Roast & Rise

Riseplan from Roast & Rise

The AI Business Case: ROI, Budget, and Where Value Stalls

Tie every AI investment to real business results, budget sanity, and the proof that matters.

Most AI projects don’t show up on the numbers that count. This plan gives founders, finance, and operators the tools to cut through busywork, surface honest ROI, and make decisions with clear evidence—not just expert hope.

Editorial tabletop arrangement with unfilled business scorecards, orange review tag, folders, and dividers under soft light—evoking an honest, numbers-first AI funding workflow, no people present.
A clean, editorial still life: a set of blank business scorecards, a single vivid orange tag marking 'review required,' and a warm palette of documents, folders, and decision dividers on a spare, matte-warm surface. Unattended. No visible text, just clear business artifacts in gentle scene light.

Course thesis

Despite major AI investment, most organizations cannot prove business impact. Only 39% report enterprise-level EBIT outcomes, and just 6% achieve meaningful returns (McKinsey, 2026). Overruns, poor link to P&L, and noise from proxy metrics stall value. The only way forward is an honest, numbers-first workflow that kills busy pilots and funds what works, from pilot review through scale.

What you leave with

By the end, you'll build a rigorous scorecard connecting every AI initiative to a real financial metric, track budgets before overrun, and write a keep/kill/scale memo that stands up to scrutiny.

For

Founders, finance leads, and operators who approve, fund, or assess AI projects after pilots are live.

Workflow

Review and decide ongoing AI project funding post-pilot with a business-level scorecard, budget tracker, and keep/kill/scale process.

Change

Move from funding AI on activity and stories to funding AI on tracked, defensible business value and budget discipline.

What you can do

Use these as checks while you move through the plan.

Diagnose where AI value stalls: EBIT vs proxies

Tie every AI initiative to a live business metric, not just technical output

Build and use an AI budget-vs-actuals tracker

Distinguish pilots that create margin from ones that just run

Produce and present a keep/kill/scale decision memo

Run a transparent scorecard review with your executive team

Chapters

01

Seeing the Value Gap: What Gets Measured, Moves (or Doesn’t)

Diagnose why most AI projects quietly miss real business impact, and systematically map each current project to the numbers that matter—P&L, margin or EBIT—using a clear value gap worksheet.

Editorial scene with index cards in soft green, yellow, and red borders arranged on a neutral workspace, each group separated by clear dividers, expressing the diagnosis of value link for AI projects, zero people, no words.
Color-coded cards representing AI projects are placed across a neutral, divided workspace—some grouped with green, yellow, or red edges, visualizing their link to real financial outcomes. The arrangement spotlights value linkage and exposes the gap.

Why this matters in the workflow

The board wants results. Operators want proof. But the reality—backed by McKinsey’s 2026 State of AI global survey—is painful: only 39% of organizations see any EBIT impact from AI. Nearly two-thirds of companies never scale past pilots. Over half blow their infrastructure budgets by 40%--or more. The reason is upstream: we measure the easy proxies—like data quality or technical wins—instead of mapping every AI project to P&L, margin, or EBIT outcomes. Proxies feel like progress but rarely move the number you’re paid to move.

You need to surface which AI initiatives truly link to financial value, and flag the rest. The Value Gap Diagnostic Worksheet is your first step. It turns a fuzzy list of AI projects into an honest map: what’s moving real numbers, and what is noise.

The working model

Quality checklist

All live AI projects are listed (none skipped, no cherry-picking)

Each project is mapped to a financial outcome metric, not just a technical one

Value status is assigned using McKinsey criteria: green = tracked P&L impact, yellow = plausible but untracked, red = proxy only

Rationales are brutally honest, concise, and not wishful

At least one relevant reviewer (finance/op leader) has reviewed for blind spots

Common mistakes

Mapping projects only to technical KPIs, skipping business outcomes

Marking 'yellow' or 'green' on optimism rather than data

Missing projects that quietly drain value or budget

Leaving rationales vague or political

Failing to run the list by someone outside the project team for bias check

Checkpoint

Have you mapped every current AI project to a financial metric, color-coded their value link, and captured honest rationales? If yes, move on to building the live scorecard.

Exercise

Map your AI projects to real numbers: Value Gap Diagnostic in 15 Minutes

Get sharp. Pull up your actual AI project list—no skipping the hard or half-finished ones.

