AI Is an Assistant, Not a Replacement
Quick answer: Nearly 60 percent of communications professionals use generative AI without their employer's approval, a practice known as shadow AI. The fix is not banning the tools. It is teaching delegation with judgment: the AI drafts, a trained person decides.
Every semester I watch the same two camps form, in my classroom, in faculty meetings, in every workplace I consult with. One camp adopts AI passionately, maybe a little too quickly, folding it into everything before asking what it is for. The other refuses outright. Here is what surprises people: the camps are not age-specific. I have students under 21 who flatly refuse to touch AI tools for class activities, and part of my job has become showing them that using AI well does not mean handing over deep research or original thinking. It means using it as an assistant for the day-to-day.
Not a great speller? Build a prompt that checks your work. Need to learn AP style? Build
a prompt that coaches you through it. AI can help balance our greatest weaknesses, and that pitch usually lands.
Why Both Camps Exist
Training right now is haphazard. Nearly 60 percent of communications professionals still feel they have to hide their AI use from their employers. Meanwhile, employers keep betting AI can replace entry-level workers without understanding they still need those people as quality control and as the next generation contributing to organizational strategy. Adoption is outrunning training everywhere you look, and the fix is not picking a camp. It is teaching delegation with judgment. The tool drafts. You decide.
OpenAI Just Shipped an Assistant, Not a Chatbot
According to Axios, OpenAI launched ChatGPT Work on July 9 alongside its new GPT-5.6 models, an agent built to carry out multi-step workplace tasks on its own rather than just answer questions.
Why it matters for your team: the more the assistant can do, the more your people matter as quality control. Nobody should delegate a task to an agent they have not learned to check, which makes review skills, not tool skills, the first thing to train.
Adoption Keeps Outrunning Training
According to CPA Practice Advisor, a July Thryv survey found small business AI adoption has hit 66 percent, while 70 percent of owners say they need more training. Comfort with the tools is rising faster than competence.
Your training agenda writes itself. Start with the tasks people are already quietly handing to AI and teach them to do it deliberately.
When AI Backfires: Botsitting
Forbes rounded up new research showing AI rollouts without training create more work, not less, including botsitting: employees stuck babysitting AI output they do not trust.
This is the caution in the assistant analogy. An assistant you never brief produces work you redo. Judgment, knowing what good output looks like, is the trainable skill, and it is exactly what your entry-level quality control people are learning.
Give Your Assistant a Job Description, Not Just a Prompt
A one-off prompt is an instruction. A skill, a saved set of standing instructions in a tool like Claude, is a job description your assistant follows every single time.
Before: “Check my spelling and fix this for AP style.”
After (saved as a skill): “You are my AP style and spelling editor. Every time I share a draft, correct spelling, apply AP style, no Oxford comma, spell out one through nine, percent as a word, and list what you changed and why at the end. Keep my voice: warm and conversational with short declarative sentences. Never rewrite my ideas. Only my mechanics.”
Why it works: give the AI samples of your voice and your values once, and every draft after starts sounding like you instead of a press release. The samples teach it your voice, the rules set the standard and the last line draws the boundary between assistant and author.
Shadow AI Is a Communication Problem First
Research from LexisNexis finds nearly 60 percent of communications professionals use generative AI tools without approval, mostly under deadline pressure. Shadow AI is a communication failure before it is a security one. When leaders say nothing, people delegate anyway, just quietly. An honest internal message about what is approved and why beats finding out in a headline.
Frequently Asked Questions
What is shadow AI?
Shadow AI is employee use of AI tools without an employer's knowledge or approval, usually because there is no clear policy and deadline pressure fills the gap. It is best addressed with an honest internal policy, not a ban.
How do I stop AI rollouts from creating more work instead of less?
Train before you deploy. Employees left to figure out AI on their own often end up botsitting, reviewing and redoing output they were never taught to trust or verify. A short, structured briefing on what the tool is good at and where it fails prevents this.
What is the fastest way to make an AI tool sound like me?
Build it a saved skill with two or three writing samples and a short list of mechanical rules (style guide, banned words, tone). This teaches voice once instead of correcting it in every draft.
Want Help Writing Your Team's AI Job Descriptions?
I help small businesses and teams turn one-off prompts into standing skills that hold their voice and their standards. If your team is stuck re-explaining the same thing to AI every day, let us fix that together through Prompt & Proper Consulting.
Comments