The pages on this site make claims about how AI implementation actually goes, why most engagements stall and what the work of getting it right looks like in practice. The findings below are the source material. If you want to verify something cited elsewhere on this site, this is where to find it.

The big picture

Six findings that frame how AI implementation actually goes, and why most engagements fail before the tools are even chosen.

95%
The implementation crisis

Approximately 95% of enterprise generative AI pilots fail to reach sustained production. Most organizations are stuck in pilot purgatory with no path to production value.

Source: MIT-backed analysis, Fast Company, April 2026†
Why it matters: AI capability is no longer the differentiator. Implementation is.
54%
The people crisis

54% of workers bypassed their company’s AI tools in the past 30 days and did the work manually. The trust gap between workers and executives sits at 52 percentage points.

Source: WalkMe State of Digital Adoption, survey of 3,750 executives and employees†
Why it matters: Tools alone do not create adoption. The work happens in the gap between what was deployed and what gets used.
42% / 20%
The strategy gap

42% of companies feel their AI strategy is highly prepared. Only 20% are equally confident about their talent readiness.

Source: Deloitte 2026 State of AI in the Enterprise†
Why it matters: A strategy that overlooks the workforce executing it is not a strategy. It is a plan in search of a problem.
$3 : $1
The ROI reality

Among companies that fully implemented a structured AI framework, McKinsey found AI delivered roughly three dollars in return for every dollar invested, with core profit up about 20% within two to four years. Top performers focused on three or fewer AI domains rather than deploying everywhere at once.

Source: McKinsey / Kate Smaje, via Business Insider, May 2026†
Why it matters: Focus beats breadth. The data on what works points consistently to discipline over scope.
+10% / +4% / +5%
The augmentation evidence

Across industries with high AI exposure between 2017 and 2024, productivity rose 10%, employment rose 4%, and wages rose 5%. AI used as a complement to human work drove employment up. AI used as a substitute slowed wage growth.

Source: Christos Makridis research, via Inc., May 2026†
Why it matters: The augmentation case is no longer theoretical. The data on exposure-mapped industries already shows the pattern.
90% / 1.5 hrs
The adoption gap

90% of executives in a 6,000-respondent NBER survey reported AI has had no measurable impact on employment or productivity at their companies. Average AI usage across the workforce sits at 1.5 hours per week.

Source: NBER survey, via Fortune, April 2026†
Why it matters: The gap between investment and usage is the gap that implementation work exists to close.

Case studies

Real companies. Real decisions. Real outcomes. The pattern across every case below points to the same lesson: implementation discipline is the variable that decides the outcome.

Vercel

Third-party AI governance failure.

Vercel was breached not through a software vulnerability but through an architectural gap. An employee had granted a third-party AI productivity tool full-access permissions to their corporate account. The tool was the entry point, and the breach followed.

One permission grant. Full corporate exposure.
Why it matters: AI governance is no longer about the AI. It is about the access permissions granted to AI on the way in.
Source: Fast Company†

PocketOS

Autonomous agent governance failure.

An AI agent at PocketOS encountered a problem during a routine task and autonomously deleted the company’s production database along with three months of backups in nine seconds. The agent then generated a written confession listing the safety rules it had violated.

Nine seconds.
Why it matters: Autonomy without bounded oversight produces a speed of loss that no human review process can match.
Source: TechRadar Pro†

Bank of America

AI as advisor amplifier.

Bank of America deployed an AI meeting-preparation tool to 18,000 financial advisors. The tool was framed and rolled out around advisor centrality, not advisor replacement. Q1 results: net income $8.6 billion, EPS up 25% year over year, with advisors saving roughly four hours per client meeting.

4 hours saved per meeting. 18,000 advisors.
Why it matters: When AI is positioned as advisor amplification, the adoption math works. The framing matters as much as the tool.
Source: Bank of America Q1 earnings reporting†

Klarna

Speed-to-layoff cautionary tale.

Klarna announced it had replaced 700 customer service agents with AI in a high-profile productivity move. Within twelve months, the company began rehiring human agents after customer satisfaction declined and unresolved inquiries rose. The CEO later reframed earlier statements about layoffs.

