Ramping AI Behavior Change With Lopaft

• Source: Report framework library

Encodes a Revenue AI Report framework so the agent applies the published method instead of improvising one.

1. 1.Copy the SKILL.md text below, or download the raw file.
2. 2.Create a folder named exactly ramping-ai-behavior-change-with-lopaft in your agent's skills directory.
3. 3.Save the file inside that folder as SKILL.md.
4. 4.Ask the agent one of the trigger requests below.
5. 5.Check the output against what you already know before it leaves your desk.

• We trained everyone on the AI assistant in June and nobody is using it on live deals
• Run a LOPAFT diagnostic on our Agentforce rollout and tell me which rung is missing
• Adoption spiked at launch and collapsed a month later, what do we fix

• The pilot ends Friday, give me the go, fix, or stop call against our criteria
• Code the reason we shut off the AI SDR and calculate the sunk spend

---
name: ramping-ai-behavior-change-with-lopaft
description: Name the single rung of the adoption ladder an AI rollout is stuck on, then prescribe the fix that rung requires.
---

# Ramping AI Behavior Change With LOPAFT

Name the single rung of the adoption ladder an AI rollout is stuck on, then prescribe the fix that rung requires.

## When to use this skill

- A tool was announced, a training was held, and behavior on live deals has not changed.
- Adoption numbers rose during the launch window and fell back afterwards.
- A CEO or CRO asks why the rollout is not working and the current answer is that people need more training.
- Designing the adoption motion for a new AI capability before it ships to the team.
- Running a post-mortem on a failed AI rollout where the tool itself worked in demo conditions.

## Inputs to collect

- The behavior you expect to see on a live deal, written as an observable action, from the rollout charter or the buying business case.
- What the team was given at launch, from the enablement calendar, covering the training deck, SOP, or launch email.
- Whether an examples library of pass and fail samples exists, from the enablement content repository.
- Whether a sandbox, role play, or dry-run environment exists, from the enablement or IT owner.
- Usage on live deals by person, from the tool's admin console and the CRM.
- Whether anyone measured a person's output against a written standard and told them the result, from manager one-to-one notes.
- Whether any team member has taught the capability to someone else, from the enablement roster.
- The stated standard the work is measured against, from the enablement rubric. If none exists, record that as the finding.

## Process

Run the six rungs in order and stop at the first one that fails. LOPAFT is a diagnostic, not a curriculum, so the goal is to name the rung rather than to rebuild every rung (https://www.therevenueaireport.com/frameworks/lopaft).

1. Learn. Someone explains it. First exposure to the content, through a training deck, SOP, or launch email. Artifact: the launch material and the date each person received it.
2. Observe. Watch what good looks like and what bad looks like, because without both a person cannot tell where they stand. Artifact: the examples library, recorded calls, and pass and fail samples.
3. Practice. A safe place to try, fail, and try again, with no stakes, no judgment, and no live customer. Artifact: the sandbox, role play, or dry-run log with reps per person.
4. Apply. Real work, real deals, real risk. The first live rep. Artifact: the count of live deals per person where the behavior appeared.
5. Feedback. An honest read on the work, measured against a standard. Not how did it go, but a measurement. Artifact: the rubric plus one delivered measurement per person.
6. Teach. Teach it to someone else, which is the rung that locks it in, because if a person can teach it they understand it. Artifact: the name of who taught whom, and when.
7. Write the diagnosis as one rung plus one fix plus one date. Artifact: the diagnosis line.
8. Re-measure the observable behavior on live deals after the fix has had time to land.