  1. List every current/live AI project or pilot. Work from a real project roster, not memory.
  2. For each, write its closest P&L/margin/cost/revenue metric. Target a number already tracked by finance, not a tech metric.
  3. Color-code status:
  • Green if its impact is tracked and shows up in live financial metrics
  • Yellow if you have a plausible connection but it’s not tracked (add a note: “needs validation”)
  • Red if there is no credible business linkage, just technical proxies
  1. Fill in simple rationale for each status—one honest line explaining why it’s green, yellow, or red.
  2. Share the worksheet with one CFO/finance or operations lead and ask for feedback or corrections.

You’ll finish with a clean gap map for the next funding review, plus surface where the business case is weakest.

Use this at work tomorrow

Pull your AI project list and run a 15-minute value gap diagnostic. Flag where evidence is thin before more money goes out the door.

02

The Honest Scorecard: Tying AI to Margins, not Metrics

How to build a scorecard that links each AI initiative to business results—not just activity or technical metrics—and drive evidence-based funding decisions.

Aerial view: an empty AI scorecard form, orange review marker on top, and colored status tabs beside it, arranged with austere neatness on a warm editorial surface, no text or people, all business.
An open, unmarked AI business scorecard template on a clean workspace, a single orange review token set on it. Close by are colored status markers—green, yellow, red—poised for assignment. The scene evokes focus and honesty in how AI value is tracked.

Why this matters in the workflow

Founders and operators are sick of dashboards packed with metrics that don’t move the number. Evidence: McKinsey’s 2026 State of AI report. 50% of enterprises track their AI on proxies like data quality and productivity. Only 39% show EBIT impact. That EBIT impact is small—under 5% for most. If your AI pilots aren’t visible in true margin, they’re noise, not value.

Steering meetings become autopilot rituals. Funding runs on the story—busy, complex, sometimes impressive. But the number doesn’t budge. A scorecard that forces linkage to EBIT or P&L brings the conversation to ground. Reduces “activity theatre.” Triggers hard decisions and clarity about owners, timing, and outcomes.

The working model

Quality checklist

Every initiative listed absorbs real resource or budget.

Each has a metric—the metric is tied to a business result, not a technical/process proxy alone.

P&L or EBIT linkage is active: either direct or a reasoned, traceable path.

Owner is a named person, not a functional group.

Review date is within 30 days—not open-ended.

Status is honest (reds don’t get buried to protect pilot ego).

At least one credible critique/review incorporated.

Common mistakes

Letting technical metrics stand without connecting to EBIT or margin.

Writing 'team', 'function', or leaving owner blank.

Review date set to 'ongoing' or left blank.

Rationalizing non-impact as 'still valuable' without numbers.

Downgrading status for optics instead of evidence.

Checkpoint

Can you show every AI project’s latest business result and trace it to the number your CFO or board actually cares about?

Exercise

Build Your Honest AI Scorecard (Live Project)

You have 15 minutes.

  1. Open a blank spreadsheet or document.
  2. List all active AI initiatives (pilots, proofs, or ongoing projects—if it absorbs budget, it’s on the list).
  3. For each, fill the following table:
  • Initiative name
  • Metric claimed to improve
  • Latest actual result (number/percentage/level)
  • P&L or margin link (how will this show up in EBIT/cost/profit?)
  • Owner (single accountable person)
  • Next review date
  1. Use red/yellow/green traffic light for each initiative:
  • Green: Clear business metric linked, measurable
  • Yellow: Proxy metric with plausible (but not certain) link
  • Red: Proxy metric or unclear owner or review point
  1. Share this draft with your CFO, finance lead, or operator for one pointed comment: “What here wouldn’t survive a board review?”
  2. Note their feedback—plan a 10-minute slot in your next AI review to fix reds, clarify owners, and chase real metrics.

Use this at work tomorrow

Draft your AI scorecard today and circle every project with a fuzzy or missing link to EBIT, cost, or revenue. Bring it to your next steering meeting for the first honest review.

03

Stopping the Spend: Budget Discipline Without Blind Overrun

AI infrastructure costs often spiral—58% of enterprises overrun their AI budgets by 40% or more (McKinsey 2026). This chapter shows how to set, track, and correct AI costs before they drain margin. Build a working budget-vs-actuals tracker that flags overruns, assigns cause, and signals stop points.

