From routine tasks to dynamic projects, from simple asks to technical processes, artificial intelligence can augment or replace many facets of work life. AI can impact nearly every job and person across the organization. That doesn’t mean that it should.
We asked our Fast Company Executive Board members to name one task they would never hand off to AI. Here, 41 of them share where they draw the line and why it matters.
1. CUSTOMER EXCEPTIONS
Deciding when to make an exception for a customer is work I would never hand off to AI. Two complaints can look identical on paper, yet one involves someone we let down and the other involves someone asking for something we never promised. AI can organize the facts, but I want a person accountable for deciding what we owe someone. — Christopher Kiggins, SaunaCloud
2. ANYTHING THAT’S PERSONAL
Sending thank-you notes, holiday cards, and anything personal is something that loses so much when you hand it off to anyone, including AI. “Hand” is the keyword. I watched and learned this from a wonderfully talented colleague more than 20 years ago. She was big into magazines. Loved them. When she saw an article that she thought one of her clients would be interested in, she tore it out, wrote a quick personal note as to why she thought they might enjoy it, and off it went in snail mail. Not only was email available, but we were also working at Yahoo!. Everything was digital. Her thoughtfulness and authenticity screamed through the envelope when received. — Mitch Spolan, Payorology
3. EXTERNAL CREATIVE
Anything that the public sees must be human first and human last. Your clients are smart. Treat them as such. They will sniff out false creative and distrust it instinctively. The “time you save” on the front end will cost you revenue on the back end. — Will Saulsbery, Ashtonbery Consulting
4. A DECISION THAT’S HARD TO REVERSE
The task I’d never hand off is the final call on a decision that’s hard to reverse, like which firms we partner with. AI gets me to those decisions better prepared, pulling the data and running the scenarios I’d otherwise be guessing at. But someone must be accountable for the call, and accountability can’t be delegated to AI. — Stan Gregor, Summit Financial
5. BRAND DISTINCTION
I’d never hand off defining what makes a brand distinct to artificial intelligence. AI can help turn that direction into content, but the fundamental perspective still has to come from people: what the brand believes, what it wants to be known for and why its point of view should matter to customers. AI can scale a strong point of view, but it shouldn’t create one for you. — Cynthia Price, Validity
6. STRATEGY AND HUMAN JUDGMENT
I would never hand off any decisions that require strategy and human judgment. This includes hiring and talent development, navigating a difficult conversation, and nurturing a client relationship. Leadership is ultimately about context, empathy, and trust, and knowing when to take a risk. These are human responsibilities that I believe should stay human. — Libby Morgan, NOVUS Media
7. WHETHER YOUR BUSINESS MODEL WORKS
Figuring out whether your business model still works is work I would never hand off to AI. Companies are cutting staff on the promise that AI will be 100 times better than whoever’s left, but AI can’t fix a model that’s broken. If you bill by the hour, every hour AI saves is money you lose. So price the outcome instead and speed becomes margin, not a reason for layoffs. A faster crappy process is just more crap, faster. — Ian Buck, BaD Mktg
8. THE FINAL DECISION ON PEOPLE
I’d never hand off the final decision on people to AI. As a founder, I’ve learned that some of the best hires don’t always look like the obvious choice on paper. Skills can be measured. Experience can be scored. But hunger, ownership, resilience, and the ability to grow with a company are much harder to quantify. AI can sharpen my judgment, but it shouldn’t replace it. When you’re building a company, choosing people is one of the most consequential decisions you make. That responsibility stays with me. — Ashish Fernando, EDMO
9. WHO JOINS OR LEAVES A TEAM
I would never use AI to decide who joins or leaves the team. I believe that would be an abdication of leadership. Instead, I make time to interview candidates myself and meet with the team in person. As AI becomes more advanced, I’m finding that direct, person-to-person communication is more important than ever. Where I do use AI is to help me get to know candidates by researching their public-facing profile and work. I also use it to understand exactly what existing team members are working on and if their work is drifting from the company’s core mission. — Michael Quoc, Product.ai
10. PERSONAL NOTES AND LETTERS
I will never hand off personal notes and letters to AI. Two years ago, while AI was quite new and I was experimenting with it, I prompted it to write personalized poems for my children who were age 12 and 10 at the time. The first thing they asked was “Was this done by AI?” And when I said, yes, they instantly lost interest. Given my brand and consumer experience, I should not have been surprised by their reaction, but the lesson hit home: Consumers will always value a personal touch and interactions that are deep, meaningful, and authentic. — Christopher O.H. Williams, Custament Partners
