by Mick Lavin, Coach, Agile Coach, Mentor
The organisations getting real, measurable value from AI aren’t the ones who’ve deployed the most tools. They’re the ones who’ve figured out how to work alongside AI in a way that multiplies, rather than merely adds to, human capability.
This distinction matters. Adding AI to your workflow can save you an hour here or there. Multiplying through AI, building genuine collaborative intelligence between human judgment and machine speed, produces a fundamentally different order of result.
The good news is that the research on how to do this well is clear, practical, and directly applicable to HR and legal teams. Today, we explore the two most successful models of human-AI collaboration, how to train your people to use them, and what multiplicative gains actually look like in practice.
758 Consultants and the Jagged Frontier – A Definitive Study
The most important research in this space comes from a Harvard and Boston Consulting Group study of 758 management consultants. It’s worth exploring this in some detail, because the findings are both counterintuitive and practically actionable.
Researchers gave some consultants access to AI tools and others none, then tracked performance across a range of tasks. For tasks sitting comfortably within AI’s capability frontier, creative ideation, broad content generation, rapid data synthesis, the results were striking:
- Consultants using AI completed 12.2% more tasks overall
- They operated 25.1% faster
- Their output was graded at 40% higher quality
- The lowest-performing consultants saw the most dramatic improvement, a 43% performance boost
But there’s a critical counterpoint. For complex strategic tasks sitting just outside AI’s capability frontier, tasks that required nuanced judgment, contextual awareness, or real-world complexity, consultants who used AI performed 19 percent worse than those who worked without it entirely.
The AI didn’t fail because it was switched off. It failed because users trusted it on tasks it couldn’t handle, and stopped engaging their own judgment.
The practical implication is important: knowing where the frontier is, and adjusting your collaboration model accordingly, is the foundational skill of effective AI-augmented work.
Two Models That Work – Centaurs and Cyborgs
Through analysis of the consultants who consistently outperformed their peers, researchers identified two successful archetypes of human-AI collaboration.
The Centaur Model
Named after the mythological half-human, half-horse creature, the Centaur model describes a macro-level division of labour. Centaurs are clear and deliberate about which tasks go entirely to AI and which stay entirely with humans.
A Centaur may delegate the following to AI:
- Structuring large datasets
- Drafting standard correspondence
- Summarising background research
- Writing initial policy drafts
They retain the following exclusively for human execution:
- Final strategic synthesis
- Nuanced stakeholder conversations
- Ethical review and accountability
- Any decision with material consequences
The key characteristic of the Centaur is strategic discernment, the ability to know clearly which category a given task falls into, and to maintain that discipline consistently.
The Cyborg Model
The Cyborg model describes deep micro-level integration. Cyborgs don’t delegate whole tasks and walk away. Instead, they maintain a continuous, interactive dialogue with AI at the granular sub-task level, prompting, refining, challenging, and steering outputs in real time.
A Cyborg working on a complex policy document might:
- Ask AI to generate a structural outline
- Critically review and revise that outline
- Use AI to draft a specific section
- Rewrite it in their own voice with contextual nuance
- Ask AI to check it against specific legal or regulatory requirements
- Apply final human judgment to the overall document
This model treats AI as a continuous cognitive exoskeleton, always present, actively used, but never running autonomously.
Both models work. Both consistently outperform the third archetype identified in the research: the Self-Automator – the person who hands a task to AI, accepts the output uncritically, and presents it as finished work.
We have seen these accepted outputs in the news headlines as legal teams confidently presented non-existent case law. I have heard about it from lecturers reviewing student papers with fictional bibliographies.
Personally, I use both methods as I create proposals for clients, structure training material, and sometimes experiment with code. For research purposes, I will use tools such as NotebookLM that grounds the research in the reference material I provide and therefore reduces any chance of hallucination.
No matter which role you choose, the most important thing is that you validate the output.
Four Domains of Multiplicative Gains in Practice
When deployed through the Centaur or Cyborg approach, AI delivers what the research describes as “multiplicative, rather than merely additive” gains across four key areas:
Creativity
AI eliminates the blank page. A Cyborg worker uses AI to generate fifty divergent concepts in seconds, rapidly discards the mediocre, and synthesises the best elements into a genuinely original final product. The creative output is better than what either the human or the AI would have produced alone.
Decision-Making
A Centaur approach to decision-making means using AI to rapidly ingest vast quantities of historical data, policy documents, and risk indicators, presenting structured probabilities that the human leader then evaluates through the lens of ethical nuance, lived experience, and strategic ambition.
Learning and Development
AI functions as an infinitely patient, highly personalised tutor available at any moment. Employees can ask complex technical questions in real time, explore concepts at their own pace, and receive immediate, tailored explanations. For L&D teams, the opportunity to embed AI-powered just-in-time learning into daily workflows is genuinely significant.
Collaboration
By automating the administrative overhead of collaboration, meeting summarisation, action-item extraction, cross-language translation, AI frees human teams to focus entirely on deep strategic debate and genuine connection.
Investment Drives Returns – The Training Imperative
There’s a stark finding that HR directors involved in L&D should take seriously. Data from Ernst & Young indicates that employees receiving over 81 hours of AI-specific training annually report productivity gains of up to 14 hours per week, nearly two full working days.
Employees receiving standard, superficial training report negligible gains.
The difference isn’t in the tool. It’s in the depth of capability to use it effectively. Shallow AI training, a one-hour lunch-and-learn and a guide document, doesn’t develop the judgment, the prompting skills, or the collaborative habits that generate real productivity gains.
This has direct implications for how HR teams design and invest in AI capability building. The question isn’t “have we trained our people on AI?” It’s “have we trained them deeply enough to unlock genuine performance gains?”
Practical Steps for HR and Legal Professionals
Map your work to the frontier
Take the most common task types in your team and honestly assess which sit safely within AI capability and which require human judgment. Be specific. “Drafting correspondence” sits within the frontier. “Advising on a complex grievance” does not.
Train for Centaur and Cyborg working explicitly
These aren’t abstract concepts, they’re practical working patterns that can be taught, practised, and refined. Build this into your AI capability development programme.
Develop prompting skills
The quality of AI output is significantly shaped by the quality of the input. Strong prompting, specific, contextual, and structured, produces dramatically better results than vague requests. This is a learnable skill, not a technical one.
Set boundaries around self-automation
Make clear in your team’s AI usage guidelines that uncritically accepting and presenting AI output as finished work is not acceptable. Build in verification habits and accountability structures.
The organisations pulling ahead in AI adoption aren’t doing so because they’ve removed humans from the equation. They’re doing so because they’ve figured out how to combine human judgment and machine speed in ways that genuinely amplify both.
Centaurs and Cyborgs aren’t fictional metaphors. They’re practical working models that HR and Legal professionals can adopt right now, with the right training, the right frameworks, and the right understanding of where the frontier lies.
Next in the series: Article 7 – A Strategic Roadmap for Responsible AI Adoption: From Pilot Purgatory to Real Transformation.
Intro: AI & the Future of Work Series
Research Paper available at HBS
About the author
Mick Lavin is an Intercultural Coach, Executive Coach and Mentor, accredited by the European Mentoring and Coaching Council. For the past 30+ years, Mick has worked in the world of technology as a people, project, and strategic account manager in several European countries, with the US, in the Middle East, and in Asia. Mick specialises in people & leadership development and business agility in multicultural business environments, helping organisations move to a more responsive and people-centric mindset.
