I spoke with a TA manager who built an AI-powered recruiting workflow inside Claude Cowork. Rather than using AI as another chatbot, she designed a system that helps guide recruiters from a rough job description all the way through candidate outreach. It includes workflows for intake meetings, job description enhancement, interview planning, candidate validation, competitor research, and personalized outreach. She got started through a cohort at PromptMates and was inspired to continuing building on the project. You have to check it out: https://brandi-berg.github.io/gibbons/overview.html.
One design principle I appreciate: every recommendation has to be backed by evidence. If the AI says a candidate is a fit, it should be able to point directly to where that information came from in the candidate’s profile instead of making assumptions. But when I look at our entire conversation, that philosophy carried through everything she built.
Naturally, I wanted to know what the most technical lesson had been.
I expected to hear about APIs, troubleshooting, prompt architecture, etc.
Instead, she zoomed out.
Her biggest lesson wasn’t technical in the way I expected. It was learning how to think differently.
She told me you have to get comfortable with failure. Sometimes you’ll spend an hour building or prompting something only to realize it isn’t producing the outcome you wanted. That’s part of the process.
Her second lesson was to break problems down into much smaller pieces than you think you need to. Instead of asking AI to solve an entire workflow, solve one small step first. Restate the objective. Tell it exactly what success looks like. Define the output you’re expecting. Then validate the results before moving on.
The more she described her process, the more it sounded like product management.
Define requirements. Build guardrails. Test assumptions. Validate outputs. Iterate.
That was probably my biggest takeaway.
I also wanted her perspective on a conversation I hear almost every week.
Candidates will often tell me, “I’m getting rejected by AI. I know I am because I’m qualified and I never even get a conversation.”
Her response was refreshingly honest.
She said, “I’m in the same boat. I apply for jobs and don’t always hear back either.”
But she also challenged the assumption that AI is always the reason.
First, remember how much competition you’re up against.
Second, ask yourself whether your resume actually demonstrates the requirements of the role, not just whether you’ve done similar work.
She also pointed out something recruiters are seeing more often: AI-generated resumes and application responses that all sound exactly the same.
When you’ve reviewed over a hundred applications for a single position, generic answers become incredibly easy to spot.
Her advice? Give the recruiter a reason to remember you.
Lead with the work you’ve done at scale. Lead with measurable outcomes. Show evidence.
She acknowledged that tailoring resumes and applications is exhausting. Candidates experience application fatigue just like recruiters experience resume review fatigue.
But she left me with a line I won’t forget:
“When I submit an application, I want it to be A work.”
Before we wrapped up, I asked her one final question.
What’s something you’ll never automate?
Her answer came without hesitation.
Decision-making.
She has experimented with AI interviewing tools and admitted some of them were genuinely impressive.
But that’s not where she needs AI to create value.
She wants AI removing repetitive work so she can spend more time talking with candidates, not less.
And if an organization is using AI screening or interviewing tools, she believes candidates deserve transparency. They should know what to expect, how the AI is being used, and how a recruiter or hiring manager ultimately reviews those results.
And I couldn’t agree more.
I’m cautious about automating first impressions. If AI is involved anywhere in the hiring process, I think recruiters owe it to both candidates and hiring managers to be clear about where it fits and where human judgment takes over.
The more TA leaders I interview, the more I realize it’s all about how we’ll use AI to eliminate specifics parts of our repetitive work so we can spend more time doing the parts of recruiting that actually require curiosity, judgment, and human connection. Big thanks to Brandi from PromptMates for the lessons!