By James F. Meyer

When I was in college, I worked in the Accounting department during the day and took computer classes at night.  Personal computers had just begun to make their way into American companies.  Even at the college, everything in our Accounting department was done on paper ledgers.  When I took classes at night and used Lotus 123 and dBase, the contrast between what was and what was coming was stark.

My first full-time computer job out of college was at a division of Alcoa Aluminum.  We manufactured parts for the US military, including the blades for the Apache helicopter.  I was one of only a few people in our office that had a desktop computer; everyone else was using paper…on their metal desks with ashtrays filled to the brim (it was a different time).

Needless to say, PCs fundamentally transformed the workplace and American business.  It wasn’t just a change in tools but a change in the way organizations were structured and operated.  This was one of the few times where technology did actually change just about everything.  There had been mainframes and mid-range computers before this, of course, but PCs in the workplace were to the Information Revolution what Henry Ford’s assembly line was to the Industrial Revolution.

There are lots of change cycles that occur; some very large, like the Internet, others are incremental like cloud computing, SaaS, and mobile; but PCs were a systemic change.  We are at that point now with AI.

AI will fundamentally change how organizations are structured, managed, and operate.  As with the PC Revolution, there will be companies that understand this structural change and make the transition while others will simply see this technology as another tool to layer on top of their current structure.  Just like with PCs, the latter approach will mark the slow deterioration of established, successful businesses because they do not recognize the moment, and by the time they do, their competitors have rushed past them.

95% AI Failure Rate?

The now famous (or infamous) MIT NAND initiative, “The GenAI Divide: State of AI in Business 2025” [1], a study of 300+ public AI initiatives, reported a 95% failure rate for corporate AI initiatives; that is, no notable ROI.  What is most interesting are the reasons for failed initiatives; these include:

• The use of generic AI tools, like ChatGPT, that do not learn or remember over time.
• Over-allocation of AI budgets to high-visibility areas.  More than half of AI budgets are devoted to Sales and Marketing tools, yet the biggest ROI comes from back-office automation and optimizing specific workflows.  Back-office gains are achieved by eliminating business process outsourcing, cutting external agency costs, and streamlining operations. [2]
• Failure to partner with specialized vendors (67% failure rate on internal-only initiatives).

Who’s Winning?

The stories of companies with challenging AI experience grab most of the headlines; however, between the hype and the noise, there are companies having huge success with their AI initiatives. [3]

JP Morgan is saving $2B annually; $200M of which was with its contract system (COIN) alone.  The COIN system can process 12,000 annual contract agreements in seconds, saving 360,000 hours of manual legal review and increasing accuracy.  JP Morgan has over 600 AI use-cases in production with another 400+ on the way. 

American Express has reduced annual fraud losses by $2B with a 99.5% accuracy rate and a <0.1% false positive rate; a 10x ROI within two years. 

Klarna saw a $40M profit improvement from AI customer service.

Octopus Energy scaled operations without scaling headcount by handling 44% customer inquiries via AI.  Customer satisfaction scores rose from 65% for human-only interaction to 80% with AI interactions.  Response times decreased from hours to seconds.  Octopus Energy was able to scale while only adding two more people vs. 250 without AI.

Walmart saved $2B by optimizing supply chain operations.  Out-of-stock incidents fell by 30%, forecasting accuracy rose from 80% to 95%, and 85% of routine inventory decisions are fully automated.

Maersk saved over $100M using AI for predictive maintenance.  Unplanned downtime was reduced by 70%.  Equipment lifespan extended by 20% to 30%.  Maintenance costs were cut by 25% and more than 400 critical failures were prevented annually.

Tesla reduced its manufacturing costs by 50%.  Factory footprint size was reduced by 40%.  Manufacturing costs were reduced by up to 50% per vehicle while quality and consistency improved.  It scaled from thousands of vehicles per year to millions.

Duolingo saw a 45% YoY revenue growth powered by AI personalization.  Their course completion rate increased by 50%, engagement time fell by 30%, and they scaled to over 500M users without a corresponding increase in content costs.

Stripe reduced fraud costs to merchants by $500M.  Fraud was detected with a 99.9% accuracy rate and the false-positive rate fell by 60%.  $20B in fraudulent transactions were prevented.

