A manager uses ChatGPT to draft customer communications. It saves her 30 minutes per week. Good outcome, right? But what has changed organisationally? Nothing. One person is more efficient. The broader team hasn't shifted how it works. The organisation hasn't captured or scaled the learning. The manager leaves the company, and the capability leaves with her.
This is the pattern in most organisations. There's plenty of individual AI use. McKinsey's 2025 research found that high-performing teams are 3 times more likely to have senior leaders actively championing AI adoption. But there's a massive gap between individual use and organisational capability. Companies where leaders express confidence in AI transformation achieve 2.3 times higher transformation success.
Deloitte's 2026 State of AI research shows that organisations are moving from experimentation to integration, but the process is messy. Watermark's 2025 research found that 90% of interim leaders use AI tools, but only 3% report extensive organisational adoption. There's a massive translation gap.
Designing Human-AI Workflows
Start with a workflow audit. Take a critical process in your organisation. Map out every step. For each step, ask three questions:
First: Where can AI add significant value? Not where could we automate, but where does AI augment human judgment or capability? Perhaps AI can rapidly synthesise data that a human then interprets. Perhaps AI can generate initial drafts that humans refine. Look for genuine augmentation, not replacement.
Second: Where must humans maintain control? Where does human judgment remain central? Where do values, ethics, or nuanced understanding matter most? Be honest about this.
Third: Where does the transition from human to AI to human happen? How does the work actually flow? Who inputs what? Who makes which decision? Who has decision-right to override or escalate? The actual mechanics matter enormously.
Once you've mapped this, document the workflow. Create clarity about roles, decision rights, and escalation protocols. Test the workflow with a small team. Learn what works. Iterate. Then scale with confidence.
Decision Rights and Accountability
One of the most fraught questions in human-AI teams is: who decides? There are three possible patterns:
Pattern 1: Human decides and can override AI. This protects human judgment and autonomy but can slow decisions. Works well for high-stakes decisions where human judgment is crucial.
Pattern 2: AI recommends and humans act unless they actively override. This accelerates decisions but can lead to thoughtless automation of AI recommendations. Works well for lower-stakes, high-volume decisions.
Pattern 3: Hybrid, where decision rights depend on confidence levels. If the AI is very confident and humans agree, implement. If the AI is confident but humans disagree, escalate. If the AI is uncertain, the human decides.
For each critical decision in your workflow, make an explicit choice about which pattern applies. Document it. Make it transparent to the teams doing the work.
Measuring Human-AI Team Performance
Don't measure by adoption rate. Measure by whether the workflow is actually improving. Are decisions faster? Are they better quality? Is the team learning? Is the organisation capturing value?
Create a simple dashboard for each human-AI workflow: speed of decision-making, quality of outcomes, team engagement, rate of learning, and cost impact. Track this quarterly.
Try This
Pick one critical workflow in your organisation. Map out every step. For each step, ask: Where can AI add value? Where must humans maintain control? How do we transition between human and AI work? Who has decision rights? Document this. Test with a small team. Iterate.
Create a ‘decision rights charter’ for how humans and AI work together on the decisions that matter most. Be explicit: In this situation, the human decides. In this situation, we implement the AI recommendation unless escalated.
Measure the performance impact of your human-AI workflow. Track: speed of decisions, quality of outcomes, team engagement, rate of learning. Review quarterly.
References
Deloitte (2024) State of AI in the Enterprise, 6th edn. Deloitte AI Institute.
McKinsey & Company (2025) The State of AI: How organisations are rewiring to capture value. McKinsey Global Institute.
Raisch, S. and Krakowski, S. (2021) 'Artificial intelligence and management: The automation-augmentation paradox', Academy of Management Review, 46(1), pp. 192-210.
Watermark Search International (2025) AI in Senior Leadership. London: Watermark Search.