The Path to Higher P/E Multiples? Scaling AI-Enabled Productivity End-to-End

Last week on CNBC, BlackRock Chief Investment Officer Rick Rieder cited “extraordinary” earnings driven by AI-enabled productivity gains in “inventory management, logistics, automation, and customer procurement” to explain surging stock markets. This view, however, seems out of whack with MIT’s much discussed finding that 95% of generative AI projects “fail.”

Maybe the disconnect is a matter of scope. “Extraordinary” earnings for the F500 crowd are not about running successful pilots, but a result of scaling transformative technologies for step change improvements in operations, end-to-end. 

The takeaway for chief operating officers orchestrating execution from supply through commercial operations and customer service is to learn to use agentic AI systems to improve metrics like revenue per employee, lifetime customer value, and time to market for new products.  

Incrementalism Is a Trap 

Operations is a discipline dedicated to continuous improvement, and while this is an admirable principle, it can revert to incrementalism. The risk is that even well-run operations might sleepwalk into an era where investors are pricing 30-50% productivity gains into equities. Slow and steady won’t win this race. 

I had a conversation with Torsten Pilz (then CSCO of Honeywell) a few years back. His view on AI was that return on investment is so massive when applied correctly that detailed business cases are a waste of time. Since then, I have seen planning teams eliminate 30% of human work in S&OP, sourcing teams automate 70% of supplier negotiations, and customer service teams quadruple labor productivity in support center work. 

Meanwhile, generative AI’s proven ability to draft a sourcing risk management plan or propose a channel segmentation strategy is so obvious that doubts about its value for business are melting. Equally mind-blowing is the power of vibe coding to quickly and cheaply build AI agents. Tools like Replit can generate rough but usable V1 agents to do work like spotting suppliers missing on SLAs and identifying root causes for delivery performance issues – in 30 minutes. 

If continuous improvement is analogous to squeezing the stone for another drop of blood, agentic AI will feel more like a magic wand.  

Back to Wall Street and the COO’s AI Mission

Everyone loves to debate whether we’re in an “AI bubble.” Who cares? For operations leaders, the important question is how to master the technology while it is still newly emerging. 

Technology booms including railroads in the nineteenth century and the internet in the 1990s often feature bursting bubbles, including the Panic of 1873 and the dotcom crash in 2000. Price-to-earnings (P/E) multiples for companies selling the new tech can go crazy, but they also spike for companies expected to use the tech to accelerate value creation in other industries. 

At its peak, the unweighted average P/E multiple for the top seven tech stocks in 2000 was 276x. The comparable figure for the Mag 7 today is about 70x, reflecting much better tech company earnings now. In both cases (and probably railroads too) the crazy multiples aren’t ultimately justified for most stocks, but they may be for the wider economy. 

In 2000, non-tech price-to-earnings ratio multiples peaked at almost the same level as they are now: ~23x, which is 40% higher than the historical average for the S&P 500. Perhaps that 40% bump reflects investors’ collective assessment of how much more productive companies will be using AI.  

Agentic AI Is All About Workflows 

The breakthrough with agentic AI is that it can do work that requires understanding, reasoning, and decision making. This is very different from past tech booms that were fundamentally infrastructural, like rail for transport and the internet for commerce.  

These human-like capabilities exist in workflows across functions, like planners consolidating forecasts or account managers creating customer quotes. Productivity gains from agentic AI are less about automating defined tasks in functional silos than handling ambiguity across silos in pursuit of shared business goals. This principle is the essence of operations research, minus the deterministic math of optimization. 

Workflow acceleration with AI agents handling huge volumes of repetitive human judgement tasks like credit checks, service diagnostics, inventory forecasts, or delivery route plans will revolutionize labor productivity in operations. 

COOs should embrace this chance to prove investors right.