700 agents replaced. Then rehired.
Why it matters: Replacement looks good in the press release. The reversal looks worse.
Source: Fortune, Reworked†

IKEA

Reskilling over replacement.

IKEA deployed an AI chatbot that handled 47% of routine customer inquiries. Rather than reduce headcount, IKEA reskilled 8,500 customer service agents into interior design consultants. Reported revenue impact reached $1.4 billion. The company subsequently launched a 30,000-employee AI literacy initiative.

$1.4 billion. 8,500 reskilled. Zero layoffs.
Why it matters: The augmentation case at scale. AI handles routine. People handle the work that builds revenue.
Source: Forbes, HR Grapevine†

By the numbers

Nine data points across three categories. Each one points to the same conclusion: implementation discipline, not tool capability, is what separates results from spend.

Implementation reality

What actually happens when organizations try to deploy AI at scale.

54% / 25%

54% of organizations expect AI to reach production within three to six months. Only 25% have actually reached it. The gap between expectation and execution is widening, not closing.

Deloitte 2026 State of Generative AI in the Enterprise†
5% → 40%

In 2024, fewer than 5% of enterprise applications had task-specific AI agents integrated. Gartner projects that figure will hit 40% by the end of 2026. The change is fast. The integration debt accumulating along the way is faster.

Gartner forecast, via Inc., April 2026†
60%

60% of AI implementation efforts fail because they cannot integrate cleanly with legacy systems. The blocker is rarely the model. It is the architecture the model is asked to live inside.

McKinsey, via Lleverage AI, April 2026†

Workforce impact

How AI is actually showing up in the labor market and inside teams.

1.7pp

Top-quartile AI exposure industries contributed 1.7 percentage points to US productivity in 2025, roughly double the prior year. AI productivity gains are concentrating in the industries that built capability first.

Morgan Stanley, Chief Economist Gapen, via Fortune, April 2026†
13% / 2.3%

Of the 840 million jobs globally exposed to AI, Bank of America research finds 13% are being augmented by it and only 2.3% are being genuinely automated away. Augmentation is the dominant pattern. Replacement is the headline.

Bank of America Global Research, via Fortune, May 2026†
56%

56% of workers using AI tools received zero formal skills development from their employer to support that use. The training gap is the implementation gap, expressed at the individual contributor level.

ManpowerGroup 2026 Global Talent Barometer†

ROI evidence

What the return data actually shows when implementation is done with discipline.

4.4x

KPMG analyzed Total Shareholder Returns across more than 1,800 US public companies. Companies that accelerated transformation during the 2024–2025 uncertainty window delivered 4.4x the TSR of companies that paused. Discipline during turbulence compounds.

KPMG analysis of 1,800+ US public companies, via Fast Company, April 2026†
0

Goldman Sachs found no reliable correlation between the size of a company’s AI spending and the productivity gains it reports. Spend size is not the variable. Implementation quality is.

Goldman Sachs research, March 2026†
20%

20% of Boston Consulting Group’s own work was AI-related in 2024, up from a fraction of that two years prior. The professional services firms with the most ROI conviction are the ones already restructuring around AI as the work, not bolting AI onto existing work.

BCG internal practice composition, via Business Insider, April 2026†

Sources

The full source list behind every claim on this site. Every entry includes publication, date, and a link to the original where one is publicly available.