Stacked tracker sheets with colored chips indicating overruns and a crisp, orange stop marker set nearby, all arranged intentionally on a bare workspace. No people, text, or numbers.
A stack of budget tracker sheets—some neat, some with visible flagged markers—sits on a matte workspace next to a vivid orange stop token. Colored chips indicate overruns but no numbers or text are present. The scene is orderly but full of tension.

Why this matters in the workflow

Most AI budgets leak in silence. McKinsey’s 2026 survey found over half of enterprises spent 40% more than planned on AI infrastructure. Money keeps flowing long after value stalls. It shows up as cloud bills, surprise compute, extra contract hours, or sticky MVPs that never die. Funding through inertia is out. Budget review, every cycle, is in.

The working model

An honest tracker does three things:

  • Breaks AI costs down: infra, data, labor, vendors, platform, training.
  • Shows variance: what was forecast, what actually hit the P&L, and by how much.
  • Captures story: why the overrun (or underrun) happened, who flagged it, and whether to stop, reroute, or keep funding.

Quality checklist

Each active AI project broken into clear cost lines (not just totals)

Forecast and actuals reported in matching time windows

Variance calculated and color coded accurately per threshold

Red or yellow lines include both a short cause/owner and a next-step action

Tracker is fit for presentation: clear, readable, no buried lines

Common mistakes

Lumping costs by project only, hiding overrun details

Copying last month’s numbers without checking actuals

Marking everything green just to look good

Forgetting to assign ownership and next steps

Letting overruns slide with no documented response

Checkpoint

Can you show a live tracker with forecast, actual, variance, color, and at least one flagged overrun with cause and response for your AI projects?

Exercise

Build Your AI Budget vs Actuals Tracker

15-Minute Output
  1. List up to three current AI projects. For each, break spend into two or three main cost lines (infrastructure, vendors, labor, or other real costs).
  2. Fill in last month's or quarter's forecast and actuals for each cost line. If you don’t have a forecast yet, use an estimate.
  3. Calculate % variance for each line: `((Actual - Forecast) / Forecast) * 100`.
  4. Code each line: Green (within budget), Yellow (up to 15% over), Red (over 15%).
  5. For any yellow/red, write one line explaining the variance and assign an owner to review.
  6. Decide: keep, correct, or flag for stop—the response goes in your tracker.

Deliverable: a simple, readable table or spreadsheet matching the template below.

Extra: Share your version in your next finance or ops meeting for live feedback.

Use this at work tomorrow

Start a simple AI cost tracker. Flag your biggest overrun, write the cause in one sentence, and force a decision at your next review.

04

Kill, Keep, Scale: The No-Bull Decision Memo

Move beyond inertia. This chapter teaches you to write an evidence-driven memo that forces a clear decision: kill, keep, or scale each AI initiative based on ROI, real costs, and business impact—not sunk effort or momentum.

Still life of three document folders with colored tabs (green, yellow, red) and an orange envelope offset to the side, arranged on a clean, minimal surface; evokes written keep/kill/scale decisions with no text or human presence.
Three weighty, document folders marked only by colored tabs—green, yellow, red—sit atop a clean desk, a single orange envelope slightly offset, symbolizing the formal decision memo ready for dispatch. The scene embodies gravity and clarity.

Why this matters in the workflow

When funding dries up or pressure mounts, most AI pilots drift. Leadership falls for the sunk cost—"just one more quarter"—or lets old experiments limp along, eating budget and attention. According to the McKinsey 2026 State of AI survey, only 6% of orgs are 'high performers' who rigorously review investments, tie them to EBIT, and cut deadweight. Most just keep swimming and hope for impact that never lands.

The only way forward: a transparent, decision-ready memo for each AI initiative. Kill what doesn't work. Keep what's promising. Scale only what you can prove. No bull. No hiding busywork under jargon. Put this decision on paper and in the room.

The working model

Quality checklist

Memo includes specific business metric, not just proxy or technical targets

Decision (kill/keep/scale) is clear and justified by results, not hopes

Budget overrun or underspend explained briefly

Rationale names both strengths and what’s missing

Owner and next step assigned, with date for next review

Circulated to the full decision group, with feedback or sign-off recorded

Common mistakes

Leaving business metric blank or using only technical progress

Using generic rationale or avoiding specifics on outcome

Ignoring budget overrun or failing to note variances/explanations

Assigning no owner, actions, or next review date

Writing but not circulating—memos must be shared and discussed

Checkpoint

Can you show a circulated memo for at least one AI initiative, with a documented and defendable decision?