## Decision rules

- Stop at the first failing rung and fix only that one. Running another expensive event to solve a Practice problem or a Feedback problem wastes the budget and leaves the rung intact (https://www.therevenueaireport.com/frameworks/lopaft).
- Suspect the Practice-to-Apply seam first. In most failed AI rollouts the problem sits between Practice and Apply, because nobody got safe reps. In the rest, Feedback is missing, because nobody told the operator whether the work was any good (https://www.therevenueaireport.com/frameworks/lopaft).
- Reject a training event as the fix for anything above Learn. As Kvarfordt puts it, most teams use training as the only thing that they do, the initial rollout of a three-day event, then another thing a month later, and that is not where behavior changes but where learning happens (https://www.therevenueaireport.com/frameworks/lopaft).
- Do not count a rollout as having an Observe rung unless both good and bad samples exist. One-sided examples leave the person unable to locate their own work.
- Require a written standard before you claim Feedback exists. Feedback is a measurement against a standard, not a conversation, and the framework quotes Thomas Monson via Kvarfordt that when performance is measured and reported back on, that is when behavior changes (https://www.therevenueaireport.com/frameworks/lopaft).
- Deliver feedback privately. Group feedback in front of peers kills adoption, because feedback has to be safe before it can be honest (https://www.therevenueaireport.com/frameworks/lopaft).
- Never send a rep to a live customer as their first attempt. Handing a rep a new tool and pointing them at a live customer burns the deal and the rep's trust in the tool (https://www.therevenueaireport.com/frameworks/lopaft).
- Treat Teach as a retention control, not a nice-to-have. Untaught skill stays fragile, and when the person leaves the role the organization loses the capability (https://www.therevenueaireport.com/frameworks/lopaft).
- Instrument the constrained step rather than the freed step. Time saved is an input claim, while stage conversion and cycle time are the test, so a rollout that frees hours into an unchanged buying process produces queue rather than throughput (https://www.therevenueaireport.com/research/proof-gap).
- Pair the ladder with a workflow change where one is available. McKinsey tested 25 organizational attributes and found fundamentally redesigning workflows had the largest effect on EBIT impact, with only 21 percent having redesigned any workflow at the time (https://www.therevenueaireport.com/research/what-works).
- Build human oversight into the Apply rung for anything agentic. In the Capgemini series, human oversight with the ability to overrule ranks fourth among trust improvers at 36 percent, behind demonstrated accuracy and reliability at 52 percent (https://www.therevenueaireport.com/research/trust).
- Use the published window as the only time expectation. The framework states behavior change on live deals in 60 to 90 days rather than a training event and a hope (https://www.therevenueaireport.com/frameworks/lopaft). No page publishes a per-rung pass rate or a required number of practice reps, so set those with the team and label them as team-set.

## Output requirements

- One named rung as the diagnosis, with the evidence that rung failed.
- One fix, owned by one person, with one date.
- The observable live-deal behavior that will be re-measured, and when.
- A rung-by-rung table showing which rungs hold.

Use this table shape.

| Rung | Artifact required | Present | Evidence |
|---|---|---|---|
| Learn | Training deck, SOP, or launch email | yes | launch email 2026-06-02 |
| Observe | Pass and fail samples | partial | good calls only, no fail samples |
| Practice | Sandbox, role play, dry runs | no | no sandbox provisioned |
| Apply | Live-deal usage per person | no | 11 of 40 reps, one deal each |
| Feedback | Written standard plus delivered measurement | no | no rubric exists |
| Teach | Named person taught another | no | none |

## Verification loop

1. Validate the diagnosis by checking the rung below it. If the rung below also fails, the diagnosis is too high on the ladder, so move down one rung and re-run this check until the rung below holds.
2. Validate the artifact, not the intention. For each rung marked present, open the artifact. If the artifact cannot be produced, mark the rung absent and return to step 1.
3. Validate that the fix matches the rung. A fix that adds content addresses Learn. If the failing rung is Practice, Apply, Feedback, or Teach and the proposed fix is content or an event, reject it and rewrite the fix.
4. Validate the behavior definition. Confirm the expected behavior is observable in the CRM or the call recording without an interview. If it is not, rewrite it and re-run steps 1 to 3.
5. Re-measure after the fix window. Compare live-deal behavior against the pre-fix baseline at the date set in the diagnosis.
6. Only proceed when the rung below the diagnosis holds on evidence, every present rung has a producible artifact, the fix is rung-matched, and the behavior is observable without an interview. Make a rollout-wide change only after that point. If any check fails, re-run the ladder rather than escalating scope.

## Quality checks

- Exactly one rung is named as the diagnosis.
- Every rung marked present has an artifact that can be opened.
- The fix does not include an additional training event unless the failing rung is Learn.
- Feedback is defined as a measurement against a written standard, not a conversation.
- The expected behavior is stated as an observable action on a live deal.
- The re-measure date is set before the fix begins.