Academic research

Harvard Business School / Boston Consulting Group
“Navigating the Jagged Technological Frontier” — Field experiment with 758 consultants, 2024.
https://www.hbs.edu/faculty/Pages/item.aspx?num=64700
Foundational evidence that expert AI guidance produces 40% quality improvement; solo adoption produces 19% degradation. Cited across the site as the core implementation-quality proof point.
Massachusetts Institute of Technology
Project NANDA — Enterprise generative AI pilot analysis, 2025–2026. Reported via Fast Company, April 2026.
https://www.fastcompany.com/91342110/95-percent-ai-pilots-fail
95% pilot failure rate. Anchor citation for pilot purgatory framing across Services S9 and Insights Section 1 Card A.
Princeton University
Generative Engine Optimization (GEO) Research — 2025–2026 series.
https://arxiv.org/abs/2311.09735
Empirical evidence that source visibility in AI responses can be improved through deliberate optimization.
University of Texas / KPMG
Analysis of more than one million AI interactions, May 2026. Reported via Inc.
https://www.inc.com/
Sparring partner versus task delegator finding: only ~5% of users show sophisticated AI usage habits.
Anna Lenhardt / Fast Company
Organizational design analysis, March 2026.
https://www.fastcompany.com/
Strong people look average inside organizational design failure. Quoted on About and Fractional Overview.
National Bureau of Economic Research
6,000-respondent executive survey on AI workforce impact, via Fortune, April 2026.
https://www.fortune.com/
90% of executives report no measurable AI impact on employment or productivity. Average usage 1.5 hours per week. Insights Section 1 Card F.

