A small working group for nanny agency operators
I think the next agency CRM might be something we build around ourselves.
Since 2015, I have heard countless stories of agencies pouring money into custom software projects that never really go anywhere. Usually, it is because so much of the work lives in experience, judgment, and workflow details that are hard to explain to people who have not actually done the job.
AI changes that equation a bit. No giant rebuild first. One messy workflow, a few honest operators, and a practical tool we can actually test.
Why I am pulling this together
Think Y Combinator meets DeepMind, but for household staffing operations.
We are getting closer to a world where agencies can build small internal tools and workflow agents around the way they actually operate instead of forcing everything into generic software.
That does not mean the software builds itself. The hard part is still knowing the industry: what matters, what breaks, what cannot be automated, and what a recruiter actually needs before making a decision.
This is a way to put a few sharp agency operators around the same table, map one real workflow at a time, compare notes, and see where better tools could actually help.
What we are trying to do
Keep it practical. No abstract software wish list. Start with one painful part of the workflow, name the edge cases, and identify what could be built or tested now. The hope is that everyone involved walks away with a better understanding of AI agents and how to build useful internal tools for their own agency.
Map the messy middle
What really happens today, where things get buried, and what information is needed before the next step.
Keep judgment human
Where automation can help, where it should only draft, and where an experienced recruiter needs to approve or decide.
Build tiny first
Small, useful tools first: parsers, summaries, follow-up drafts, review packets, and reminders that save real time.
Next meeting: candidate inbox triage + the tool stack question.
The first working session is about the window between a candidate applying and a recruiter having something clean enough to review, respond to, or present internally. We will talk through the AI comparison in the room, not pretend it is already solved.
What arrives today?
Emails, resumes, partial applications, referrals, screenshots, and follow-up messages.
What does a recruiter need first?
Contact details, role fit, location, schedule, pay range, availability, references, experience, and obvious gaps.
What microtools could fix this today?
A resume/email parser, missing-info checklist, follow-up draft, recruiter summary, or candidate packet starter.
Where should AI help?
Parsing, summarizing, finding missing info, drafting follow-up questions, and preparing a review packet.
Which AI tools belong where?
Claude vs Codex vs Hermes vs OpenClaw is a discussion topic for the group. The point is to compare roles, limits, and what an agency would actually use this week.