3 Foundational Steps to Design for Effective Human-AI Collaboration

Most organisations are investing in AI. Far fewer are seeing meaningful gains.

McKinsey research found that nearly 90% of organisations have invested in AI, and fewer than 40% report measurable gains.

So, what’s getting in the way?

For many organisations, the challenge isn’t the technology itself. It’s the way we’re introducing it. Rather than redesigning work around the strengths of both people and technology, we’re often adding AI to existing workflows and expecting productivity gains to follow.

We’re finding that teams are quickly adopting AI to drive speed. And, indeed, it does seem that individuals are getting more done, faster. Unfortunately, it may come at a cost, weakening a group’s capacity to think critically and exercise sound judgement.

The question facing businesses today is: How can we optimise AI for effectiveness, not just efficiency? Efficiency asks: How quickly we can get something done? Effectiveness asks: Are we doing the right thing, in the right way, using the right judgement?

That’s where thoughtful human-AI collaboration comes in.

In a recent professional development video that I shared with Emergenetics Associates, I recommended starting with working agreements. These guidelines define how a group will collaborate using a set of rules, norms or standards. While typically used to align a team of humans together on the principles they will apply to coordinate their work, the concept applies just as well to a department made up of people and technology.

3 Questions to Shape Human-AI Working Agreements

To get started, consider each of the following three questions at three levels:

• Individual level, so that AI is adding value beyond simply helping us work faster.
• Team level, to ensure that collective human intelligence is being integrated to improve outcomes.
• Organisational level, in order to prioritise effectiveness and sustainability rather than efficiency alone.

#1 — What should AI accelerate?

Technology can be a great tool for any task where speed adds value without losing accuracy. For example, when I create professional development videos, AI often helps me structure ideas to get the first draft moving along.

For individual work, it can support tasks like data sorting and summarising. For teams, it may improve process-heavy workflows like scheduling or delivering status updates. For companies, it might offer analytics dashboards to enhance operational consistency at scale.

#2 — What should humans still own?

Overreliance on AI can lead to several negative outcomes. While it can stretch our thinking, it often reinforces defaults. For instance, if you lead with Analytical thinking, AI can feed you more data and keep you in analysis mode too long. If you lead with Conceptual thinking, it can generate so many possibilities that you’re overwhelmed by options. AI doesn’t automatically make our thinking more balanced; it often reinforces our defaults – and does so quickly!

Additionally, while effective at past pattern recognition, it tends to oversimplify complexity, fails to understand interconnections and struggles with ambiguity. Humans are far better equipped to navigate these nuances.

In the example of designing my PD video, I applied this principle by having AI create the first draft while my colleagues and I owned the decisions about how to shape the content, refine the messaging and ensure it met our goals. The real work is still human.

In coordination with two of colleagues, each with distinct Emergenetics Profiles, we brought cognitive diversity to the process, challenging various assumptions and considering what would resonate most with the audience of Associates. At the organisational level, this same principle holds: humans still need to own culture, vision and where accountability sits.

#3 — What decisions require human judgement before action?

When complexity, context and multiple perspectives need to be brought together to reach a meaningful outcome, human judgement matters. Anything affecting strategy, people, trust, fairness, relationships, reputation, safety or accountability still need a human touch.

AI can highlight risks and considerations; however, the ethical weight of a choice, even one as seemingly small as a word that’s used, needs to have a real person behind it.

Considering these questions, a working agreement for a team might sound like: We use AI to add to our thinking, not replace it. Humans must hold the judgement.

Implementing Your Working Agreements

The guidelines your team will define may sound different because we all bring unique perspectives, experiences and ways of thinking to the conversation.

What matters is that you have the conversation and get the discussion going.

The gap we’re seeing in the efficacy of our fast-changing workplaces usually does not stem from technology itself. The obstacle is that we are not having a dialogue about how we will integrate the tools productively.

Collaborative intelligence is produced when people know how to use AI, when to question it and when to bring human reasoning back in. Moreover, it’s important to note that these conversations and working agreements must be regularly revisited, especially as roles change and new technology is introduced.

Consider AI as your newest team member. To get the best out of this new colleague, they must be thoughtfully onboarded and integrated into the department. So, the next time someone says that “we need to adopt AI,” consider how you’ll actually make it a part of the team.

Emergenetics Associates: Be sure to watch our July Mastery Mondays session to learn more about this topic and download our 3×3 Collaboration Design Resource — a grid mapping each question against individual, team and organisational examples — so you may initiate human-AI collaboration conversations in your workplace!

 

References: McKinsey State of AI in 2025 survey

 Print This Post