed input and a finished example. |
| Trust | Outputs need unpredictable correction. | The review decision, in order. |
| Permission | Data boundaries are unclear. | What may be pasted, and what must stay out. |
| Workflow | Extra steps outweigh the benefit. | The path inside the tool they already have open. |
| Reinforcement | Managers ignore the new behavior. | A separate manager video of the same policy. |
Two roles, one policy
This worked example is hypothetical. A support organization has an approved drafting assistant. Agents may draft a customer reply from an approved help article. They may not paste ticket text that includes payment data. Every policy claim in the draft is checked against the article. If the draft states a refund amount the article does not state, the agent escalates to the team lead.
| Support agent | Team lead | |
|---|---|---|
| Moment of use | A ticket needs a customer reply. | A draft is waiting in the review queue. |
| Real screen | The reply box, inserted from your product. | The review queue, inserted from your product. |
| Allowed input | The approved article title and the question. Payment data stays in the ticket. | The agent’s draft plus the article. The lead does not re-paste the raw ticket. |
| Review decision | Each claim matches a sentence in the article. | Send back any draft that states an amount the article does not. |
| Escalation | Team lead, when the amount is not in the article. | Policy owner, when the same error keeps appearing. |
The team-lead video uses the same rule and a different decision. Changing “the agent” to “the lead” in an otherwise identical script is not that second video.
How Golpo builds the pair
- Source. Attach the approved task card and the policy excerpt. In Prompt Mode, ask for a two-minute agent video that shows the reply box, the allowed input, the article check, and the escalation. Repeat with a team-lead brief. When a policy sentence must be spoken exactly, approve it and render in Script Mode. Put visual directions in Video Instructions. About 1,050 characters of narration fit in one selected minute.
- Settings. Insert the real screens with Own Media. Save one Custom Style from an approved brand image or a written description and select it for both videos. Use Canvas Compact when the allowed input and the review decision should stay in one view, or Expanded when each click needs its own space. Notebook is Compact only. For a voice the teams will recognize, select an authorized voice clone of a manager who gave permission, or direct a generated voice with Voice Instructions.
- Script review. Use Edit script before creating video. Confirm the prohibited paste, the article check, and the escalation owner are actually in the narration. Then render.
- Result. You should have two screen-plus-diagram videos: the agent at the reply box, and the lead at the queue. Same policy, same style, same approved voice, different screens and decisions.
- One correction. If the agent video’s picture says “paste the ticket” while the narration correctly says “paste the article title,” fix that label with frame editing. Frame editing charges the selected duration and does not change the soundtrack. If the spoken line is wrong too, revise the script and render again.
- One reuse. The team-lead video is the reuse. Keep the saved style and the saved voice. Swap the screen, the input, and the decision. When a later language is required, render that language separately and review it. An Own Narration recording is not translated by a language setting.
What each control changes
| Control | Why it matters at rollout | Plan |
|---|---|---|
| Own Media | The reply box and the review queue are your screens, so people recognize the moment of use. A generated mock of the interface is not evidence of how the tool behaves. | Business and higher, and Pay As You Go |
| Custom Style | One named style, saved in My Styles, makes the agent video and the lead video look like the same program. It follows the reference. It does not reproduce exact typography. | Business and higher, and Pay As You Go. Custom Style |
| Voice Clone | An authorized saved voice generates new narration for both roles, so the manager does not re-record every script. Get permission, check pronunciation, and keep a fallback voice. | Business and higher, and Pay As You Go. Voice Clone |
| Voice Instructions | Tone, pace, and pronunciation for a generated voice. Useful when you are not cloning anyone. They do not apply to a finished Own Narration recording. | Creator and higher, and Pay As You Go |
| Own Narration and Picture-in-Picture | When recognition should be the actual person, the manager’s recording drives the visuals. Picture-in-Picture keeps them on screen. Audio-only does not create a face. The recording sets the duration and the spoken language. | Business and higher, and Pay As You Go. Own Narration · Picture-in-Picture |
| AI Presenter | When a guide beside the screen helps, place a stock or custom presenter next to the explainer. Side-by-side keeps the diagram primary. A presenter is a generated performance, not a recording of your manager. Presenter API access is still described as coming soon, so build these two videos in the dashboard. | Creator and higher, and Pay As You Go. AI Presenter |
| Canvas Compact or Expanded | Compact keeps the input and the review decision together. Expanded gives each click room. Color starts at Growth. | Compact and Expanded |
Plan, duration, and credits
Each role video in this walkthrough is 2 selected minutes. A standard video uses 1 credit per selected minute, so the agent video is 2 credits and the team-lead video is 2 credits. One frame-edit correction of the agent video is another charged version at that duration: 2 credits. The screen inserts, Custom Style, and voice clone used for the pair require Business or higher, or Pay As You Go. On Pay As You Go, edit-script-before-video bills 10% after the script and 90% after the video. Amounts are on pricing. The Free plan is not a downloadable publishing workflow.
Managers and measurement
Managers need their own video and their own behavior: set the expected task, review the result, discuss a bad draft without punishing the experiment, and report friction that blocks the work. Their behavior teaches more than a launch message.
Define qualified adoption before the meeting: an eligible person uses the tool on an approved task, applies the verification step, and produces an acceptable result. Pair that with task time, rework, exceptions, help requests, and the business outcome you already track. Active-user counts can rise while reply quality falls. Those measures live in your telemetry and ticket system. Golpo does not collect them, and publishing the videos does not authorize a new class of customer data.
Rollout sequence
- Observe one role and choose two valuable tasks.
- Write the allowed-use card: trigger, input, output, verification, prohibited data, escalation, and done.
- Produce the agent video and the manager video in Golpo from that card.
- Correct one wrong label or one wrong spoken line, then reuse the style and voice on the second role.
- Remove the extra click that makes the tool harder than the old path.
- Review qualified adoption in your own telemetry and keep, revise, or stop.
What to leave out
- A usage target with no task and no quality bar.
- The same examples for every role.
- Hidden limits, used to protect enthusiasm.
- A campaign that treats non-use as a communication failure.
- A reward for speed that skips the article check.
Frequently asked questions
Why are employees not using our AI tools?
Common causes are weak task relevance, workflow friction, unclear permission, low trust, insufficient practice, and absent manager reinforcement. A visible, role-specific example of the allowed task gives people something they can follow at the moment of use.
Should we mandate AI usage?
Mandating raw usage can reward unsafe or pointless behavior. Define approved tasks and quality conditions first, then show those tasks in short role-specific videos.
What is qualified adoption?
Use of an approved AI workflow by an eligible user with required verification and an acceptable task result. Measure it in your own telemetry. Golpo does not collect that usage data.
How can managers improve adoption?
Managers can select relevant tasks, model safe behavior, review outcomes, normalize escalation, and surface friction. A manager-approved voice and one saved visual style can carry both the employee video and the manager video.
Can Golpo track employee AI usage?
Golpo creates the enablement videos. Usage counts, task success, and business outcomes stay in your own tools and analytics. The videos also do not set policy or authorize sensitive data.
Can one video train every role?
Use the same approved policy, saved style, and approved voice, and change the screen, the allowed input, and the decision. A pronoun change is not a role variant. Produce a distinct video for each role that must act differently.
Before the next rollout meeting
Create the two approved task videos — the person doing the work, and the person reviewing it — and agree how you will read successful task use in the telemetry you already have. If the underlying initiative is stalled, use the pilot failure diagnostic. For the policy that feeds the task card, see policy-to-role training.
Tags
