Roast & Rise

Riseplan from Roast & Rise

Performance and Promotion in the Age of AI-Assisted Work

Build review and promotion criteria that capture judgement, verification, and originality— not just output volume.

AI is changing what counts as great work. This plan shows founders, managers, and HR leads how to evaluate, review, and promote people whose output is deeply AI-assisted. You’ll replace outdated volume metrics with rubrics for judgement, verification quality, and original thinking—so you can recognize what really matters.

A warm-lit workspace table viewed from above. On the table: a carefully arranged set of blank review rubrics, a single orange-gold card standing out among whites and off-whites, and a threshold of light crossing the surface. There are no visible people, devices, or text—just evidence of process, selection, and focused judgement. The scene is minimal, with generous negative space and editorial focus on the artifacts and light, using the Roast & Rise orange and warm neutrals palette. No words, labels, numbers, or logos.
In the AI era, what matters isn’t output volume—it’s clarity of judgement, verification, and originality, revealed through the human decisions layered on top of machine work.

Course thesis

In AI-assisted roles, output is no longer a proxy for performance. Performance must be judged by how well a person directs, verifies, and builds on AI—not just what or how much gets produced. Key skills, like judgement and creativity, have become more valuable, but most review systems still reward speed and volume. This plan gives you the structure and tools to recognize and reward what matters.

What you leave with

By the end, you’ll know how to design review and promotion decisions that reward judgement, originality, and verification in AI-driven roles—backed by clear rubrics, manager guides, and checklists you can use now.

For

Founders, managers, and HR/people leads overseeing teams whose outputs are significantly AI-assisted.

Workflow

Annual or biannual performance review and promotion cycles for teams using AI tools for core deliverables.

Change

Move from reviewing people by visible volume or hours to evaluating the quality of human judgement, verification, and originality applied to AI-supported work, and make promotion decisions that reward these higher-order skills.

What you can do

Use these as checks while you move through the plan.

Diagnose when traditional review criteria no longer match the work being done.

Design rubrics that capture judgement, verification, and originality in AI-supported work.

Practice evaluating AI-assisted outputs with revised review tools.

Implement new performance and promotion criteria in a live review cycle.

Measure the impact on team clarity, performance decisions, and development of core skills.

Chapters

01

See What Changed: Diagnosing the Shift in Performance Signals

Diagnose where AI has changed the true signals of strong performance in your team and identify which old review criteria no longer fit. Build a clear before/after map to spot gaps and target your next review cycle.

A top-down view of a blank worksheet: two adjacent columns, with the first side faint and pale, the second side stronger and bordered in orange. There are abstract boxes and check marks but no text or numbers. Additional review cards are scattered, with two outdated boxes gently crossed by a stroke of orange. No people or hands, just editorially arranged artifacts on a light surface. Warm shadows and material detail suggest a moment of critical reflection.
A diagnostic worksheet shows the before-and-after map of performance criteria—old review boxes faded, new highlighted skill domains ready to surface what AI cannot.

Why this matters in the workflow

AI is under the hood on almost every desk now. What your people do most days no longer matches what your old review template measured. Volume and programmed output are easy to automate, but judgement, verification, originality—these rise to the top when prediction becomes cheap.

The 2026 PwC AI Jobs Barometer shows it plainly: tasks in AI-exposed roles are over twice as likely to need empathy, judgement, or creativity. Junior roles, when AI is a co-pilot, suddenly take on senior skill demands—leadership, gut-calls, context sense. If you don’t map these changes, your next review cycle will reward the wrong things, promote prompt monkeys, and leave real talent unseen.

The working model

Quality checklist

Mapped at least three key tasks with specific detail—not generic labels

Showed how AI changed human involvement for each task

Flagged at least two outdated criteria with clear reasoning

Named skills like judgement, verification, creativity, leadership, empathy—linked to concrete work situations

Language is honest—no vague claims or idealized descriptions

Common mistakes

Writing generic task labels instead of live work

Missing the human layer now required for higher value

Downplaying how much review criteria need to change

Failing to tie new skill needs to specific AI workflow changes

Flagging only technical criteria but ignoring creative or strategic ones

Checkpoint

Did you map at least three tasks, capture both AI changes and new skill demands, and flag two outdated criteria ready for replacement?

Exercise

Map the Shift: From Old Criteria to Real Value

Goal: Create a before/after map for one team to reveal where your review process needs to evolve.

Steps:

  1. Pick a real team or function. List up to five core tasks or deliverables for their last cycle.
  2. For each task, write the actual criteria you used in their last review or reward process (don’t idealize).
  3. For each, document what changed as AI entered the workflow—where does the person spend less effort, and where do they face new challenges?
  4. Using the PwC Barometer (see summary below), write the new key skill(s) each task now demands from a human owner (examples: judgement, verification, creativity, leadership, empathy).
  5. Flag at least two old criteria that are now unreliable proxies for value or growth.
  6. Briefly note which updated skills should be measured next review cycle.