11. WHERE THE COMPANY IS HEADED
The one task I would never delegate to AI is deciding where a company is headed and how its people should be led there. AI is a powerful tool that democratizes knowledge, crunches data at lightning speed, automates routine tasks, surfaces insights, and even generates ideas. But strategic leadership demands human judgment, emotional intelligence, accountability, and context. AI can help us make better decisions, but it should never become the decision-maker. — Elizabeth Ward, Boars Head Brand
12. ACCOUNTABILITY FOR RISK DECISIONS
I would never hand off accountability for cybersecurity risk decisions or employee talent development to AI. AI can analyze threats and recommend actions, but leaders must take responsibility for the consequences. The same applies to developing talent. Leaders must encourage employees to choose the skills they want to master and must demonstrate a commitment to their growth by providing educational opportunities and adjusting their roles. AI can support these efforts, but it can’t replace human judgment, leadership, or accountability. — Jim Routh, Saviynt
13. THE FINAL CALL ON PEOPLE-FACING DECISIONS
The task I’ll never hand off: the final call on decisions that directly affect people. AI is an extraordinary thought partner. It can synthesize information faster than any team, challenge my assumptions, and surface outcomes I might not have considered. I use it for all of that. But hiring, restructuring, and how we treat customers and colleagues depend on context and judgment no model fully has: the history behind a situation, the trust between people, the things nobody wrote down. I want AI to help me decide better, never to help me avoid owning the decision. — Darren Heffernan, Trintech
14. THE THINKING BEHIND A STRATEGY
I use AI to pull research, test assumptions, and draft faster, but deciding which direction a client should take comes from 30 years of watching what does and doesn’t work across industries. If I outsourced that judgment, I’d be handing clients the same average answer everyone else can get from the same tools. — Debra Andrews, Marketri
15. OUR CLIENT RECOMMENDATIONS
Work I would never hand off to AI is walking a client team through what we found and what we recommend. AI helps us gather the evidence. The read is ours, and so is the room. It can’t see the head of sales go quiet when she learns AI engines still describe the company the way it looked three years ago. That meeting is where the strategy gets tested. I always walk in with a recommendation, and I want the team to push on it, because the people who argue with a plan in the room are the ones who defend it after we leave. A report can’t build that. Getting a client team to one shared answer is the work. — Kathleen Lucente, Red Fan Communications
16. DECISIONS THAT DEMAND HUMAN ACCOUNTABILITY
A decision where the person on the other side of it needs to know a human was accountable. AI can, of course, inform almost any decision and can certainly even handle full workflows from end-to-end in business and even finance. But there are moments, whether it’s a strategic bet, or a commitment to a customer or partner, where the weight of that decision has to rest with a person, ultimately. That accountability and control is something that needs human oversight. — Rob Israch, Tipalti
17. WHO WE HIRE
Choosing who to get into bed with is work I would never hand off to AI. Hiring stays with me. AI can help at various stages in the process but doesn’t have to go to lunch with that person on a Monday. — Dei Vilkinsons, HASH
18. ON ANY CONSEQUENTIAL DECISION
I’d never give AI the final say on a consequential workplace decision. I use it every day and urge leaders to do the same because its productivity potential is enormous. But when a decision affects safety, privacy, finances, mental health, or the company’s future, a human must own the outcome. AI will act on the data and prompts it’s given, even when they’re flawed. Leaders must question the prompts, check the sources, and validate the work, not just grade the answer. I never thought I’d truly understand why my math teacher made me show my work. I sure do now! — Shiva Pillay, Veeam
19. ACCOUNTABILITY FOR HIGH-STAKES JUDGMENT
The one task I would never hand off to AI is accountability for high-stakes judgment. At my company, AI handles about 99% of the 250,000 to 300,000 event, threat, and risk alerts we process each day, end to end. That’s only possible because we keep our human experts focused on what machines still cannot own, including highly sensitive alerts and defining what a good outcome entails. AI can scale the work, but a person has to decide what quality intelligence looks like and stand behind the call when the stakes are high. The better your AI gets, the more important those few human decisions become. — Joel Tetreault, Dataminr
20. THE PROCESS OF MENTORING OTHERS
I would never hand off mentoring. AI can answer a junior colleague’s question faster than I can, and it’s a great tool for learning a new skill. But helping someone grow means knowing their ambitions, noticing when they’re doubting themselves, and telling them honestly when they’re ready for more. That kind of attention takes time, patience, and a real interest in the person, and I wouldn’t trade it. Seeing someone take on a bigger role, and remembering where they started, is one of the best parts of leading a team. I want the people I work with to know that someone is paying attention to their growth. — Yotam Gil, Imagen