Industry reports and surveys

Deloitte 2026 State of AI in the Enterprise
Annual research report, Q1 2026.
https://www2.deloitte.com/us/en/insights/focus/cognitive-technologies/state-of-ai-and-intelligent-automation-in-business-survey.html
Source for two cards: 74% chasing AI revenue / 20% finding it; 42% strategy-ready / 20% talent-ready. Used on Home and Insights Section 1 Card C.
McKinsey & Company
AI implementation framework analysis. Kate Smaje commentary, via Business Insider, May 2026.
https://www.businessinsider.com/
Core ROI evidence: roughly $3 return per $1 invested for fully implemented programs. Top performers focus on three or fewer AI domains. Anchor for Insights Card D.
McKinsey via Lleverage AI
Mid-market AI implementation costs analysis, April 2026.
60% integration failure rate with legacy systems. Cited in Services S4 and S5.
KPMG
Total Shareholder Returns analysis of 1,800+ US public companies. Via Fast Company, April 2026.
https://www.fastcompany.com/
4.4x TSR for companies that accelerated transformation during uncertainty. Used in Home page proof block.
Capgemini
Survey of 1,500 senior executives on AI agent deployment, April 2026. Via Inc.
https://www.inc.com/
Only 2% of organizations have deployed AI agents at full scale. Trust dropped 43% to 27% in one year. Lead citation for Services S2.
Boston Consulting Group
Internal practice composition disclosure via Business Insider, April 2026.
https://www.businessinsider.com/
20% of BCG work was AI-related in 2024. Used for Fractional differentiation framing.
Forrester Research
Enterprise AI decision-maker survey, March 2026.
https://www.forrester.com/
85% of decision-makers cannot show a positive return on AI investment. Anchor citation on Home page Quote Card 3.
Gartner
Enterprise AI agent integration forecast, via Inc., April 2026.
https://www.gartner.com/
40% of enterprise applications will integrate task-specific AI agents by end 2026, up from less than 5%. Services S1 and S2.
WalkMe State of Digital Adoption
Survey of 3,750 executives and employees, Q1 2026.
https://www.walkme.com/state-of-digital-adoption/
54% bypass rate; 52-point trust gap. Insights Section 1 Card B and Services S5/S7.
Edelman / LinkedIn
Joint study on B2B thought leadership trust, May 2026. Via Inc.
https://www.linkedin.com/business/marketing/blog/b2b-marketing/edelman-linkedin-b2b-thought-leadership
64% of buyers trust thought leadership content over marketing materials. About page support.
Semrush / LinkedIn
Joint study on AI citation rates across platforms, March 2026. Via Inc.
https://www.semrush.com/
LinkedIn cited in 11% of AI responses average / 14.3% on ChatGPT Search. AEO infrastructure case.
MBO Partners State of Independence
Annual workforce report, April 2026. Via Inc.
https://www.mbopartners.com/state-of-independence/
US independent workforce grew 78% from 41M to 73M (2019–2025). Fractional page market validation.
LinkedIn Jobs on the Rise 2026
Annual fastest-growing role report, April 2026. Via Business Insider.
https://www.linkedin.com/pulse/jobs-on-the-rise-2026/
AI consultants, strategists, engineers and researchers in the top five fastest-growing US roles. Fractional page authority lead.
Theory Ventures
Analysis of consulting partnership timing for AI-era startups, via Business Insider, April 2026.
https://www.businessinsider.com/
Partnership formation timing has shifted from $10M+ revenue to $2M–$5M. Fractional page competitive timing.
Vendux Fractional Leadership Report
68% year-over-year growth in fractional executive placements. The fractional executive market doubled between 2024 and 2025.
https://www.vendux.org/
68% YoY growth in fractional executive placements. Fractional Overview Section 1.
PwC 2026 CEO Survey
Annual global CEO survey, January 2026.
https://www.pwc.com/gx/en/issues/c-suite-insights/ceo-survey.html
Implementation as bottleneck data point. Foundational reference.
Bank of America Global Research
AI exposure analysis using ILO data. Via Fortune, May 2026.
https://www.fortune.com/
840 million jobs exposed globally. 13% augmentation versus 2.3% genuine automation. Workforce framing reference.
Morgan Stanley
AI productivity contribution analysis, Chief Economist Gapen. Via Fortune, April 2026.
https://www.fortune.com/
1.7 percentage points contribution to US productivity from top-quartile AI exposure industries.
JPMorgan Private Bank
AI capability versus reliability threshold analysis, Manoukian and Biemann. Via TheStreet, May 2026.
https://www.thestreet.com/
Reliability is the current ceiling, not capability. Used for executive briefing context.
Goldman Sachs
AI investment versus productivity correlation analysis, March 2026.
https://www.goldmansachs.com/
No reliable correlation between AI investment size and productivity gains. Implementation quality is the variable.
OpenAI Economic Research / Harvard / Deming
ChatGPT enterprise usage analysis. Via Inc., April 2026.
https://www.openai.com/
~80% of organizational ChatGPT usage falls into three basic categories. Maturity framing.
World Economic Forum
Future of Jobs Report 2025.
https://www.weforum.org/reports/the-future-of-jobs-report-2025/
170M jobs created, 92M displaced, net +78M by 2030. Macro-economic anchor.
ManpowerGroup Global Talent Barometer
2026 global workforce study.
https://www.manpowergroup.com/
AI usage and confidence inversion. Training crisis evidence.
PwC Fractional Executive ROI Analysis
Survey of CEOs who hired fractional executives, 2025.
https://www.pwc.com/
96% of CEOs report fractional ROI met or exceeded expectations. Services S7 and Fractional Model tile 1. Distinct from existing PwC 2026 CEO Survey entry.
Gartner Fractional Executive Forecast
Mid-market workforce projection, via Fortune, March 2026.
https://www.gartner.com/
More than 30% of midsize enterprises will retain at least one fractional executive by 2027. Fractional Overview tile 2.
Future of Consulting Report
Analysis of consulting firm restructuring around AI, March 2026. Publication name to be confirmed.
https://www.businessinsider.com/
Largest consulting firms rebuilding delivery around senior layer plus AI layer. “The pyramid is becoming an obelisk.” Fractional Overview tile 3.
McKinsey QuantumBlack
Enterprise AI implementation analysis, 2025.
https://www.mckinsey.com/capabilities/quantumblack
“Every organization has AI. One percent has figured out AI.” Fractional Model tile 3. Distinct from McKinsey/Smaje and McKinsey/Lleverage entries.
Morgan Stanley AI Team-Scale Analysis
Sam Altman commentary captured in Morgan Stanley research, 2026.
https://www.morganstanley.com/
“The organizational unit of the future is a small team with AI multiplying their output.” Fractional Model tile 2. Distinct from existing Morgan Stanley/Gapen productivity entry.
Bank of America Q1 2026 Earnings Disclosure
Earnings reporting and advisor productivity disclosures, April 2026.
https://newsroom.bankofamerica.com/
18,000 advisors deployed AI meeting-prep tool. Net income $8.6 billion, EPS up 25% YoY, four hours saved per client meeting. Insights Case Study Block 3. Distinct from existing BofA Global Research entry.

