AI’s Real ROI: Augment First, Automate Later

  2025-08-29
  4 min read

The MIT NANDA Program recently published The State of AI in Business 2025 report, stating that “95% of organizations are getting zero return” from Generative AI (GenAI). This news was unexpected and led to a short but sharp drop in financial markets.

But reading through MIT’s 26-page report carefully paints a different story. Eighty percent of organizations have explored or piloted GenAI, and half of those have purchased a subscription; however, almost every single person has used an LLM in some form at work. These tools primarily enhance individual productivity, not P&L performance. Organizations reporting success in AI implementations evaluate tools based on business outcomes rather than software benchmarks.

The study highlights an unforeseen point: it may be hard to measure, but organizations are already reaping the benefits of GenAI, as employees adopt it autonomously.

We’ve Been Here Before

In a 1987 article in the Times Book Review, Robert Solow, a Nobel-winning economist at MIT, commented, “You can see the computer age everywhere but in the productivity statistics.” The “productivity paradox,” as it came to be known, was never satisfactorily solved. Three years after OpenAI released ChatGPT, we may be facing a new economic paradox. 

There is pressure on all levels of leadership to leap into the benefits of GenAI and increase organizational profits. The default leadership view is that replacing people with technology will yield savings that speak for themselves, and the money will roll in.

But there are limitations to the technology. The MIT report mentions brittle workflows, lack of contextual learning, and misalignment with day-to-day operations. Barriers to scaling GenAI systems are the inability to retain feedback, adapt to context, or improve over time. I will add two more: cost and hallucinations.

Even at $200/month, OpenAI loses money. This suggests that OpenAI’s pricing strategy follows a market penetration approach, prioritizing user adoption over immediate profitability. Long-term, sustained daily use pricing may rival salaries.

Established organizations are particularly sensitive to GenAI hallucinations. Depending on whether GenAI is exposed to customers, the organization may lose customer trust in its operations or accumulate an insurmountable amount of bad internal data that is fed into decision-making. What if humans are involved in validating and refining GenAI responses? Well, that’s exactly what’s happening today when employees use GenAI to augment their performance.

The Case for Augmentation

Augmentation with AI is happening now. It brings clear business value that is hard to measure, but reduces the risk of hallucination impacts and other limitations, and decouples your organization’s internal operations from future AI price hikes. It also brings increased staff satisfaction and retention, as people tend to stay (longer) at organizations not engaged in layoffs. This doesn’t mean you can’t automate parts of your business with GenAI, but the default should be employee augmentation. Imagine it as doubling your head count almost for free.

If you’re a CTO/CEO, here are three things you can start doing tomorrow:

1. Provide enterprise-grade GenAI access — Don’t rely on employees’ personal accounts; Ensure confidentiality and data protection from the provider.
2. Invest in people, not just tech — Bring in a GenAI training partner. Keep leadership close to training so real business use cases flow back into adoption.
3. Redefine your metrics — Don’t expect immediate gains in revenue per employee. Track task-level efficiencies and business outcomes.

The leaders who win with AI will be those who augment first, automate second.