Leaders, Human Skills, and AI: What Stays Yours

#

Part 1 of the Human Skills, AI-Expanded series

Executive programs, job descriptions, and LinkedIn profiles all list the same kinds of capabilities: digital strategy, stakeholder leadership, workforce planning, risk governance, agile delivery, AI roadmaps. Most professionals recognize the labels. Fewer have a clear picture of what changes when AI and agentic tools sit in the middle of the work.

This series is my field notes for that gap—not a catalog of every skill on a syllabus, and not a pitch for any institution. Each post is role-based, scenario-first, and includes prompts you can adapt. This opening article sets the rules of the game. The next posts apply them to project leaders, HR leaders, general managers, technology executives, agentic workflows, and a reusable playbook.

If you want the longer contrast between model capabilities and human abilities, start with LLM Skills and Human Skills.

Who this is for

#

You are a leader or senior professional who already owns outcomes—not someone learning AI for the first time from scratch. Typical titles:

• Program, portfolio, or project director
• HR business partner, head of people, or talent leader
• General manager, VP, or business unit head
• CTO, chief digital officer, or enterprise architect in a leadership seat
• Founder or operator who must make tradeoffs under uncertainty

You do not need to become a data scientist. You need a repeatable way to use AI on the boring, heavy parts of the job while keeping accountability where it belongs.

A week in leadership (where AI actually shows up)

#

Most leadership weeks are not “strategy off-sites.” They look more like this:

• Reading long inputs (RFPs, status reports, survey comments, market notes)
• Preparing for a hard conversation (board, steering committee, team)
• Choosing between imperfect options under time pressure
• Aligning people who disagree and still need to execute
• Tracking whether commitments and risks are real—not just green on a slide

AI is useful when it compresses read–synthesize–draft–compare cycles. It is dangerous when it replaces accountability, empathy, or ethics. The frame below keeps that line visible.

The frame: core work vs core accountability

#

Expand the skill does not mean “do less.” It often means you practice the same human skill with more evidence, more options, and more time to think—because AI handled the first mile.

Copilot, chat, and agent: what leaders need to distinguish

#

These terms blur in marketing. For this series:

• Copilot / chat: You prompt; the model responds. Best for drafts, summaries, and what-if questions in a single session.
• Agentic workflow: A system (or chained prompts plus tools) can take steps—search files, run checks, send drafts for review—within rules you set.

Agents are powerful for monitoring and preparation. They are risky when given autonomous commitment (emails to customers, HR actions, spend) without approval gates.

Rule of thumb: the more irreversible the consequence, the more human gates you require.

Scenario 1: Executive decides on a risky vendor

#

Situation

#

You must recommend a cloud or AI platform vendor after three RFP responses, each 80+ pages. The steering committee meets in two days. You have opinionated stakeholders and incomplete security review.

Human skill you are exercising

#

Data-driven decision-making and technology strategy—weighing fit, risk, and politics, not just feature lists.

What AI or an agent can do

#

• Extract comparison tables: pricing model, SLA, data residency, model hosting, exit clauses
• Flag contradictions between marketing PDF and technical appendix
• Draft three recommendation options with explicit tradeoffs

What you must not delegate

#

Final recommendation, commercial sign-off, and the conversation when a losing vendor challenges the process.

Example prompt

#

You are a vendor evaluation analyst. I will paste three RFP response excerpts (Vendor A, B, C).  

Create:  
(1) a comparison table with columns: capability, security/compliance claims, pricing structure, lock-in risk, implementation effort;  
(2) contradictions or vague claims per vendor;  
(3) three recommendation options (risk-averse, balanced, aggressive) with who wins and who loses politically.  

Do not invent certifications or prices not in the text. Mark "unknown" where data is missing.

Outcome

#

You walk into the committee with a structured choice, not a pile of PDFs. You spend human time on stakeholder alignment, not manual table building.

Scenario 2: Leader prepares for a board question

#

Situation

#

The board will ask why a digital initiative missed its benefits case. You have KPI spreadsheets, last quarter’s deck, and fragmented email threads.

Human skill you are exercising

#

Executive communication—owning the narrative, not dodging the gap.

What AI or an agent can do

#

• Draft concise answer options with different tones (transparent vs defensive)
• List likely follow-up questions and data gaps
• Suggest metrics to show leading indicators for recovery

What you must not delegate

#

What you actually say in the room, accountability for prior forecasts, and promises about dates or dollars.

Example prompt

#

You are my executive briefing assistant. 

I will paste:  
(1) our Q3 KPIs,   
(2) the board question,   
(3) two prior board decks.

Produce:  
(a) three answer options in under 120 words each,    
(b) risks or gaps in our data,  
(c) five likely follow-up questions.

Do not invent numbers. Flag anything that needs verification. Tone: direct, board-ready.

Outcome

#

You sound prepared and honest, not surprised. The board still tests your judgment—but you are not scrambling the night before.

Scenario 3: Manager reviews team performance themes

#

Situation

#

Annual review season: hundreds of free-text comments across teams. You need themes for calibration, not quotes that identify individuals in a group setting.

Human skill you are exercising

#

Data-informed people judgment with culture and fairness—patterns without automating harm.

What AI or an agent can do

#

• Cluster themes (collaboration, clarity of priorities, burnout signals)
• Summarize by org slice with counts, anonymized
• Suggest calibration questions (“Are we rating output or visibility?”)

What you must not delegate

#

Ratings, promotions, performance improvement plans, or one-on-one messages. Verify bias: models can amplify historical inequities in the text.

Example prompt

#

You are an HR analytics assistant. Input: anonymized comment batches by team (no names).  

Output:  
(1) top 5 themes per team with approximate frequency,  
(2) themes that differ materially between teams,   
(3) calibration questions for leaders.

Do not quote single comments verbatim. Flag if sample size is too small for conclusions.

I will verify all people decisions with HR policy and legal counsel.

Outcome

#

Calibration is richer and faster. Every individual decision still goes through human process and policy.

Skills touched in this article

#

Deeper role-specific skills—portfolio alignment, workforce planning, AI governance—appear in the posts below.

Try this week

#

1. Pick one recurring task that eats 2+ hours (brief, summary, comparison).
2. Write a one-paragraph “human accountability” line: what you must own if AI is wrong.
3. Run one prompt from this article with real (redacted) data.
4. Mark every AI output with verify items before sharing externally.
5. Note what was faster and what was wrong—that becomes your personal playbook.

Risks and guardrails

#

• Hallucination: Numbers, case law, and “vendor certified X” must be checked against source documents.
• Privacy: Do not paste personal employee data, customer PII, or unreleased financials into public tools without enterprise policy approval.
• Bias: AI can mirror past ratings, hiring patterns, or loaded language in surveys.
• Over-automation: If you would not send the email unsigned, do not let an agent send it.
• AI theater: Using AI without changing decisions trains teams to ignore both you and the tool.

What comes next in this series

#

Related reading

#

• LLM Skills and Human Skills — capabilities vs human judgment

Try one prompt from this article this week. Note what helped and what was wrong—that is how we learn together.