Typical savings by sector / use-case:

• Customer Service: 60% to 70% cost reduction with a 6 to 12 month ROI.
• Supply Chain & Logistics: 20% to 30% efficiency gains with a 12 to 18 month ROI.
• Healthcare: 20% to 40% capacity improvement with an 18 to 24 month ROI.
• Manufacturing & Quality Control: 40% to 50% reduction in defects with a 12 to 18 month ROI.

The difference is in the approach.  Organizations that had success made these crucial decisions, including:

• Focusing on specific pain-points and workflows rather than attempting broad, generic AI deployments. [2]
• Selecting tools that can integrate and learn over time.
• Partnering with specialized, trustworthy AI vendors.

The MIT report also makes recommendations for successful AI implementations, including:

• Focus on back-office automation.
• Target specific workflows.
• Empower operational managers to help drive organizational change.

Strategy Wins

As with any systemic change, basic business logic prevails:  have a strategy, focus on operations, examine and streamline business processes, build on incremental success.  Whether it’s assembly lines, or PCs, or AI, targeting internal operational pain-points that can produce the highest ROI in the shortest amount of time, with the least amount of disruption vastly improves a company’s chances of success.

Organizational change…and that’s what AI is bringing…requires a strategy and a specific tactical execution plan.  The strategy has specific, incremental success milestones upon which the next efficiency can be built.  “Big-bang” projects are expensive and exhausting, and, because of the speed with which targeted AI solutions can be implemented, completely unnecessary.  A company may have a million-dollar AI budget but there’s no need to spend it all at once.  Manage the change itself as a process.

The Strategic Plan

There are many aspects to any organizational change.  Unfortunately, in part, because of the rapid implementation potential of AI, and technology in general, companies often deal with the full breadth of considerations as they occur.  The problem with this ad-hoc approach is that poor decisions are easier to make when timelines and budgets have already been established and these decisions are made under duress.  This forces companies to have reoccurring discussions about timeline and budget.  No one enjoys sliding schedules or rising costs.

Companies need a comprehensive strategy that includes a tactical execution plan for each element.  Partnering with vendors who have extensive business operations and organizational change experience is as important as their technical capabilities, perhaps more so.  There are thousands of technology implementors but few of those are experienced in business operations and the nuances of organizational change.

Some of the questions a comprehensive AI strategy should answer, include:

• How can we organize and structure our internal data to make it more AI-useful?
• What core business processes do we need to identify and prioritize?
• Where should our AI live; on-premise, hosted in a VPN, or with a commercial AI cloud company?  What are the token metering and bandwidth costs of these options?
• Should we have on-premise or private hosted chat models, replacing tools like ChatGPT and Copilot, in order to protect our…and our customer’s…sensitive information?
• How do we size our AI usage costs for on-premise vs. private hosting vs. commercial AI?
• How do we approach AI agents and automated workflows?  What tools should we use?  Where should those tools live?
• How do we harden (air-gap) on-premise or private hosted implementations?
• How do we track the success metrics of our AI implementations?
• How do we manage AI agentic workflows?
• How will our models and AI agents be updated?
• What is our corporate AI governance, regulatory, and security plans?
• What are the geopolitical risks to AI and system availability?

Conclusion

There is a clear difference between successful AI initiatives and ones that fail, between losing huge sums of money and gaining huge cost and productivity savings.  As with any systemic change, the approach makes all the difference.

Just like the systemic changes that preceded it, the AI Revolution will have winners and losers.  Established companies that take a disciplined, strategic approach will succeed and those who don’t will be passed by current competitors and nimble new ones.  If we turn down the noise and avoid being overwhelmed by the hype-cycle, there is a fundamental business approach to leveraging AI to move your company forward.

A clear strategy, a tactical execution plan, knowledgeable, experienced business partners; these make the difference.

At the end of the day, it’s just business and the fundamentals never change.

James F. Meyer is the Director of Products & Innovation at ELYON International. You can reach James at James, [email protected] and their website is  www.Elyon.International (no “.com”).

[1] The GenAI Divide STATE OF AI IN BUSINESS 2025. https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf

[2] MIT Report Reveals Shocking 95% Failure Rate for Corporate AI Projects Despite $30-40 Billion Investment; Aug 25, 2025. https://eprnews.com/mit-report-reveals-shocking-95-failure-rate-for-corporate-ai-projects-686740/

[3] Show Me the Money: 17 Companies That Proved Agentic AI ROI. https://www.linkedin.com/pulse/show-me-money-17-companies-proved-agentic-ai-roi-john-chavner-bojkc