News and trade press

Gary Marcus, Fortune
Emeritus professor of psychology and neural science, NYU. Fortune column, April 2026.
https://www.fortune.com/
Anchor philosophical citation for the practice. Quoted on About hero and Fractional persistent intro.
Boylan / Avature / Gallup, via Inc.
“Coach, train, and reward” framing for AI-era workforce strategy.
https://www.inc.com/
Tier-1 positioning anchor. Locked. Same source set: only 9% of workers have broad AI expertise; structured upskilling increases retention 47%. Services S5 lead citation.
Inc.
Multiple 2026 features on AI agent deployment, AI usage habits, and workforce trends.
https://www.inc.com/
Primary publication for Capgemini, KPMG, MBO Partners, Edelman/LinkedIn, OpenAI/Harvard, BCG citations.
Fast Company
Multiple 2026 features on AI implementation, organizational design, and incident reporting.
https://www.fastcompany.com/
Source for MIT pilot data, Lenhardt commentary, KPMG TSR analysis, Vercel breach (Hoque).
Fortune
Multiple 2026 features on AI economic impact and executive perspectives.
https://www.fortune.com/
Source for Marcus column, Morgan Stanley productivity, BofA jobs analysis.
Business Insider
Multiple 2026 features on AI consulting market and venture timing shifts.
https://www.businessinsider.com/
Source for McKinsey/Smaje, Theory Ventures, BCG, LinkedIn Jobs on the Rise.
Harvard Business Review
“The Last Mile Problem” — March 2026.
https://hbr.org/
Implementation gap framing. Quote Card 2 on Home page (replaces earlier Reid Hoffman card).
TechRadar Pro
Multiple 2026 features on AI governance and workforce transformation.
https://www.techradar.com/pro
Source for Connolly governance quote and Cisco/Robbins commentary.
Entrepreneur
Surkiz organizational design feature, March 2026; Princeton GEO research coverage, April 2026.
https://www.entrepreneur.com/
Surkiz quote referenced on Fractional FAQ governance section.
TheStreet
JPMorgan Private Bank AI analysis, May 2026.
https://www.thestreet.com/
Reliability threshold framing for executive briefing context.
Forbes
Coverage of IKEA AI deployment and reskilling initiative, 2024–2025.
https://www.forbes.com/
IKEA case study, Insights Section 2 Block 5.
HR Grapevine
Coverage of IKEA AI literacy initiative and Klarna customer service reversal, 2024–2025.
https://www.hrgrapevine.com/
IKEA and Klarna case studies, Insights Section 2 Blocks 4 and 5.
Reworked
Coverage of Klarna AI customer service reversal, 2025.
https://www.reworked.co/
Klarna case study, Insights Section 2 Block 4.
Fast Company Workforce AI Anxiety Report
Coverage of AI use disclosure and workforce confidence, March 2026.
https://www.fastcompany.com/
More than half of workers are hiding AI use; half report faking fluency. About page second pull-quote.
Fast Company Organizational Design Feature
Article attribution and date to be confirmed.
https://www.fastcompany.com/
“Your company doesn’t have a talent problem. It has a design problem.” Services S6 inline pull-quote.
Fortune Fractional Executive Feature
Author and exact date to be confirmed.
https://www.fortune.com/
“AI has separated the judgment from the full-time seat.” Fractional persistent intro pull-quote, March 2026.
Inc. / Prezi CEO Interview
Peter Szafranski, Prezi CEO, on AI implementation, March 2026.
https://www.inc.com/
“Finding the right problem, that’s the hard part.” Services S1 inline pull-quote.
Inc. / Christos Makridis Industry Productivity Analysis
Coverage of Makridis research on AI exposure and labor market outcomes, May 2026.
https://www.inc.com/
+10% productivity, +4% employment, +5% wages across high AI-exposure industries 2017–2024. Insights Section 1 Card E.

Want to apply this research to your team?

The data above describes what most organizations are getting wrong. The next step is figuring out what your organization should be doing differently — implementation that keeps teams intact and delivers ROI without Redundancy.™

No sales pitch. Just a conversation about what you’re trying to accomplish. Schedule a conversation or reach out at hello@markphelps.com.
Schedule a conversation Explore services
Based in the Boston area, working on-site and remotely with US and UK clients.