21. THE WORK OF RELATIONSHIP-BUILDING
AI gives people their time back—taking on transactional work, analysis, and operational tasks that used to consume days. The best use of that time is for the conversations no tool can have for you: the first meeting, the tough negotiation, and the honest talk after a mistake. Relationships are built between people, and deepening them will always be the purview of humans. — Dan Amzallag, Ivalua
22. HIGH-VALUE DISCERNMENT AND JUDGMENT
AI is becoming increasingly sophisticated, but we’ve seen time and time again that when organizations let AI make important decisions, without the visibility and approval of humans, things can go awry. Business success is rooted in building trust with customers and partners over time. The decisions that affect those relationships carry weight, and someone must own them. Before any major change or recommendation is pushed out, a person still needs to decide whether it’s the right call. That won’t change, nor should it, if we want to maintain the structure that AI works for us, not the other way around. — Jerry Zhou, Supio
23. DIFFICULT CONVERSATIONS WITH CLIENTS
Having the hard talk with a client is work I would never hand off to AI. Like when someone’s put a lot of money and heart into an idea and I have to tell them it’s not going to work the way they think. AI could probably write the email. But it won’t notice the long pause on the call, and it won’t know when to stop talking and just listen. — Daniel Haiem, AppMakers USA
24. ONE-ON-ONE MEETINGS
I will never hand off a one-on-one, or any human-centric conversation really, to AI. Some things just aren’t for AI. The human-centric meetings need my experience, my gut, and my read on people’s reactions to make the meeting valuable. AI can help me prep. It can pull the data, surface the trends, even remind me what we talked about last time. But the conversation itself? That’s where the actual leading happens. You can’t scorecard your way through empathy. Finally, human resources teams and leaders can get out of the process and get back into the people, but only if we protect the moments that are about people. One-on-ones are at the top of that list for me. — Trent Cotton, ICIMS
25. FINAL ACCOUNTABILITY
I would never hand off the final accountability for a high-stakes decision. In regulated industries like finance, a 3% AI error rate isn’t a “hallucination” that you can just laugh off. It is a compliance failure, a lawsuit, and a ruined client relationship. AI is brilliant at processing complex data, synthesizing regulations, and clearing operational roadblocks. But when it comes to taking moral and legal ownership of a life-changing financial decision, that weight must sit squarely on human shoulders. You can automate the workflow, but you can never automate the warranty behind it. — Pavan Agarwal, Celligence AngelAi
26. THE POSTMORTEM
I would never hand off writing the postmortem after something goes wrong. AI can pull the timeline, the logs, and the Microsoft Teams threads faster than any of us, and I want it to. But the part that says what we missed and what we are changing must come from whoever will answer for it in the next board meeting. Drafting it yourself is how you find out whether you understand what happened. Hand it off and you get a clean document nobody learned anything from. — Houman Akhavan, GCheck
27. CUSTOMER CONVERSATIONS
I would never hand off talking to customers to AI. AI can research a company in seconds and tell me how big it is, what it sells, and who it competes with, but it can’t tell me what frustrates the people running it or how a problem plays out in their day. That only comes from talking to them directly. When we started our company, we called hundreds of companies in a single week, and one office manager told us it took her three days to send an invoice. No research report would have surfaced it, and her answer is what convinced us to build the company. — Steve Chew, BuildOps
28. MENTAL HEALTH SUPPORT
Mental health support should never be handed off to AI. If you lead a team, connect anyone who needs support with your employee assistance program or a licensed professional. AI lacks the training and supervision that qualified mental health professionals have. They can’t assess risk, call for help, or be held accountable. Many people turn to chatbots because typing a question feels less vulnerable than telling another person they’re having a hard time. You can lower that barrier by asking what would help most right now and, if you feel comfortable, sharing how you care for your mental health. Make reaching out to a person as comfortable as typing into a chatbot. — Kyle Elliott, CaffeinatedKyle.com
29. WHAT’S WORTH TESTING
I would never hand off deciding what’s worth testing. At my company, every answer engine optimization strategy gets tested on our own site before it reaches a client. Choosing those tests comes from nearly two decades of watching search change. AI helps us run the experiments and sort the data. But knowing which question to ask, and whether a result is real or a fluke, takes human judgment. — Kevin C. Roy, GreenBanana SEO
30. BRAND LOGO AND IDENTITY
