Climbing OAR Levels With Sling

• 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 climbing-oar-levels-with-sling 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.

• Run SLING on this agent rollout before we push it to the whole sales team.
• Our AI pilot stalled after the first month. Which step did we skip?
• Should we rip out our sequencing platform and replace it with an agent stack?
• We flipped the switch and never captured a before number. What now?

• Label each of our AI initiatives as Optimize, Amplify, or Reinvent.
• Give us a scorecard for whether this initiative is real work or theater.
• Draft an email to the vendor asking for a discount.

---
name: climbing-oar-levels-with-sling
description: Sequence AI change through a live revenue org in five moves so the organization gains levels without a rollback catastrophe, and name the skipped move when an initiative has already failed.
---

# Climbing OAR Levels With SLING

Sequence AI change through a live revenue org in five moves so the organization gains levels without a rollback catastrophe, and name the skipped move when an initiative has already failed.

## When to use this skill

- Optimize work is producing results and the move to Amplify has stalled.
- A rollout is about to touch a production revenue system and the blast radius has not been contained.
- An initiative failed and the team needs to name which move was skipped rather than blame the vendor.
- Someone is proposing to rip out a working system and replace it with a new agent platform.
- A result is being claimed and there is no pre-AI baseline behind it.

## Inputs to collect

- Current OAR level for the function, from classifying-ai-ambition-with-oar or an equivalent audit.
- Data health for the workflow: field completeness, duplicate rate, identity resolution, and pipeline hygiene. Source: RevOps and the CRM admin.
- Shared definitions for the key objects and stages, and who owns each definition. Source: the RevOps data dictionary.
- The current stack diagram with every integration point and the team that bought each tool. Source: RevOps or IT asset inventory.
- The list of live Optimize bets and their budgets, plus the proposed larger bet. Source: the AI program tracker and finance.
- Rollout mechanics: the smallest group the change can be released to and how it is rolled back. Source: the systems owner.
- The pre-AI baseline for the target metric, its value, and the date it was captured. Source: CRM or finance report dated before the change.

## Process

Run the five moves in order. When the skill is being used as a post-mortem, run the same five in order and stop at the first move that was skipped, because a failure usually traces to the first or the last one (https://www.therevenueaireport.com/frameworks/sling).

1. **Set the foundation.** Fix data, definitions, identity, and pipeline hygiene before buying anything. Artifact: a data health report with the fixes closed and dated.
2. **Layer it in.** Thread AI through the seams of the stack you already own before ripping anything out. Different teams buy different tools for different reasons and nobody connects them, so the seam is the work. Artifact: an integration map naming each seam and its owner.
3. **Iterate in parallel.** Run small Optimize bets and one larger Reinvent bet at the same time. Small wins fund the big bet, and the big bet pulls the organization forward before the small wins run out of budget. Artifact: a two-track funding plan.
4. **Nudge, don't shove.** Move each production system one step at a time. Contained rollouts contain failures. Artifact: a staged rollout schedule with a rollback procedure per stage.
5. **Ground truth every number.** Measure against a pre-AI baseline captured before the switch was flipped. Artifact: a baseline record with a value, a capture date, and a named owner.

## Decision rules

- Do the foundation work first, even though it is unglamorous. Data preparation absorbs most AI project time, and most projects fail on data quality before they fail on the model (https://www.therevenueaireport.com/frameworks/sling). Jonathan Kvarfordt states the mechanism as "It requires a lot of what people call unsexy work, because no one gets on LinkedIn and says, hey, I just mapped out 30 workflows" (https://www.therevenueaireport.com/frameworks/scale).
- Do not automate a broken process. Skipping the foundation move builds faster broken systems (https://www.therevenueaireport.com/frameworks/sling).
- Thread rather than replace. Most systems can be threaded, and tearing out a working system for a new agent platform is the named failure mode (https://www.therevenueaireport.com/frameworks/sling).
- Do not run small bets only. Three Optimize wins do not answer one competitor's structural advantage (https://www.therevenueaireport.com/frameworks/sling).
- Roll changes in contained stages. Rolling a large change to a whole team at once means that when it breaks, it breaks everywhere (https://www.therevenueaireport.com/frameworks/sling).
- Refuse to claim a result without a pre-AI baseline. Without the pre-AI number the post-AI number cannot be claimed, and you cannot tell whether AI moved the number, the market moved it, or a rep got lucky (https://www.therevenueaireport.com/frameworks/sling).
- In a post-mortem, check Set the foundation and Ground truth first. Those are the two moves the site names as most often skipped (https://www.therevenueaireport.com/frameworks/sling).
- Budget the reversal rather than treating it as scandal. In a 10-country sample of 2,527 senior decision makers, 74 percent had already rolled back or shut down a deployed AI customer communications agent over a governance failure, and 98 percent were increasing AI investment anyway (Sinch, The AI Production Paradox, fielded January to February 2026, vendor research, https://www.therevenueaireport.com/research/rollback).
- Diligence the data boundary before model quality. Data leakage pulled more agents out of production than hallucination did, which makes it an architecture and permissions problem before it is a model problem (https://www.therevenueaireport.com/research/rollback).
- Expect the pilot-to-production drop and plan the stages around it. S&P Global Market Intelligence, 451 Research, n=1,006, margin of error plus or minus 3 points, found firms abandoning the majority of AI initiatives before production rose from 17 percent to 42 percent year over year, with an average of 46 percent of proofs of concept scrapped (https://www.therevenueaireport.com/research/rollback).
- Fix the specification before the model. RAND, based on 65 practitioner interviews, found misunderstanding of the problem to be solved was raised in 84 percent of leadership-driven failure cases (https://www.therevenueaireport.com/research/rollback).
- The framework publishes no stage size, no data-quality pass mark, and no ratio between small bets and the big bet. Set each with the systems owner and finance, write them into the plan, and hold them for the whole rollout so a bad stage cannot be redefined as a good one.

