The CFO's question: how much does it cost and how much does it return

Enthusiasm for AI does not convince a CFO. Numbers do. This module gives you the tools to answer with data, not opinions.

Three benchmark figures you need to know:

• Average AI spend per employee (2026): $2,068/year. Top 10% spends >$2,800. Bottom 50% spends <$200.
• Average ROI of AI training: 3.70x per EUR invested (Fed Reserve Atlanta 2025).
• Average time savings: 2h/day per employee actively using AI (McKinsey 2025).

These figures are your starting point. Your business case must customize them with your company's real numbers.

ROI formula for executives

Basic formula

Annual ROI = (Value generated - Total investment) / Total investment x 100

Value generated = Hours saved x Average hourly cost x 12 months
                + Additional revenue attributable to AI
                + Costs avoided (errors, fines, turnover)

Total investment = AI licenses x 12 months
                + Training (one-time or annual)
                + Adaptation time (first month)
                + Management and support (monthly)

Example with real numbers

Company: 30 employees. 15 actively use AI.
Tools: Claude Team ($25/person/month)

VALUE GENERATED:
- Time savings: 15 people x 1.5h/day x 220 days x EUR 35/h = EUR 173,250/year
- Quality: 20% fewer document errors = ~EUR 8,000 (conservative estimate)
- Speed: proposals 3x faster = 2 extra deals/year x EUR 15,000 = EUR 30,000
Total value: ~EUR 211,000/year

INVESTMENT:
- Licenses: 15 x $25 x 12 = $4,500 (~EUR 4,200)
- IAcademy training: 15 x EUR 399 (Starter) = EUR 5,985
- Adaptation: 15 x 10h x EUR 35 = EUR 5,250 (one-time, first month)
- Management: 2h/week x EUR 50 x 52 = EUR 5,200
Total first-year investment: ~EUR 20,635

ROI = (211,000 - 20,635) / 20,635 x 100 = 922%
Multiplier: 10.2x
Break-even: month 2

Prioritizing initiatives: the impact-effort matrix

You cannot do everything at once. The 2x2 matrix helps you prioritize:

Quadrant 1: Quick Wins (high impact, low effort)

Do first. Example: automate follow-up emails (immediate savings, 1-hour setup).

Quadrant 2: Strategic projects (high impact, high effort)

Plan ahead. Example: AI for M&A due diligence (major savings, requires integration and training).

Quadrant 3: Incremental improvements (low impact, low effort)

Do if there is time. Example: AI for document formatting (saves some time, easy to implement).

Quadrant 4: Discardable (low impact, high effort)

Do not do. Example: building a custom chatbot when an off-the-shelf solution solves the problem.

How to score

For each AI initiative, score:

IMPACT (1-5):
- Hours saved/week (1: <2h, 5: >20h)
- People benefited (1: 1-2, 5: >20)
- Effect on revenue/costs (1: marginal, 5: >10% impact)

EFFORT (1-5):
- Implementation time (1: <1 week, 5: >3 months)
- Technical complexity (1: plug-and-play, 5: custom development)
- Organizational change (1: 1 person, 5: entire company)

Score = Impact / Effort. Higher score = higher priority.

AI budget: how much to invest

Benchmarks by size

The 1-3% rule

As a starting point: invest between 1% and 3% of your gross payroll in AI. If payroll is EUR 2M, your AI budget should be between EUR 20,000 and EUR 60,000/year. This includes tools, training and management.

Where NOT to cut

1. Training: it represents 3-6% of the AI budget but generates 70% of the value (because without training, tools go unused).
2. Enterprise plans: the difference between a free plan (data used for training, no control) and a Team plan (data isolation, admin panel) is $25/person/month. The cost of a data breach is 10,000x more.

Board presentation template

BOARD PRESENTATION: AI STRATEGY [YEAR]

SLIDE 1: Context
- % of competitors already using AI
- Regulatory obligations (AI Act)
- Cost of inaction (competitive gap)

SLIDE 2: Opportunity
- Top 3 highest-impact areas in our company
- Potential savings in hours and EUR (base scenario)
- Qualitative benefits (quality, speed, innovation)

SLIDE 3: Proposal
- Recommended tools and why
- Training plan (IAcademy Base Program + Specialization)
- Governance: committee, policy, responsible party

SLIDE 4: Investment and ROI
- Detailed budget (tools + training + management)
- ROI in 3 scenarios (pessimistic, base, optimistic)
- Break-even (estimated month)

SLIDE 5: Roadmap
- Phase 1 (months 1-2): pilot with 1 department
- Phase 2 (months 3-4): scale to 2-3 departments
- Phase 3 (months 5-6): entire organization
- Tracking KPIs

SLIDE 6: Risks and mitigation
- Privacy: enterprise plans + policy
- Regulatory: AI Act compliance + training
- Adoption: champions + structured training
- Decision: approval requested [amount] EUR

AI metrics dashboard for leadership

6 metrics the executive committee should review monthly:

1. Adoption rate: % of employees using AI at least 3x/week. Target: >70% at 6 months.
2. Hours saved: total hours saved by the organization. Convert to EUR.
3. Cumulative ROI: value generated vs investment to date.
4. AI incidents: number of incidents (data, bias, critical errors). Target: 0 serious.
5. Satisfaction: quarterly team survey (1-5). Target: >3.5.
6. Compliance: % of employees with completed training (AI Literacy). Target: 100%.

5 traps when calculating AI ROI

1. Counting only time savings. AI also improves quality, reduces errors, accelerates revenue and prevents fines. If you only count time, you underestimate 40% of the value.
2. Ignoring the cost of not adopting. Your competitors are adopting. The gap widens every quarter. The cost of "waiting" is not zero.
3. Assuming 100% adoption. In practice, 70-80% of the team will adopt well. The remaining 20-30% will need more time. Calculate with 70% adoption.
4. Not including adaptation costs. The first month has a learning curve. Each person invests ~10h learning. That cost is real but one-time.
5. Presenting a single scenario. Always 3: pessimistic, base, optimistic. The CFO will distrust a single perfect number.

Case study: 80-employee company

Professional services firm (consulting), 80 employees, regulated industry.

Situation

• 50 people do knowledge work (reports, proposals, analysis)
• Average hourly cost: EUR 45 (including employer contributions)
• Shadow AI detected: 35% of the team uses free ChatGPT with client data
• No AI policy, no training, no governance

Proposal

• Tool: Claude Team (50 licenses x $25/month = $15,000/year)
• Training: IAcademy Enterprise (EUR 799/person x 50 = EUR 39,950)
• Governance: AI Lead at 20% (no additional cost, role reassignment)
• Total first year: ~EUR 54,000

Estimated return (base scenario, 75% adoption)

• 38 people save 1.5h/day x 220 days x EUR 45 = EUR 564,300
• Shadow AI elimination: risk avoided (not quantifiable but real)
• AI Act compliance: training as evidence (avoids potential sanction)
• ROI: 564,300 / 54,000 = 10.4x
• Break-even: month 2

Practical exercise

Puntos clave