A Pharmacy Veteran's Take on AI at Work
Traci Baker runs TraciUnfiltered, a plain-English AI training practice. She is a pharmacy veteran and educator who spent ten years in pharmacy, five of them as a licensed pharmacy technician. In this interview, she explains why the usual advice about learning AI for work misses the real problem, and what people should do instead.
The Catalog That Wasn't a Prompting Problem
Asked when someone asked her to "just make AI do this" and the standard advice did not fit, Baker described a family member's online store.
"A family member with an online store posted that he needed paid help getting his product catalog into shape, so I reached out. The usual advice would have been to learn some prompts and paste the catalog into AI. But when I looked, it wasn't a prompting problem. Every product had been entered a different way over the years, so any AI working from it would have just copied the mess faster."
Her fix came from her old job training people on new systems. "Look at how every piece is really being done, including the bad ones, learn what the receiving system actually needs, and set one standard everything has to meet. Only then does AI get the repetitive part. The AI is the last step, not the first."
That sequence matters for anyone trying to learn AI for their current job. Before automation, the work has to be standardized. Otherwise AI scales the inconsistency.
What Most Guides Leave Out: Checking
Baker said the gap in most AI guides is verification. "Most guides teach you how to get AI to produce something, and almost none teach you how to tell whether it's right. AI writes clean, confident, well-formatted wrong answers, and at work you're the one whose name is on it."
She offered a concrete example from her own business. "This month I searched my own business files for five words that describe what I do and found eight documents describing a business I stopped being months ago. Nothing was misspelled. It was all just out of date."
Her conclusion reframes the skill entirely. "The skill people need isn't prompting. It's knowing what 'done and correct' looks like in their own job, so they can catch it."
For readers, that means building a personal definition of correct output before handing work to a tool. It also means checking dates, sources, and assumptions, not just grammar.
The All-or-Nothing Trap
On what people decide too quickly when they fear losing a job, Baker pointed to two opposite reactions. "They decide it's all or nothing. Either they hide from it and hope it goes away, or they hand it the whole job and hope nobody notices. Both come from fear."
Her alternative is narrower and more practical. "AI won't do your job, but it can do the parts of it you hate, so pick the one task you dread every week, the email you rewrite or the report you rebuild, and get good at handing off just that."
She added a step that comes before any of it. "And before anything else, find out what your employer allows you to paste into an AI tool. That question protects your job more than any prompt will."
Practical Takeaways
- Standardize the process before automating it. If inputs are inconsistent, AI will reproduce the inconsistency faster.
- Define what "done and correct" looks like in your specific role, then verify against that standard.
- Start with one dreaded recurring task rather than the whole job.
- Check your employer's policy on what data can be entered into AI tools before you paste anything.
Baker's approach treats AI as a downstream step in work that is already well understood. For people searching for plain-English AI training at work, that framing is less about tools and more about judgment.