What I Learned at the AI Briefing — and What It Means for Us

Executive Summary

• AI produces real, sustained value at scale — UPS saves $400M/year, JPMorgan prevents $1.5B in fraud, IKEA generated $1.4B in new revenue by reskilling rather than cutting. These are not pilots. They are multi-year, audited results.
• Only 5% of companies capture this value. The other 95% spend the money and abandon the work within 18 months. The difference is not the technology — it is how the organization deploys it.
• The path from where most mid-market companies sit today to the 5% is a 90-day sequence: audit what already exists, deploy on the right tasks with honest cost modeling, and define success metrics before spending.

Three Things Worth Knowing

1. AI works — but 95% of companies deploy it wrong

The proof that AI delivers returns is no longer in question. UPS has compounded $400M/year in savings from ML route optimization for over a decade. JPMorgan runs 400+ AI use cases preventing $1.5B in annual fraud losses. Citi reached 70% AI adoption across 182,000 employees through a peer champions network — no mandates, no forced rollouts.

The failure rate is equally well-documented. MIT’s study of 300+ deployments (August 2025) puts the generative AI pilot failure rate at 95%. McKinsey (n=1,993, July 2025) finds only 6% of organizations show EBIT impact above 5%. S&P Global (n=1,006, March 2025) reports 42% of companies abandoned most AI initiatives in 2025, up from 17% the prior year.

The 5% that succeed do three things differently:

2. The real cost is multiples of the subscription — and that number is manageable if you plan for it

A 10-person team pays $8,400/year in AI tool licenses. The actual Year 1 cost — debugging AI-generated output, additional code review time, training, governance, and process redesign — runs roughly 2.5x the license (DX Research/Atlan, 2025). License fees represent 10-20% of the total investment.

This sounds alarming until you see BCG’s formula: companies that allocate 10% to algorithms, 20% to technology and data, and 70% to people and process achieve 1.7x revenue growth and 3.6x total shareholder return (BCG, n=1,250+, September 2025). The cost is real. The return is also real — but only for organizations that budget honestly from day one.

For a company with 200-500 employees, the realistic Year 1 investment for a targeted AI deployment is $140K-$420K, inclusive of everything. That number should appear in the business case before the first purchase order.

3. There is a 90-day path — and it does not start with buying technology

The briefing outlined a specific sequence:

Weeks 1-2: Audit. Run a shadow AI audit. Employees are already using AI tools — 78% through unapproved channels (WalkMe, n=1,000, 2025), 77% pasting company data through personal accounts (LayerX, October 2025). The audit reveals 3-5x the expected tool footprint and quantifies the security exposure. This takes two weeks, not two months.

Weeks 3-4: Decide. Pick one workflow for the first deployment — not three, not five. Write a two-page acceptable use policy. Build the honest cost model. Organizations with pre-defined success metrics achieve 54% pilot success versus 12% without (Pertama Partners, n=2,400+ initiatives, 2026).

Weeks 5-8: Pilot. 15-25 people, mixing enthusiasts and skeptics. Train through internal champions (1 per 10-20 people), not mandatory webinars. Expect an adoption dip at weeks 3-5 — this is documented and normal.

Weeks 9-12: Evaluate. Compile results into a business case with real numbers: what was spent, what was measured, what the recommendation is for scale. The pilot either proved it or it did not. Either answer is valuable.

Key Data Points

What This Means for Your Organization

This one-pager is designed to travel. The executive who attended the briefing cannot relay 30 minutes of data and nuance in a hallway conversation — but a board member or CEO who reads the three arguments above has the essential picture: AI produces proven value, most companies fail to capture it for identifiable reasons, and a disciplined 90-day path exists to get from where most mid-market organizations are today to the 5% that succeed.

The decision this document supports is not “should we invest in AI.” That question is already settled — employees are using AI tools whether or not the organization has sanctioned them. The decision is whether to deploy AI deliberately, with honest cost modeling and defined success criteria, or to let shadow adoption continue accumulating risk without capturing value.

If the person reading this wants the full 90-day plan, the shadow AI audit framework, or the 25-question organizational assessment referenced in the briefing — brandon@brandonsneider.com is the fastest path to those materials.

Sources

1. BCG — “AI at Work 2025.” n=10,600+ workers, 11 countries. Only 5% of organizations achieving substantial AI gains. Independent survey. High credibility. https://www.bcg.com/publications/2025/ai-at-work-momentum-builds-but-gaps-remain
2. McKinsey — “The State of AI in 2025.” n=1,993. June-July 2025. EBIT impact and adoption data. Independent survey. High credibility. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
3. BCG — “From Potential to Profit.” n=1,250+, September 2025. 10-20-70 framework and revenue/TSR data. Consulting survey. Moderate-high credibility. https://www.bcg.com/publications/2025/closing-the-ai-impact-gap
4. S&P Global 451 Research — n=1,006, March 2025. Initiative abandonment rates. Independent. High credibility. https://www.spglobal.com/market-intelligence/en/news-insights/research/ai-experiences-rapid-adoption-but-with-mixed-outcomes-highlights-from-vote-ai-machine-learning
5. Faros AI / AlterSquare — 10,000+ developers, 1,255 teams (Faros); 20+ client deployments (AlterSquare). True cost analysis. Practitioner data. Moderate credibility. https://www.faros.ai
6. Pertama Partners — n=2,400+ AI initiatives, 2025-2026. Success metrics and C-level sponsorship data. Independent. High credibility. https://www.pertamapartners.com/insights/ai-project-failure-statistics-2026
7. WalkMe/Propeller Insights — Shadow AI survey. n=1,000 U.S. workers, July 2025. ±3% margin of error. Independent polling. High credibility. https://news.sap.com/2025/08/new-walkme-survey-shadow-ai-rampant-training-gaps-undermine-roi/
8. LayerX — Enterprise AI & SaaS Data Security Report. October 2025. Browser telemetry. Independent security vendor. High credibility. https://layerxsecurity.com/blog/layerxs-enterprise-genai-security-report-2025-exposing-hidden-ai-security-blind-spots/
9. IBM — Cost of a Data Breach 2025. n=604 organizations, 17 countries. Gold standard for breach cost data. https://www.ibm.com/reports/data-breach
10. UPS — Investor presentations, 2015-2025. ORION route optimization. Public filings.
11. JPMorgan — Reuters, 2025. 400+ AI/ML use cases. Public reporting.
12. Citi — Chief Information Officer public remarks, 2024-2025. 182,000-employee champions network.
13. IKEA — Ingka Group public reporting, 2023-2025. Call center reskilling and revenue data.

Brandon Sneider | brandon@brandonsneider.com March 2026