Building an AI Business Case

Why Most AI Business Cases Fail to Get Approved

Leadership rejects AI business cases for three reasons, almost without exception: the numbers aren't credible, the risks aren't addressed honestly, or the ask isn't clear. Most AI enthusiasts make all three mistakes simultaneously — they oversell the upside, wave away the risks, and ask for resources without specifying what they'll do with them.

A business case that gets approved does the opposite. It uses conservative estimates that leadership will trust. It names the risks and shows that you've thought them through. And it makes a specific, bounded ask with a clear definition of success.

Quantifying Time Savings: The Right Way

Time savings are the most reliable ROI driver for most AI business cases, and the easiest to calculate. Here is the method:

The Risk Assessment: Name It Before They Do

Leadership's biggest concern with AI is risk. If your business case doesn't address risk proactively, they will invent their own risk scenarios — usually worse than reality. The manager who names the risks, explains them honestly, and describes the mitigations looks like the person who should be trusted with AI. The manager who doesn't mention risk looks naive.

For each AI initiative, assess four risk categories:

• Accuracy risk: What happens when the AI gets something wrong? Who catches it? What is the blast radius of an error? For low-stakes tasks (meeting notes, internal drafts), errors are low-cost. For customer-facing or compliance-sensitive tasks, errors can be expensive.
• Data risk: What data does the AI process? Who has access to AI outputs? Does using this tool create any data privacy or security exposure?
• Adoption risk: Will your team actually use this? What is the resistance level? What happens to the AI investment if 60% of your team refuses to change their workflow?
• Dependency risk: What happens if the vendor changes pricing, shuts down, or the product degrades? How locked in are you?

For each risk, write one sentence on the mitigation. "Accuracy risk: all AI outputs will be reviewed by [role] before any external use." That is sufficient. You do not need a 20-page risk management plan — you need to show you've thought through the obvious concerns.

The One-Page Business Case Format

AI INITIATIVE: [Name of initiative]
Submitted by: [Your name/title]  |  Date: [Date]
Sponsor needed: [Who you need to approve this]

THE PROBLEM
[2-3 sentences describing the specific problem. Quantify it.
"Our team spends approximately X hours per week on [task].
This delays [outcome] and prevents us from [higher-value work]."]

THE PROPOSED SOLUTION
[1-2 sentences. Specific tool and use case. What the AI does,
not how it works.]

THE NUMBERS
  Current state cost:    $[X]/year in staff time
  Tool cost:             $[Y]/year
  Annual savings:        $[X-Y]/year
  Payback period:        [Z] months

  Calculation basis: [brief note — e.g., "Based on 5 people
  spending 3 hrs/week on [task] at $75/hr burdened cost,
  50% reduction assumption."]

KEY RISKS AND MITIGATIONS
  Risk 1: [Name] — Mitigation: [Specific control]
  Risk 2: [Name] — Mitigation: [Specific control]
  Risk 3: [Name] — Mitigation: [Specific control]

SUCCESS METRICS (90-DAY PILOT)
  Primary: [Specific measurable outcome]
  Secondary: [Adoption rate or quality metric]
  Go/No-Go criteria: [When would we stop the pilot?]

THE ASK
Budget:    $[Amount] for [X-month pilot]
Resources: [Specific personnel time required]
Timeline:  Start [date] | Evaluate [date] | Decide [date]

NEXT STEP
Approve pilot by [date] to [specific action that unblocks].

Getting Budget Approval: What Actually Works

The mechanics of approval vary by organization, but the psychology doesn't. Here is what works across virtually every type of leadership team:

Pre-sell before you present

Share a draft of your business case with your most AI-skeptical stakeholder before the formal presentation. Ask for their pushback. Then address it in the final version. You arrive at the meeting having already answered the hardest questions.

Start with a pilot, not a program

Asking for $500K for an enterprise AI transformation will get you a committee. Asking for $15,000 for a 60-day pilot will get you a signature. Every enterprise AI deployment that succeeded started as a small pilot that proved value. Design your ask to be the one that can say yes.

Use a reference that's already done it

If a peer organization, competitor, or respected company in your industry has deployed this tool successfully, mention it. Specifically. "Company X deployed this for their [team] and achieved [result]." Credible peer examples remove the "unproven" objection faster than any amount of your own argument.

Key Takeaways from Day 3

• The three reasons AI business cases fail: non-credible numbers, risks not addressed, unclear ask. Address all three proactively.
• Calculate time savings with the 9-step template. Use 50% reduction as your conservative assumption. Show all math.
• Name your risks before leadership does — accuracy, data, adoption, and dependency. One sentence of mitigation per risk is sufficient.
• The one-page format works because it informs first and asks second. Leadership approves what they understand, not what sounds impressive.
• Start with a pilot, not a program. Design the ask to be the one that can get a yes.