PwC 2026 AI Jobs Barometer summary: New tasks in AI-exposed roles are 2.5x more likely to need empathy, judgement, and creativity; junior roles are 7x more likely to demand senior skills like leadership and strategic thinking.

You have 15 minutes. Be specific and honest.

Use this at work tomorrow

Use this mapping to prep for your next performance review—see where to stop rewarding old proxies and start measuring what your team actually contributes now.

02

Design for Judgement: Building the New Review Rubric

How to rewrite your review process to measure the skills that matter in AI-assisted roles. Build a practical rubric anchored in real behaviours: judgement, verification, and originality.

A minimal, tri-partite card or worksheet on a clean background, physically divided into three sections with slight elevation, each section rimmed with orange. Blank but for tick-boxes or dots (no text or numbers), suggesting judgement, verification, and originality. A sharpened pencil rests nearby. No people, screens, or digital UI. Editorial composition, focused light draws attention to physical separation and structure of rubric space. Palette: white, off-white, orange, with shadow for depth.
A redesigned review rubric: three sections—each distinct, each inviting focused evidence, not vague effort—awaits practical use in the new performance cycle.

Why this matters in the workflow

AI has changed who does what, but most review rubrics are still locked in a pre-AI mindset. If you measure people by output volume or raw speed, you penalize depth, learning, and the human edge. According to the PwC 2026 AI Jobs Barometer, AI-exposed tasks now demand 2.5x more empathy, judgement, and creativity, with junior roles showing a 7x jump in demand for senior skills. This is not just a shift in language—it's a shift in value. Your review criteria must match it or you will reward the wrong things, and the right behaviours will go elsewhere.

The working model

A good AI-era review rubric answers two questions:

  • What does high-quality human contribution look like in AI-assisted work?
  • How can we observe, describe, and score it fairly—without over- or under-crediting the AI?

Quality checklist

Behaviours are observable in real work (not abstract qualities)

Each level (high/medium/low) is clearly and specifically described

The rubric focuses on human direction, not just the final output

Tested on at least one real work sample for clarity

Notes identify where the rubric gave insight and where not

Common mistakes

Leaving rubrics vague or generic (e.g. 'good judgement')

Focusing only on final deliverable, ignoring how AI was directed and checked

Using the same statements for multiple roles without calibration

Over-complicating: too many categories or too much detail to use practically

Skipping the real-work sample test and only writing the rubric in theory

Checkpoint

Can you now write and apply a behaviourally anchored rubric for any AI-assisted role on your team?

Exercise

Rewrite a Review Rubric for AI-Assisted Work

Steps
  1. Pick one role or deliverable from your team that now relies on AI for core output.
  2. For Judgement, Verification Quality, and Originality, write three levels (high/medium/low) of behaviourally specific review statements for that role. Use real phrases and work signals you’ve seen.
  3. Apply your rubric to a recent, real work sample. Score it—then note where your rubric description gave clarity or confusion.
  4. Capture your rubric and notes as your working output.

Use this at work tomorrow

Pick your team's last AI-assisted deliverable and draft a three-skill rubric (judgement, verification, originality) before your next review.

03

Evaluate the Hybrid: Manager Guide to AI-Assisted Work

Managers learn to evaluate and credit real human input in AI-supported outputs—moving beyond surface-level prompt use to properly assess judgement, verification, and originality.

A warm-lit, editorial tabletop with four small, translucent gates or standing dividers arranged along a path, each with an orange-tinted card placed in front of them. The path curves past stacked, nondescript files or sheets, visually representing checkpoints of review. No people, no text, no digital screens. Negative space and material contrasts highlight where human judgement checkpoints are built into the process.
Four review gates—physical dividers and evidence cards—make visible the points where managers isolate and credit human direction, not just file output.

Why this matters in the workflow

Your team’s work is shaped by AI. Volume and polish don’t reveal the real contributor anymore. As the PwC 2026 AI Jobs Barometer showed, judgement, empathy, and originality are now 2–7x more important in AI-exposed roles—but most review processes can’t see past the surface result. When you run a 1:1 review or give formal feedback, you need to cut through the noise: what did the human actually direct, check, or invent?

The working model: Human Direction Over Routine Prompting

Reviewing hybrid AI-human work means inspecting how the person:

  • Directed the AI to shape the real outcome (not just ran a prompt)
  • Verified accuracy and risks, rather than assuming the tool is right
  • Added unique judgement or creativity—not just the first answer the tool gave

This is the difference between a manager spotting a spark, and rubber-stamping template output.