## Limitations

- LOPAFT diagnoses adoption. It cannot tell you whether the tool was worth buying, and a well-adopted tool with no revenue effect is still a miss on the original thesis.
- The ladder assumes the capability works. If reliability is the real problem, adoption work will not fix it, and agents without a cheap automatic check before an irreversible action collapse across repeated attempts (https://www.therevenueaireport.com/research/agent-reliability).
- Self-reported usage overstates behavior change. Prefer CRM and call-recording evidence.
- The framework publishes no per-rung pass thresholds, so any numeric gate is a team decision and must be labeled as one.
- The 60-to-90-day window is the framework's stated expectation for behavior change, not a measured benchmark from a survey panel (https://www.therevenueaireport.com/frameworks/lopaft).
- Supporting figures on workflow redesign and trust come from self-reported executive surveys. They are consistent with each other, which is evidence, and they share a common bias, which is a limit (https://www.therevenueaireport.com/research/what-works).

## Example input

Illustrative case, synthetic figures. An AI deal-research assistant rolled out to 40 account executives on June 2, 2026 with a two-hour launch session and a recorded walkthrough. At day 75, admin logs show 11 of 40 reps used it on at least one live deal, none used it twice, no sandbox was provisioned, no rubric exists, and managers report they have not reviewed anyone's output against a standard.

## Example output

Diagnosis: the rollout is stuck at Practice. Learn holds, because the launch session and walkthrough exist. Observe is partial, because only good examples were shared. Practice fails, because no sandbox, role play, or dry run was ever provisioned, so the 11 reps who tried it did so on live deals as their first attempt. That is the published failure pattern, since in most failed AI rollouts the problem sits between Practice and Apply (https://www.therevenueaireport.com/frameworks/lopaft).

Fix: one sandbox with three scripted dry runs against closed-won deals from last year, owned by the enablement manager, live by day 90. Add the missing fail samples to the examples library in the same pass, because a person cannot tell where they stand without both.

Do not schedule a refresher event. A refresher fixes Learn, and Learn is not the failing rung.

Re-measure: repeat live-deal usage per rep at day 150, against the day-75 baseline of 11 of 40 reps at one deal each. The framework's stated expectation is behavior change on live deals in 60 to 90 days from the point the ladder is intact (https://www.therevenueaireport.com/frameworks/lopaft).

Escalation note: Feedback and Teach are also absent and will become the next diagnosis once Practice holds. Do not fix them in the same sprint, because the ladder is sequential.

## Rules of conduct

- Write for a Director, VP, or operator. Short sentences. Explain uncommon terms.
- Separate facts from assumptions. Never hide uncertainty.
- Do not invent numbers, benchmarks, quotes, or customer names.
- Do not send messages, change CRM records, or publish anything unless the user explicitly asks.
- Flag when a decision needs human review.

## Evidence

- https://www.therevenueaireport.com/frameworks/lopaft
- https://www.therevenueaireport.com/research/what-works
- https://www.therevenueaireport.com/research/proof-gap
- https://www.therevenueaireport.com/research/trust
- https://www.therevenueaireport.com/research/agent-reliability
- https://www.therevenueaireport.com/blog/just-in-time-enablement-with-ai
- https://www.therevenueaireport.com/blog/enablement-in-the-age-of-agents

## Cite this framework

Kvarfordt, Jonathan. "LOPAFT." The Revenue AI Report. https://www.therevenueaireport.com/frameworks/lopaft

What does the Ramping AI Behavior Change With Lopaft skill do?

Name the single rung of the adoption ladder an AI rollout is stuck on, then prescribe the fix that rung requires.

Where does the Ramping AI Behavior Change With Lopaft skill come from?

Report framework library. It was written by The Revenue AI Report against a 12 criterion quality rubric and graded in an independent scoring pass.

Why was the Ramping AI Behavior Change With Lopaft skill chosen for this library?

Encodes a Revenue AI Report framework so the agent applies the published method instead of improvising one.

When should the Ramping AI Behavior Change With Lopaft skill not be used?

Do not use it for: The pilot ends Friday, give me the go, fix, or stop call against our criteria Or: Code the reason we shut off the AI SDR and calculate the sunk spend

How do I install the Ramping AI Behavior Change With Lopaft SKILL.md file?

Download the file, create a folder named exactly ramping-ai-behavior-change-with-lopaft inside your agent's skills directory, and save the file inside it as SKILL.md. The agent loads it when a request matches the description.