Skip to main content
    Back to The Human Advantage
    Adaptation
    5 min read19 January 2026

    Whether people can direct, combine and check AI agents in real work.

    Building Human-AI Teams That Actually Work

    Individual AI use rarely becomes team capability on its own. Here is how to design human-AI workflows, agree who decides, protect how the team works together and measure what matters.

    Ben George

    Founder & Director

    A manager uses AI to draft customer emails and saves half an hour a week. One person is faster. The team works the same way, and the learning leaves when she does.

    That gap between individual use and team capability is common. McKinsey's State of AI 2026 found 80% of firms report better individual productivity. Yet only 37% report any profit effect, the same share as a year earlier.

    Design closes the gap. Deloitte's 2026 Global Human Capital Trends found organisations focused on work design were twice as likely to exceed their AI return expectations.

    Design the workflow

    Map a key process step by step. For each step, ask three questions.

    Where does AI add value? Look for augmentation. AI drafts or summarises, and people interpret and decide.

    Where must people stay in control? Values, ethics, relationships and high-stakes judgement.

    Where are the handoffs? Who supplies what, who checks and who can override?

    Test each step. Dell'Acqua and colleagues found AI capability is uneven, strong on some tasks and misleading on others.

    Document roles, decision rights and escalation routes. Pilot with a small team, adjust, then scale.

    Build this capability in your organisation

    Growth Performance designs bespoke programmes that turn insight into measurable behaviour change.

    Explore Adaptation

    Agree who decides

    Three patterns cover most decisions:

    • A person decides. AI informs and a person chooses. This suits high-stakes calls.
    • AI recommends, a person approves. This is fast but risks rubber-stamping. It suits low-stakes, high-volume work.
    • Conditional. When the person disagrees or the AI is uncertain, the person decides or escalates.

    Make the choice explicit for each critical decision, and tell the team.

    Teams behave differently with AI

    In Ju and Aral's 2026 experiment, human-AI teams produced 50% more ads per worker. They also sent 18% fewer interpersonal messages, and their outputs were more alike.

    Protect the human exchange that builds ideas and relationships.

    Checking matters too. A 2026 study of 962 US desk workers found 38% had received AI "workslop" from colleagues. Each lost about 3.4 hours a month fixing it.

    Measure the right things

    Adoption rates say little. Track decision speed, quality, rework, engagement and what the team is learning. Review quarterly.

    Redesigning roles, hand-offs and checking around AI agents is Adaptation.

    Try this

    Pick one workflow. Map each step and mark where AI helps, where people decide and who checks. Agree one change and test it for a month.

    Then use the free Human-AI Workflow Design Canvas.


    Want to explore these ideas further?

    Tell us about your people and we will shape options around them.

    Get options for your team

    Trusted by McDonald's, Nissan, the NHS and Munich Re. 97% average satisfaction across 500+ programmes and workshops.Source: Growth Performance internal research data and surveys, June 2016 to June 2026.