← Ideas & guides

Reflection

AI makes the task easier. How should collaboration change?

A reflection on Benedict Evans’s conversation: what individual results tell us about team composition, and how a new way of working can be examined.

By Dr. Sven JungmannRetrospective reference date: · Published: · Reviewed:
A cog is connected to a loom through rearranged threads.

At a glance

  • A convincing individual result does not establish a permanent staffing decision.
  • The sequence of collaboration can be changed deliberately.
  • Preparation, joint judgement and transition each deserve observation.

When someone produces a convincing proposal with AI, a leader will soon face another question: will this work need the same group of people in future? An organisational decision sits between the quality of a proposal and the composition of a team. It deserves a justification of its own.

In his conversation with Lenny Rachitsky on 31 May 2026, Benedict Evans distinguishes an executable task from the wider contribution of a professional role. The published chapter on professional services begins at 09:44. [1] This prompts a concrete question for hospitals and businesses: which collaboration should change when preparing a task becomes easier?

My answer starts with the sequence in which people work together. A technical improvement could make shared time more valuable earlier, later or elsewhere in the process. These possibilities can be investigated before a convincing individual result becomes a permanent staffing decision. The approach below is my interpretation for organisational design; neither the conversation nor the studies tested it as an implementation method.

What an experiment can contribute to team design

Dell'Acqua and colleagues' 2026 publication analyses 791 Procter & Gamble professionals in a one-day product-development exercise conducted in 2024. Randomly assigned individuals with AI achieved comparable rated proposal quality to two-person teams without AI. Proposals were assessed; this established neither lasting collaboration nor subsequent market outcomes. [2]

For leadership, this opens a possibility that can be examined: perhaps some preparatory activities will need less simultaneous coordination. Whether professional exchange consequently becomes unnecessary is a further question. The ability to formulate a proposal and different participants' willingness to implement it together should be observed separately in a local investigation.

In a product-development meeting, someone might bring technical and commercial considerations together in an initial draft. Later, people still need to decide which assumptions to investigate, what resources are available and which customer question takes priority. The effect of the preparation on those decisions would be a separate subject for assessment, requiring observations beyond the first document's quality.

Examine the different contributions of collaboration

Before changing who participates, I would ask why the current group comes together. Some people provide information. Others expose conflicting requirements. Others can make binding resource decisions. The same person may provide several contributions. A list of job titles captures these differences only partially and can make quite different kinds of participation look interchangeable.

For each contribution, it is then possible to describe when it is needed. Should a clinical perspective participate in defining the problem? Can a commercial assumption be prepared first and examined together later? Which uncertainty becomes intelligible only through speaking to the intended users? Such questions produce a more concrete organisational choice than a general demand for smaller teams.

A useful result could be a redesigned meeting. Participants receive a prepared question, identify their own assumptions in advance and use the shared time to examine them. Alternatively, early joint work might prove essential because the participants initially mean different things by the problem. Both possibilities should remain available for investigation before a new arrangement is adopted.

A constructed medical device example

A fictional medical device manufacturer wants to revise the education it provides for an existing device. A product specialist and a clinical application expert currently develop the initial concept together. The company is considering asking one person to prepare it with AI and bringing the two together afterwards.

I would compare three arrangements: the existing joint approach, individual preparation followed by a joint discussion, and two independent preparations brought together later. The third is deliberately an alternative of my own. It examines whether separate preparation reveals additional questions that may receive too little attention when everyone starts from one early draft.

Every arrangement receives the same bounded task and publicly usable or explicitly authorised materials. A useful draft is defined in advance: it should identify the intended audience, mark unresolved assumptions and formulate a question about the education that can be investigated. A proposal can sound plausible while still being inadequate on one of those dimensions.

I would then observe what happens in the joint discussion. Are fundamental assumptions reopened? Do missing user perspectives emerge? Can both participants explain why the chosen direction should be pursued? These are proposed observations for this fictional case. They provide information about collaboration without turning a short exercise into a reliable forecast of organisational performance.

Plan for the work of changing the arrangement

Brynjolfsson, Rock and Syverson's productivity J-curve paper models complementary intangible investments, including processes and skills. Such investment can make measured productivity temporarily diverge from the accumulation of future productive capacity. This is an economic model with historical applications; it guarantees no successful outcome for a present-day AI initiative. [3]

For the fictional company, I would derive a straightforward planning rule: designing the new collaboration also takes time. Someone needs to make the task clear, select appropriate materials and try the revised meeting format. If that occupies capacity, the work should be visible in the decision paper so the investment can be considered as a whole.

The company should also specify the reusable result being created. This might be a better task description, a workable process for joint decisions or a clear expectation for preparation. Merely referring to a necessary learning period does not explain why this particular initiative deserves continued support. The intermediate contribution needs to be describable and open to assessment.

Make a reversible change before a permanent commitment

As a first organisational step, I would change when people participate in a clearly bounded task. The assignment remains comprehensible, the necessary perspectives remain accessible and participants know when they will come together again. After several comparable tasks, the organisation can examine whether the new sequence continues to help, rather than relying solely on an encouraging first occasion.

I would also look for consequences that an initial output assessment might miss. Does someone absorb additional follow-up questions? Are important choices effectively made in the draft before others can introduce their assumptions? Does the joint meeting become shorter while later coordination increases? These questions form part of my proposed assessment; they are not reported findings from the P&G experiment.

This offers investors a useful distinction as well. A demonstration shows a capability. A repeatedly successful way of working supplies additional evidence about how an organisation can use that capability. A company's account should make clear which of those levels it can already substantiate. Progress can then be recognised while the remaining organisational work stays visible and available for evaluation.

Evans' distinction therefore leads me to a question that can be answered in the next planning meeting: where do we need to exercise judgement together, and what preparation could improve it? Developing a precise answer allows tasks and collaboration to evolve together. A decision about the permanent composition of the team then has a basis that others can examine.

Sources and further reading

  1. Benedict Evans with Lenny RachitskyLenny’s Podcast

    Starting point: tasks and professional contributions.

  2. Dell’Acqua and colleagues (2026)Organization Science

    Proposal quality in a bounded experiment.

  3. Brynjolfsson, Rock and SyversonNBER

    Model of complementary intangible investment.

The starting point for this reflection

Benedict Evans with Lenny Rachitsky

Perspective and interests

This article was developed with AI assistance. The organisational examples are fictional. The practical proposals are the author’s inferences from the bounded sources.

I am the founder and CEO of aiomics and have a commercial interest in responsible AI adoption in medicine.

Keep reading

What should your event make possible?

Tell me about your audience, occasion and timing. We can shape a talk around the questions that matter to them.

Enquire about a talk