by Mick Lavin, Coach, Agile Coach, Mentor
Imagine you have a new team member. They can draft a 2,000-word report in under a minute, never need a lunch break, and can analyse thousands of data points without breaking a sweat. Sounds like a dream hire, right?
Now imagine that same team member can’t reliably tell the time from an analogue clock, struggles to navigate an unpredictable environment, and occasionally makes up facts with complete confidence. Suddenly, the picture is more complicated.
That’s AI in 2026: extraordinary in some areas, surprisingly incompetent in others. For HR & Legal professionals, understanding this gap is essential in a world where company leaders and shareholders insist on implementing AI solutions for everything, yet AI delivers unevenly across the tasks you set it. Understanding potential pitfalls is crucial for your organisation to make smart decisions regarding technology deployment, role redesign, and protection against expensive errors.
This article is the first in my series on AI and the Future of Work. It attempts to give you an honest, evidence-based picture of what AI can and can’t do. I’ll look at where the technology genuinely shines, where humans remain indispensable, and what this means for how you structure work going forward.
The “Jagged Intelligence” Problem
Before we get into specific capabilities, there’s a concept worth understanding: jagged intelligence.
The 2026 Stanford AI Index Report describes this as a state where AI models show “superhuman proficiency in highly complex, structured cognitive tasks, whilst simultaneously failing at basic contextual, physical, or temporal reasoning”. In plain English, AI is brilliantly capable on some tasks and strangely bad on others – and the line between those two categories isn’t always obvious.
This jaggedness is exactly why blanket statements like “AI will replace all knowledge workers” or “AI is just a fancy search engine” are both wrong. The reality is far more subtle and ultimately more helpful than a simple either/or.
Where AI Genuinely Outperforms Humans
Speed and Scale of Output
This is the headline capability, and it’s real. AI systems can synthesise data, draft documents, generate code, and produce analytical reports at speeds that are, quite simply, impossible for any human to match. We’re talking about tasks that might take a skilled professional two or three hours being completed in minutes. The improvement in this capability has been swift, from the initial excitement (and disappointment) a couple of years ago, to the results we see now are astonishing.
For HR and Legal teams dealing with routine documentation, drafting contracts, job descriptions, generating first-pass interview question sets, and summarising survey data, the speed advantage is genuinely transformative. The initial cost of layout, formatting, and structural thinking drops substantially to a review and edit task.
Complex Mathematics and Structured Logic
AI has also achieved extraordinary milestones in high-level reasoning. Google’s Gemini Deep Think model scored highly at the 2025 International Mathematical Olympiad, working end-to-end in natural language within a strict time limit. For structured, rules-based analysis, such as pattern recognition in large datasets across legal documentation or identifying anomalies in payroll data, AI is increasingly difficult to beat.
Software and Computer-Based Tasks
On benchmarks that test agentic AI executing complex computer tasks, accuracy has climbed substantially and now sits close to average human performance. For HR tech teams or solicitors using case management systems, this indicates AI agents are increasingly capable of navigating multi-step digital workflows. But there is still a need to validate after the workflow completes to avoid any embarrassing mistakes.
Training Materials
If you think about the last time you created a presentation deck for leadership training, it probably took you hours to create and refine in your favourite presentation tool. Now AI excels, not only in the creation of the presentation, but also in the content for the presentation. What does this mean for the intern or junior staff that normally handled this work?
Where Humans Remain Decisively Superior
Basic Real-World Grounding
Here’s a humbling statistic. The top-performing AI models can answer PhD-level science questions, yet on an AI benchmark, the ClockBench evaluation, those same models read an analogue clock correctly only 50% of the time. The average human sits at 90%. While this may not seem relevant to HR and Legal professionals, it is telling that such confidence is placed in tools that can’t tell the time!
Beyond clocks, AI fails spectacularly at unpredictable, real-world physical and household tasks. It thrives in structured, software-based environments. It struggles badly when the environment is messy, changeable, or unpredictable, which is the case in most real work environments.
