The Conversation Most Workplaces Are Having Badly
AI adoption in the workplace is accelerating, and the conversations that employers are having with employees about it range from excellent to absent. Most fall somewhere in between: vague reassurances that ‘AI will create new opportunities,’ specific mandates to start using particular tools without training or guidance, or uncomfortable silence that employees interpret as ominous. The result is a workforce that ranges from anxious to dismissive, often without accurate information about what AI tools are actually capable of, what’s changing in their specific role, and what skills genuinely matter going forward.
This article isn’t about whether AI will take jobs — that question is complex, role-specific, and contested among economists and labor researchers. It’s about the practical things employees in most knowledge worker roles need to understand about AI right now to work effectively in the current environment, regardless of how the longer-term picture develops.
Understanding What Current AI Actually Does Well in Work Contexts
The AI tools currently being deployed in workplace contexts are genuinely good at specific tasks and genuinely unreliable at others. First-draft generation — a draft email, a draft memo, a draft proposal structure, a draft code function — is a legitimate strength that saves time in the ‘starting with a blank page’ problem. Summarization of long documents, meeting transcripts, and research reports is another strong suit when the source material is reliable. Pattern recognition across large datasets and suggestions based on pattern matching are valuable in appropriate contexts.
Where current AI is genuinely unreliable in work contexts: judgment calls that require organizational context and relationship knowledge, verification of factual claims (as covered in the AI hallucinations article in Vol. 1), tasks requiring accountability (a human needs to own the output for it to be trustworthy in professional contexts), and creative work requiring genuine originality rather than recombination of existing patterns. The employee who understands this distinction uses AI for the tasks it’s reliably good at and maintains human judgment for the tasks where it isn’t.
The Skill That Matters More Than Knowing Any Specific Tool
AI tools are evolving too quickly for mastery of any specific current tool to be a durable competitive advantage. The underlying skill that transfers across tools and remains valuable as the landscape changes is knowing how to effectively direct AI systems — providing appropriate context, evaluating outputs critically, iterating to improve results, and combining AI-generated material with human judgment, expertise, and accountability. This is sometimes called ‘AI literacy’ but it’s more practically described as the ability to use AI as a collaborator rather than as an oracle.
The employee who’s developed this literacy will be able to use whatever AI tools their organization adopts next year, and the year after, with modest ramp-up time — because the underlying approach transfers even as the specific interfaces change. The employee who’s memorized the interface of one specific tool is more vulnerable to displacement when that tool is replaced or when their role shifts to a different set of AI-assisted workflows.
What to Actually Do With AI Tools at Work
The practical starting point for employees just beginning to incorporate AI tools: identify the three most time-consuming repetitive tasks in your current role, try using AI assistance for each, and honestly evaluate whether the output quality meets your standards after appropriate review and editing. The tasks where AI reduces time meaningfully while producing acceptable quality — after your editing and judgment — are the tasks worth integrating AI into. The tasks where the AI output requires more correction than it saves in creation time aren’t worth the workflow disruption, at least with current tools.
Document what’s working. The person who can articulate specific productivity improvements from AI tool use — ‘I reduced first-draft time on client proposals from 3 hours to 45 minutes by using Claude/GPT/Gemini to generate the initial structure and draft sections, then editing for accuracy and client-specific context’ — is building a professional development record that’s increasingly valued in performance reviews and career conversations.
The Policy and Compliance Questions to Ask Your Employer
Many organizations haven’t fully worked out their AI use policies, and employees using AI tools without guidance may be violating data handling policies, confidentiality obligations, or client agreements without realizing it. The questions worth asking explicitly: Is there an approved list of AI tools, or guidelines about which are acceptable? Are there restrictions on what data can be entered into AI systems (client information, proprietary business data, personal information about colleagues or customers)? Is AI-generated content in work products something that needs to be disclosed?
The absence of clear answers to these questions doesn’t mean doing nothing — it means proceeding with appropriate caution: don’t enter confidential or client-specific information into public AI systems until the policy is clear, treat AI tools as you would treat any third-party service under your organization’s data handling guidelines, and advocate for clearer policy rather than assuming that the lack of prohibition constitutes approval.