Can AI 2x Human Productivity?

Most small companies do not have a shortage of ideas. They have a shortage of people who can carry an idea across functions. A technical founder can build the product but stall on positioning, while an agency owner can see a client opportunity but lack the time to research, package, and validate it.

The usual choices are another hire or slower execution. The Cybernetic Teammate, a large field experiment at Procter & Gamble, found another option: one professional using AI produced work as good as a two-person commercial and technical team. The individual also finished 16.4% faster than someone working alone without AI.

One person with AI matched a two-person team

The researchers gave P&G professionals real product innovation challenges from their own business units. Commercial and R&D specialists worked either alone or in cross-functional pairs, with and without AI. Independent evaluators scored the resulting proposals on quality, novelty, impact, business potential, and feasibility.

The numbers measure how much better each group performed than one person working alone. A larger number means a larger improvement, not a percentage increase. One person with AI reached team-level quality and landed almost level with a team that also used AI.

The AI-assisted individual reached that quality while using 16.4% less time. The proposals were also substantially longer, so the time saving came alongside more developed work rather than a shorter answer.

In knowledge work, 2x productivity can mean doing work that previously depended on two sets of expertise. The person gains range as well as speed.

AI gave specialists the perspective they were missing

Without AI, commercial professionals produced more commercially oriented ideas, while R&D professionals produced more technical ones. Their proposals reflected the function they already knew.

That divide largely disappeared when individuals used AI. Both groups produced a more balanced mix of technical and commercial ideas, and people whose normal role sat further from product development gained the most. With AI, they reached a quality level comparable to teams that included a product-development specialist.

Small companies face the same constraint every day. A founder may understand software but lack a strong sales perspective, while an agency owner may know delivery but have little time for positioning. AI can bring the missing perspective into the work before another specialist is available. AI makes cross-functional capacity available without waiting for another hire, allowing a one-person company to add functions without adding headcount.

More capacity makes human judgment more valuable

The peer-reviewed article separates idea generation from idea selection. AI raised the quality of the ideas people produced, while people remained better at deciding which idea deserved commitment.

Generation rewards breadth and speed. Selection depends on knowing the company, the strategy, the customer, and the consequences of choosing one direction over another. Producing another credible option becomes cheap, which makes judgment more valuable rather than less.

The AI-assisted proposals also became more similar to one another. AI helped people reach strong options, but several people were pulled toward related patterns. A founder still has to recognize which option fits the business instead of accepting the most polished answer.

Participants reported feeling more enthusiastic and less anxious after working with AI. They produced better work, finished faster, and enjoyed the process more, which makes the productivity gain easier to sustain in normal work.

A small company can gain capacity before it gains headcount

A large company can assemble specialists around an important problem. A small company has to choose which functions receive attention and which ones wait.

AI lets a founder test the commercial side of a technical idea, examine the operational consequences of a campaign, or challenge a product plan from the perspective of support and sales. More research briefs, campaign directions, product concepts, and operating plans can reach the point where a real decision is possible.

Headcount therefore becomes a weaker measure of what a company can attempt. One person can carry more functions and move several bounded jobs forward without pretending to be an expert in all of them. The founder spends less time producing every option and more time applying direction, judgment, and responsibility.

How Task Machine turns individual leverage into company capacity

The P&G study shows how much range one person can gain from AI. A company captures that gain only when the context, work, decisions, and results survive beyond one chat session.

Task Machine gives that work a durable operating structure. Chat is where people and agents set direction and develop new work, Tasks preserve the assignment and result, and repeatable jobs become workflows with checks and human approvals. When an agent has prepared several directions or reached a consequential action, the decision returns to the Inbox with the relevant context attached.

This follows the division of work visible in the study. Agents expand the number and quality of options, while people select direction and remain accountable for the outcome. The company gains more capacity without forcing the founder to supervise every step that produced it.

One strong AI session is useful. A repeatable process with context, verification, decisions, and history becomes part of how the company operates.

The experiment covered product innovation inside one company, so other work will produce different gains. Its central result is strong: one experienced person with AI matched a cross-functional pair, finished faster, reached beyond their own expertise, and enjoyed the process more.

For a concrete example, read how to run a structured brainstorming session with an agent. It turns the same generation-and-selection split into a repeatable process with a human decision at the end.

Source

• Fabrizio Dell'Acqua et al., “The Cybernetic Teammate: A Field Experiment on Generative AI and Teamwork”, Organization Science, published online June 2026.
• Fabrizio Dell'Acqua et al., NBER Working Paper 33641, April 2025.