“AI Reduces the Cost of Executing Work, but Not the Cost of Coordination,” Says Ian Beacraft

One of the most influential voices in the discussion about AI and work argues that today we invest 90% in execution and 10% in coordination—and that this ratio is about to reverse.

mission [EXP + PROS]

Artificial intelligence is no longer just a tool; it is beginning to reshape the very logic of organizations. That was the central argument of the panel “How to Design a Company That AI Can’t Outpace,” featuring Ian Beacraft, CEO of Signal and Cipher and one of the most influential voices today in the debate around AI and the future of work.

Rather than simply asking how to use AI, Beacraft proposes something more radical: how to design entire organizations for a world where executing tasks has become cheap and nearly instantaneous. The challenge, he argues, is no longer doing the work itself, but coordinating it and designing systems that make sense.

A recent turning point has been the rise of AI agents accessible to everyone, symbolized by the success of OpenClaw—an open system of AI agents capable of autonomously executing complex tasks while coordinating multiple digital processes. This marks the transition from a model where a person interacts with a chatbot to a scenario in which multiple agents operate simultaneously.

Treating AI merely as a tool, however, is a mistake. “A good knife doesn’t make a good chef. Great AI tools don’t make an AI-native worker,” Beacraft says. Many companies use AI simply to accelerate existing processes, producing small gains that quickly turn into nothing more than more work.

The physics of work

“For 150 years, we built our organizations around the idea that humans were the limiting factor,” Beacraft explains. Approval chains, departments, and sequential workflows all emerged from those human limitations. With AI, that logic begins to change—and organizational structures will need to evolve along with it.

“AI drastically reduces the cost of executing work, but it doesn’t reduce the cost of coordination,” he says. “Today we invest 90% in execution and 10% in coordination. That ratio will likely flip.” In many cases, building a prototype now costs less than the meeting required to plan it.

AI agents also struggle with vague feedback such as “this looks good.” According to Beacraft, organizations must define success much more clearly. “It’s like evaluating the work before it even begins—where everyone is extremely clear about what success looks like, how it will be achieved, and the moment anything indicates success is not being reached, the work stops,” he explains.

Over time, the rules and patterns learned by agents stop being mere memory and become part of the system’s operational infrastructure.

Culture as data

For agents to act correctly, everything that used to be implicit must become explicit: values, tone of voice, quality standards, and decision-making processes. More than a technological challenge, this requires organizations to clearly define who they are.

The major shift, Beacraft argues, is moving from a mindset of “doing the work” to “designing the work.” Competing with AI on speed and scale is futile. Human value increasingly lies in intent, judgment, and the ability to architect systems.

Want more?

In The Second Machine Age, Erik Brynjolfsson and Andrew McAfee argue that we are living through a new industrial revolution—this time not driven by physical power but by cognitive capabilities. The difference now is that machines are not just helping humans work; they are beginning to participate in designing the work itself.