AI economics, plainly
How much does AI actually cost?
For Omaha mid-market leaders. The real cost ranges, what drives them, and the line items most companies forget until they show up on the invoice.
Text Rosey · Schedule a call →Definition
AI cost has three layers, and most cost surprises come from the second and third.
**Layer 1 — Direct tool cost.** Enterprise tier of a major AI assistant runs roughly $20–$60/seat/month for ChatGPT Enterprise, Microsoft Copilot for Business, or Claude Enterprise. For a 100-person mid-market company with broad rollout, that's $24K–$72K/year. For specialized tools (call summarization, marketing copy, code assistance), add $5–$50/seat/month per tool. Most companies underestimate by counting only the first tool.
**Layer 2 — API and consumption cost.** When AI is embedded in custom workflows (claims processing, RFP drafting, document summarization), the cost shifts from per-seat licenses to per-token API consumption. Variable, hard to predict at first, often surprising. Budget for 50–200% of your initial estimate while you tune; converges to a steady state once usage stabilizes.
**Layer 3 — Governance and change-management overhead.** Often the largest line item, almost always invisible at procurement time. AI use policy drafting, approved tool list maintenance, training rollout, attestation tracking, audit trail infrastructure, manager enablement, IT/Security review per new tool. For a mid-market rollout, this typically runs $50K–$250K in the first year — outsourced or as internal time.
Why it matters for Omaha companies
Deloitte's Q4 2024 GenAI survey found 78% of organizations expect to increase their overall AI spending in the next fiscal year. That spending is real and growing — but the ROI gap is the bigger story. McKinsey 2025 found only 39% of organizations report any EBIT impact at the enterprise level from AI, and only ~6% qualify as 'AI high performers'.
The gap between spending and impact comes mostly from layer 3 being underfunded. Companies that budget for tool licenses and API costs but skip the governance and change-management investment end up in the 92% — adopting but not transforming. Companies that fund all three layers tend to land closer to the 6–7% who get strategic impact.
For Omaha mid-market specifically, the line items that surprise most often: BAA negotiation costs (healthcare), IT/Security review of each tool, AI use policy drafting and Legal review, training infrastructure, the audit trail that makes future regulator examination defensible.
Common follow-up questions
- What's a realistic budget for a mid-market AI rollout?
- For one wedge workflow shipped in 90 days with proper governance: roughly $80K–$250K all-in (consultant fees + tool costs + internal time at burdened rates). Enterprise-wide rollout over 12–24 months: $400K–$2M+ depending on regulatory complexity and number of workflows.
- Will it pay for itself?
- Tier 1 (one workflow): usually within 12 months if the workflow is high-volume and document-heavy. Tier 2 (multiple workflows): 18–24 months. Tier 3 (EBIT-level impact): McKinsey's data shows only ~6% of organizations achieve this, and the differentiator is workflow redesign, not spending.
- What's the biggest cost surprise?
- Almost always the governance and change-management work in layer 3. Most companies allocate 5–10% of the rollout budget to it; high performers allocate 25–40%. The math doesn't change — what changes is whether you pay it upfront or in failure-mode costs later.
- How does this compare to other tech rollouts?
- Closer to a CRM or ERP rollout than a SaaS subscription. The tool is the smallest part of the cost; the workflow redesign, training, governance, and ongoing tuning are the bigger parts. Companies that treat AI as a SaaS subscription end up disappointed; companies that treat it as a transformation program tend to capture the value.
- Should we wait for prices to come down?
- Token API costs have fallen ~10–20x since 2023 and will keep falling. Per-seat enterprise licenses have been roughly flat. The biggest cost component (governance and change-management) is labor and won't fall. Waiting saves money on the smallest cost component while the largest component continues to compound.
Sources
- 78% of respondents expect to increase their overall AI spending in the next fiscal year — Now decides next: Generating a new future — State of Generative AI in the Enterprise Quarter four, Deloitte AI Institute, 2025
- About 6% of organizations qualify as 'AI high performers' — those attributing 5%+ EBIT impact to AI — The state of AI in 2025: Agents, innovation, and transformation, McKinsey & Company (QuantumBlack, AI by McKinsey), 2025
- 39% of organizations report any EBIT impact at the enterprise level from AI — The state of AI in 2025: Agents, innovation, and transformation, McKinsey & Company (QuantumBlack, AI by McKinsey), 2025
Related
→ Start here
Text Rosey to begin.
Rosey is our executive-assistant bot. Text the number below — she'll ask two questions, offer three calendar slots, and put a 30-minute call on Jim's calendar.
Text Rosey · Schedule a call →