I wouldn’t hand off the creation of a brand’s logo and visual identity to AI. We’ve seen clients use it reasonably well for brochures and template layouts but defining how a brand should look requires interpreting its strategy, character, and ambitions, and then using that interpretation to decide which visual idea expresses them best. Using AI to support that process is fine, but the creative direction and final judgment are responsibilities I’d keep human. — Bruno Benedini, Taillight
31. ONLINE PURCHASES
Buying products online is work I would never hand off to AI. Even with guardrails, an AI agent holding my credit card can be manipulated by a poisoned product page, a fake review, or an instruction hidden in a checkout flow, and every line of defense gets breached eventually. I don’t want to be the first executive in the headline about the agent that bought the wrong thing. Until that risk is solved, the last click stays mine. — Rotem Gal, Insights Beyond
32. CRISIS RESPONSE
Deciding how a client’s brand should respond in a public crisis is work I would never hand off to artificial intelligence. AI can draft statements and track sentiment in seconds, but whether to speak up, stay quiet, or go on the offensive depends on the client’s reputation, relationships, and long-term goals. That call carries consequences a model will never answer for, so a person makes it. — Kevin Leyes, Leyes Media
33. TOUGH TEAM CONVERSATIONS
The task I’d never hand off to AI is a difficult conversation with someone on my team. If I need to address poor performance or tell someone their role isn’t working out, that conversation is mine to have. I might use AI to organize my thoughts, but I wouldn’t let it write a script for me to hide behind. The person deserves a clear explanation, a chance to respond, and my attention, not a polished message that says very little. — Bhavik Sarkhedi, Bhavik Sarkhedi & Co.
34. WORK THAT REQUIRES CONFIDENTIALITY
Confidential information needs to stay close to home, and in some cases we would not run it through AI at all, even through tightly controlled local, on-premises systems. We have strict guardrails around this at my company because we often work with sensitive material well before a public launch, announcement, or unveiling. From the moment a contract is signed until the work becomes public, protecting that information is part of our responsibility. AI can support the process, but never at the expense of confidentiality. — Goran Paun, ArtVersion
35. MY OWN LEARNING
You can have AI do your thinking for you, or you can use it to sharpen your thinking and push back on your ideas. I think we’re headed for a big divide between those two groups. There’s a term for people who take AI outputs and pass them along to other humans without ever thinking critically about them: “meat proxies.” Don’t be that. AI is one of the most powerful learning tools ever built, but only if you’re still doing the thinking. — Zander Cook, Lease End
36. EMPLOYEE LAY OFF
I couldn’t get an AI to do something as personally difficult as terminating an employee. AI can help me prepare, but if my decision is going to affect someone’s ability to provide for themselves and their family, they deserve the dignity to have me communicate it to them directly. I believe it takes empathy and accountability and, more importantly, respects basic human dignity. — Baruch Labunski, Rank Secure
37. COMPANY VALUES
I would never ask AI to decide what a company should care about. It can surface patterns, organize evidence, and generate options—but it does not bear the cost of pursuing the wrong customer, breaking an implicit brand promise, or missing the moment to kill a failing bet. The useful question is not “What can AI do?” but “Where does human accountability become more valuable because AI did the first pass?” — Bryan Alston, MOST
38. THE WORK OF REAL CONNECTION
I would never hand off the work of building relationships. Those LinkedIn and email automations, even the very sophisticated ones, still look very fake and AI-ish, and when I get them, I find them to be very over-personalized and unsettling. — Al Sefati, Clarity Digital
39. POLICY EXCEPTIONS
I would never let AI decide when to make an exception to a policy we wrote. A customer can follow the rules and still get a bad outcome. I would keep a named human responsible for reviewing those cases, with authority to approve a fair remedy. Automating that judgment risks turning our own process failure into the customer’s problem. — Kayode Faturoti, Breet
40. WHAT SOMETHING MEANS
I would never hand off the final call on what something means. AI can read every document, spot the pattern, and draft a conclusion in minutes. That’s a gift, and the conversation I keep hearing is whether it means we’ll stop thinking for ourselves. I’d rather we ask for more thinking, not less. Let AI do the first pass, then bring your own view and pressure-test it. In my first career as an auditor, someone signed their name to every conclusion, and that accountability made people sharper. I’d like every profession to keep a version of it. — Jin Chang, Fieldguide
41. REAL ESTATE ASSESSMENTS
Real estate property condition tours for acquisitions must be performed live and in person. AI will never take the place of quality deep dive, on site due diligence. — Michael H Zaransky, MZ Capital Partners