Exercise

Draft and Circulate a Real Keep/Kill/Scale Memo

Your task:

  1. Choose one ongoing AI initiative in your pipeline.
  2. Gather the latest scorecard, budget reports, and decision history.
  3. Write a keep/kill/scale memo, using the template structure below, focused on real business impact—not just progress or activity.
  4. Circulate the memo to your decision group (minimum: your direct exec or project sponsor). Request written feedback or sign-off.

You should produce:

  • A completed, honest decision memo ready for circulation.
  • A plan for the next funding or review meeting, with this decision memo attached as agenda material.

Use this at work tomorrow

Write a memo for your most ambiguous AI project and bring it to the next funding meeting—make the decision public.

05

Proving It Over Time: Scorecard Review and Course Correction

Build a repeatable, honest rhythm for reviewing (and correcting) your AI scorecard. Make course correction a habit, not a scramble. Tie every review to live business impact, documented decisions, and visible next actions.

Open review playbook with orange flags and action tokens arranged alongside, under cinematic light—encapsulating the ritual of scheduled review, no people or text present.
A recurring scorecard review playbook rests open, marked by layers of transparent orange page flags and adjacent tokens for 'action' and 'review.' The ensemble is illuminated with warm, scheduled light—suggesting methodical cycles and commitment.

Why this matters in the workflow

The only way to keep AI funding honest is to build a review habit. A single scorecard review won’t move the needle. Without a cycle, pilots drift; budgets balloon. McKinsey’s 2026 survey shows most AI projects slip into autopilot—fewer than half have ever been reviewed against a real outcome post-launch. The small group that outperforms keeps a live feedback loop. If you want to capture value, you need a routine: every 30/60/90 days, review what the number says, and act.

The working model

A Scorecard Review Playbook creates this cadence. It’s an agenda, a checklist, and a next-actions log in one. Every time, you:

  • Revisit the AI Business Scorecard for each project.
  • Surface wins, stalls, and overruns.
  • Take clear action: double down, kill, or fix.
  • Log ownership and the next check-in.

Quality checklist

All major AI projects reviewed, not just high-performers or favorites

Metrics updated (no backdated or estimated numbers)

P&L/margin linkage front and center for every project

Every decision backed by data from scorecard or budget tracker

Actions, owners, and deadlines documented and shared

Next review meeting scheduled and on calendar

Common mistakes

Forgetting to schedule the next review

Skipping difficult projects (review only 'winners')

Letting gut feel override the scorecard

Actions with no owner or clear follow-up date

Missing real P&L tie in discussions

Checkpoint

Can you show a completed review playbook with updated metrics, clear decisions, and next steps?

Exercise

Run Your First Live Scorecard Review

In the next 15 minutes:

  1. Choose 1–2 key AI projects with some operational history (no vaporware, no greenfield ideas).
  2. Download or copy the Scorecard Review Playbook template.
  3. Invite at least one decision-maker (yourself and 1+ founder, operator, or finance owner).
  4. Fill in the agenda using live scorecard data (see template below).
  5. Log at least one concrete next action, owner, and date per project.
  6. Send the output to all who influence funding decisions. Set a date for next review right now.

This output—your Playbook—is now a live tool: use it to drive every AI funding review. Update it each cycle. It will prove whether your projects are value creators or driftwood.

Use this at work tomorrow

Block a 30-minute meeting, fill the Playbook for your highest-spend AI project, and send your review summary to the exec team today.

30-day path

Run the Value Gap Diagnostic on all ongoing AI projects

Draft and complete the AI Business Scorecard with your team

Set up a budget-vs-actuals tracker for AI spend

Prepare and review your first keep/kill/scale memo with leadership

Schedule the first scorecard review meeting; log action items and revisit every 30 days

Success signals

All AI projects mapped to at least one P&L or EBIT metric (zero left on proxy metrics)

Scorecard adoption in team/exec review meetings

Time to flag and halt or reroute AI cost overruns shortened by at least 30%

At least one keep/kill/scale decision made per review cycle

Scorecard reviews scheduled and run at real intervals (not one-off events)

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