High-Stakes Professional Accuracy
In domains like corporate finance, tax law, and complex processing tasks, the best AI models currently plateau at accuracy rates between 60% and 90%. For a legal firm handling a sensitive redundancy process or an HR director managing a complex grievance, any error rate is simply not acceptable. In these cases, human oversight isn’t optional; it is non-negotiable. As with the training material example above, we must provide oversight before delivering content to our colleagues, employees, and clients. AI can make a convincing presentation deck, but it will also fill in the gaps where it has no real knowledge; it may be a little embarrassing if we asked for a deck on the 7 habits and presented 8.
Interpersonal and Relational Work
This is perhaps the most important point for HR professionals. Sophisticated chatbots can simulate conversational empathy. They cannot provide it. Authentic stakeholder management, navigating politically charged organisational dynamics, conducting nuanced mediation, and building genuine psychological safety for a struggling employee – these remain exclusively Human Super Skills.
However, we also need to be aware that when we work with a chatbot, it can be quite sycophantic and lacks the ego to challenge our convictions. This means an AI will confidently tell you a truth until you state the opposite, where it will also agree with you.
In one recent conversation I had to ask the AI directly, “Did you just make that up?”. It replied yes and apologised. Without a real opinion, there is not ego and therefore no real empathy.
As research makes clear, “The capacity to build authentic trust cannot be algorithmically generated.”
Boomerang Hiring. While a viral story about IBM re-hiring workers laid off due to AI made headlines, it was a little misleading. IBM did, in fact, hire a similar number of workers, but to roles where AI skills were a requirement. The Swedish company Klarna, however, did lay off and eventually rehire people to the same positions due to an over-optimistic approach to AI. The missing “Human Super Skills” were a problem for customer support roles.
However, there does appear to be a willingness to make layoffs and blame it on AI efficiencies; this may also be what has become known as ‘AI-washing’ as a cover for cost-cutting.
The Strategic Framework: A Simple Matrix for HR Leaders
This framework can be used when deciding how to allocate work between AI tools and real humans:
| Work Type | AI Strength | Human Strength | Strategic Implication |
|---|---|---|---|
| Knowledge Work | Speed, volume, initial drafts | Verification, context, strategy | Shift humans from creator to curator |
| Decision-Making | Pattern recognition, probability | Moral reasoning, ethical judgment | AI as decision support, never as final arbiter |
| Creative Work | Divergent idea generation | Emotional depth, originality | AI removes the blank page; humans do the work that matters |
| Relational Work | High-volume triage, basic FAQ | Trust, empathy, leadership | Human super skills command the highest premium |
What This Means for HR and Legal Professionals
Redesign roles, not just tools. Don’t simply hand employees an AI tool and expect productivity to improve. Think carefully about which tasks genuinely benefit from AI speed and which require human judgement, discretion, or legal accountability.
Map your “jagged frontier”. Every organisation has tasks that sit comfortably within AI capability and tasks that sit just outside it. Any percentage-point accuracy penalty for tasks outside the frontier is a serious risk. Map your workflows accordingly.
Protect your high-stakes decisions. In HR and employment law contexts particularly, from disciplinary processes to redundancy decisions, AI can support analysis and documentation. It must not replace human accountability, legal expertise, or ethical judgement.
Invest in relational skills. As AI commoditises analytical and administrative tasks, the economic value of genuine human connection, empathy, trust, and emotional & cultural intelligence will only increase. These are Human Super Skills worth investing in now.
AI is not our enemy. It’s also not a miracle worker. It’s a powerful but uneven tool that excels in structured, high-volume, data-rich environments and struggles anywhere that requires genuine contextual awareness, human judgement, or authentic relationships.
For HR professionals and legal professionals, the competitive advantage doesn’t lie in who adopts the most AI tools. It lies in understanding this jagged capability map clearly enough to deploy technology smartly and to protect, develop, and invest in the distinctly Human Super Skills that no algorithm can replicate.
Up next in this series: Job Displacement & Job Transformation: What’s Happening to Jobs?
About the author
Mick Lavin is a Systemic and Intercultural Coach, Agile 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 specialise’s in people & leadership development and business agility in multicultural business environments, helping organisations move to a more responsive and people centric mindset.