## Output requirements

Deliver a five-row sequencing table, or the same table filled in retrospectively for a post-mortem.

| Move | Status | Artifact | Owner | Evidence it is done | Blocking the next move |
|---|---|---|---|---|---|
| Set the foundation | Done or not started | Data health report | | Duplicate rate and stage-timestamp completeness measured and dated | |
| Layer it in | | Integration map | | Every seam named with an owning team | |
| Iterate in parallel | | Two-track funding plan | | Small bets and one larger bet funded in the same period | |
| Nudge, don't shove | | Staged rollout schedule | | Stage 1 group size and rollback procedure written | |
| Ground truth every number | | Baseline record | | Baseline value with a capture date before the change | |

For a post-mortem, add one line naming the skipped move, the evidence that it was skipped, and the specific failure it produced. Do not name more than one primary skipped move, because a post-mortem that blames everything changes nothing.

## Verification loop

Validate the sequence before any production change is scheduled.

1. Walk the five rows in order and confirm each artifact exists as a file or a record, not as an intention.
2. Confirm the baseline capture date is earlier than the first rollout stage date. If it is not, the baseline is contaminated.
3. Confirm the stage one group is the smallest group that can produce a readable signal, and that a rollback procedure is written for it.
4. Confirm the integration map names an owning team for every seam. An unowned seam is where the pilot breaks on live data.
5. Fix any failure and repeat checks 1 through 4 in full, because moving a rollout date usually invalidates the baseline check too.

Only proceed to schedule the production change when all four checks pass on the same version of the plan and the rollback owner has confirmed in writing that they can trigger it alone. If a data-quality fix is still open, present the plan with the foundation row marked not done and stop there rather than layering onto a broken foundation.

## Quality checks

- Every move has an artifact and a named owner.
- The baseline has a value and a capture date that precedes the change.
- The rollout is staged rather than switched on at once, and stage one is contained. The framework publishes no stage count, so the team sets it; two stages is the minimum that makes the word staged mean anything, and the number chosen must be stated with the reason.
- The integration map covers every tool that touches the workflow.
- The post-mortem names exactly one primary skipped move.
- No claimed result appears without its baseline next to it.

## Limitations

- SLING sequences change. It does not choose the ambition level, which is the job of OAR, and it does not decide whether the initiative is worth doing.
- A clean sequence does not prevent a vendor from underperforming. It contains the damage and makes the failure legible.
- Post-mortem attribution to a single skipped move is a simplification. It is chosen deliberately, because a single named move produces a fix and a list of five does not.
- The rollback base rates cited here come from vendor-published research with stated samples. Read them as direction, not as your own rate.

## Example input

A mid-market company at Optimize on call summarization. Proposal is to replace the current sequencing tool with an agent platform across all 60 reps next month. No data audit has been run. The pipeline number to be claimed is meetings-to-opportunity conversion. Illustrative and synthetic, provided to show output shape.

## Example output

Set the foundation: not done. No duplicate rate or stage-timestamp completeness has been measured, so any conversion claim is unverifiable. This blocks every later move.

Layer it in: the proposal is a rip and replace, which is the named failure mode. Recommendation is to thread the agent into the existing sequencing tool at the two seams RevOps already owns.

Iterate in parallel: the plan funds one large bet and no small bets, so there is nothing funding the big bet if the quarter turns.

Nudge, don't shove: 60 reps at once is a shove. Recommendation is stage one at six reps in one segment with a written rollback to the current tool.

Ground truth every number: no baseline captured. Capture meetings-to-opportunity conversion for the prior two quarters before stage one starts.

Verdict: do not schedule the production change. Two of five moves are unmet and both are the two the framework names as most often skipped. Flagging for human review by the RevOps director and the systems owner.

## 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/sling
- https://www.therevenueaireport.com/frameworks/oar
- https://www.therevenueaireport.com/frameworks/scale
- https://www.therevenueaireport.com/frameworks/reversal-ledger
- https://www.therevenueaireport.com/research/rollback
- https://www.therevenueaireport.com/research/what-works
- https://www.therevenueaireport.com/research/spend-vs-attribution
- https://www.therevenueaireport.com/data/reversal-ledger
- https://www.therevenueaireport.com/methodology/reversal-ledger

## Cite this framework

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

What does the Climbing OAR Levels With Sling skill do?

Sequence AI change through a live revenue org in five moves so the organization gains levels without a rollback catastrophe, and name the skipped move when an initiative has already failed.

Where does the Climbing OAR Levels With Sling 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 Climbing OAR Levels With Sling 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 Climbing OAR Levels With Sling skill not be used?

Do not use it for: Label each of our AI initiatives as Optimize, Amplify, or Reinvent. Or: Give us a scorecard for whether this initiative is real work or theater. Or: Draft an email to the vendor asking for a discount.

How do I install the Climbing OAR Levels With Sling SKILL.md file?

Download the file, create a folder named exactly climbing-oar-levels-with-sling 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.