Quality checklist

Describes at least one clear human intervention (direction, verification, or creative input)

Evidence is specific—shows what the person did, not just what the tool produced

Originality or judgement is shown in a concrete decision, insight, or improvement

No over-crediting of routine prompting or minor edits

Review note links evidence to real value in the outcome

Common mistakes

Crediting only visible AI output, glossing over human input

Failing to distinguish between simple prompt use and true direction

Overstating originality when no evidence is shown

Skipping the ‘walkthrough’ and guessing at process

Writing vague review feedback that doesn’t specify actions

Checkpoint

Can you point to a specific place in an AI-assisted work product where human input shaped, verified, or improved the outcome? If not, repeat the guided review.

Exercise

Spot Real Judgement in an AI-Assisted Output

Your 15-Minute Task

Goal: Use the manager evaluation guide to review a real, recent AI-assisted deliverable from your team. Capture evidence of direction, verification, and originality.

Steps:
  1. Pick a recent deliverable produced with AI assistance (e.g., report, campaign, analysis, code).
  2. Ask the contributor to walk you through their process briefly—what steps did they take, what decisions did they make?
  3. Use the guide (template below) to note:
  • Direction: Where was human framing or prompting key?
  • Verification: Where was output checked, challenged, or improved?
  • Originality: Where did unique insight or creative change appear?
  1. Write a short review note using your findings: call out one place where the contributor showed real judgement (not just routine prompting).
  2. Share the note with the contributor in your next feedback meeting or message.

By the end, you’ll have a real work example evaluated through a modern lens.

Use this at work tomorrow

Use the review note structure with your next AI-assisted deliverable—ask for the process, not just the file.

04

Promotion in Practice: Raising the Bar for Judgement

Redesign your promotion pathways to prioritize growth in judgement and responsibility—ensuring junior team members build, and are recognized for, genuinely higher-order skills in the age of AI-assisted work.

On a clean, sunlit table: a small, neat stack of heavyweight orange and white cards marked with abstract dots or check marks, alongside a faded, thinner set of cards pushed aside. There are no words or numbers. The prominence of the new stack signals readiness and rigor; the faded cards signal what’s being left behind. No people, no hands, no device screens. Editorial composition, restrained palette, strong natural light.
A checklist for promotion: new skill markers surface on thick cards, while a separate pile of faded, thin cards set aside represents retired criteria—drawing focus to the importance of evidence, not output count.

Why this matters in the workflow Promotion criteria are the engine of growth. In AI-heavy work, old ladders don't fit. When speed and prompt fluency dominate, junior team members risk becoming eternal AI wranglers—fast, compliant, but stalled. PwC’s 2026 AI Jobs Barometer: AI-exposed junior roles are seven times more likely to demand leadership and strategic thinking now. The mistake: continuing to promote for quick outputs instead of deeper human skill.

The working model A modern promotion checklist is anchored in evidence of judgement, verification, original thinking, and contribution to team or business direction—not just volume or prompt tricks. Candidates for advancement must show they can:

  • Design and troubleshoot their workflow, not just execute given prompts.
  • Challenge AI outputs, credit and explain improvements, and take accountability for results.
  • Spot risks, escalation moments, or teamwide opportunities—signalling readiness for wider scope.

How to apply it Promotion review now means:

  • Reviewing AI-assisted deliverables with the candidate, surfacing where they improved, verified, or re-directed AI output.
  • Collecting evidence of peer influence, process improvement, independent verification, and original framing—not just time saved or content produced.
  • Documenting gaps: If a team member only cycles prompts, this is support—not promotion—material. Growth gaps become coaching, not skipped checkboxes.

Example on live work Case: A content associate whose briefs are AI-generated.

  • Old logic: Promoted when output is frequent, error-free, on time.
  • New logic: Promoted only if, on actual projects, the associate refines the AI’s structure, spots output errors, proposes new angles, and shares a verification checklist with the team. Evidence: Annotated versions, added brief sections, and peer feedback. Their coachable mistake surfaces: Inconsistent judgement or missing obvious fact errors.

Quality checklist

Each new criterion names a core human skill, not an output metric.

Observable behaviors are specific, practical, and visible in real work.

Checklist forces reviewers to ask for direct evidence or real examples.

Checklist surfaces gaps for coaching, not just pass/fail.

Retires at least one outdated volume-or-speed metric.

Common mistakes

Checking off time/volume boxes without seeing real judgement applied.

Using generic skill words ("good communicator") with no tied evidence.

Not asking for specific examples: templates filled, but real practice untested.

Defaulting to comfort: choosing criteria everyone already meets.

Checkpoint

Can you produce a promotion-readiness checklist for a real AI-exposed role, with clear evidence requirements for judgement, verification, and originality?

Exercise

Draft a Promotion-Readiness Checklist for One Role

Steps
  1. Choose a real, AI-exposed role (e.g., associate, junior analyst) up for potential promotion.
  2. List traditional promotion criteria currently in use.
  3. For each, write the new, AI-era skill that actually signals readiness: judgement, verification, originality, or direction.
  4. Draft 3-5 specific, observable behaviors that show the candidate demonstrates each skill in live work.
  5. Identify one piece of real deliverable evidence or feedback for each behavior.
  6. Review with a peer or manager: does this checklist genuinely raise the bar for judgement—not just output?
Use the template below to structure your checklist.

Use this at work tomorrow

Pick a junior team member's last AI-assisted deliverable and use this checklist to test for real promotion readiness—not just prompt fluency.

05

Check, Learn, Adapt: Measuring Success and Next Steps

Close the loop: assess if your new review and promotion criteria work in practice for AI-assisted roles, capture lessons, and decide what to change for the next cycle. Build a feedback snapshot and set up a simple improvement loop.

On a pristine, editorial table: a checklist with completed check marks, a separate blank survey sheet with empty circles, and a beam of orange light crossing both, forming a subtle arrow pointing off the surface. No visible people, handwriting, text, numbers, devices, dashboards, or logos. Composition focuses attention on the tracking and feedback objects, with negative space and warm, directional light.
A split workspace: checked checklist and a blank feedback survey rest side by side, illuminated by an arrow of light signaling the path to next steps—a concrete close to an adaptive review cycle.

Why this matters in the workflow

You revised your review criteria. You ran the new process. The real test: did it work? Measurement is where bold intentions either get lost or help you course-correct. AI-assisted work is new terrain. Teams need proof that new review practices increase clarity, reward the right skills, and don’t just reshuffle confusion. Without a tight feedback loop, outdated habits creep back or new mistakes quietly take root.

The working model

End-of-cycle review isn’t paperwork. It’s the rapid learning engine for your people system. You’re looking for three things:

Quality checklist

Checklist is filled honestly—not all 'yes' unless real examples support it.

At least one example per checklist item.

Survey responses include at least one specific, non-generic insight from reviewer or team member.

Summary notes two changes and one lesson that ties to review quality or clarity.

Next step/action is concrete and relevant to the observed lesson.

Common mistakes

Filling the checklist with 'yes'/'no' only—no examples or evidence.

Skipping the team member/reviewer survey—missing real perceptions.

Summarizing vague 'improvements' without stating concrete changes or lessons.

Treating the process as compliance, not a real source of feedback to improve next cycle.

Ignoring open issues or not planning a next step for iteration.

Checkpoint

Did your checklist and survey reveal at least two actual changes and one lesson or open issue you can act on next cycle?

Exercise

Run a Rapid Review: Your End-of-Cycle Checklist and Survey

Steps (aim for 15 minutes)
  1. Download or copy the template below.
  2. Complete the checklist: For your most recent review cycle, note which items are fully met, partly met, or not met, and jot one example per case.
  3. Send the survey questions by email or chat: Pick 2 reviewers and 2 team members (or more if possible).
  4. Summarize: Write down two observed changes and one open lesson/issue.
  5. Journal a next step: What needs to improve or be clarified before the next cycle?
Output

A completed checklist—plus a paragraph summary of two changes, one lesson, and the next planned action.

Template below you can copy-paste into your note, Slack, or doc.

Use this at work tomorrow

Copy the checklist into a staff note, fill it out, and share your team’s two changes and one next step for the next review cycle.

30-day path

Week 1: Diagnose the current state and map new skill demands.

Week 2: Draft and calibrate the revised performance-review rubric.

Week 3: Train managers and test the AI-assisted work guide in real reviews.

Week 4: Roll out the new checklists and criteria in promotion cycles.

End-of-month: Complete the checklist, collect feedback, and set up retrospective for iteration.

Success signals

Traditional criteria replaced by new, skill-focused criteria in 100% of templates.

Managers and reviewers correctly apply the rubric in at least 80% of calibration cases.

75% of team members report increased clarity on what 'good' looks like in AI-assisted work.

All promotion cases reference evidence from revised criteria.

Observable reduction in over-crediting AI output alone in review feedback.

Reflection prompts

Where does this topic show up in real work?

What behavior should change first?

What evidence would prove this Riseplan worked?

Manager checklist

Choose one owner for the behavior change.

Use the exercise on live work.

Review the output before scaling the habit.

Decide what changes after 30 days.

Want this shaped around your company?

Risey can research your company foundation first, then build a version of this path around your real workflows, customers, and culture.

Start with your company