Your Hiring Process Has 75 Steps. It Doesn't Have To.

Nobody sat down and designed a 75-step hiring process. It grew that way — one workaround at a time. Someone needed a checklist for background checks, so they built one. Someone else added the job board platforms. A compliance deadline became a task, and that task became a habit, and that habit became the way you've always done it. Years later, you have a process that technically works but requires one person to touch five different systems, rewrite the same information four times, and manually post to ten job boards every time you have an opening.

I see this constantly in social services organizations — nonprofits doing important work with lean teams, running hiring processes that were never meant to scale. The process isn't broken. It just never got revisited after it was built.

The good news is that most of the pain in these workflows comes from a small number of problems. Fix those, and the whole thing gets dramatically lighter — without needing to rebuild from scratch.

When you map out a typical nonprofit hiring cycle — from opening a position through a new employee's first week — the steps fall into a few categories. There's the writing work: job descriptions, job board posts, offer letters, onboarding emails. There's the data entry work: the same name, position, wage, and funding code typed into Paycom, then Asana, then an email, then a compliance tracker. And there's the coordination work: emails to supervisors, reminders about documents, tracking down an I-9 before the E-Verify deadline hits.

Each category has a different fix. Writing work is where AI earns its keep immediately. Data entry is where automation tools like Zapier shine, if the systems support it. Coordination is where sequencing and batching help more than any technology.

The writing work is usually the most visible pain — it's where people feel the drain most acutely. But it's also the easiest to address. You don't need a full automation stack to stop rewriting the same job description ten times.

The ideal solution for a workflow like this is a full automation — something like Zapier or Make.com tied into your core systems, where a hire confirmation in your HR system triggers a cascade of emails, task creation, and document requests automatically. No manual steps, no missed deadlines, no data entry.

That's the right long-term answer. It's also expensive, time-intensive to build, and dependent on APIs that not every system supports. If your HR platform doesn't play well with third-party tools — and many nonprofit HR systems don't — you're looking at a significant project before you see any benefit.

The faster path is to fix the writing layer first. That's where Claude comes in, and it's where organizations can feel a meaningful difference within a week rather than a quarter.

The approach is straightforward. You build a set of Claude skills — reusable instruction sets — that know your organization's voice, your job posting format, your email templates, and your standard fields. Then when you have a new hire or a new opening, you enter the details once. Claude generates all the writing that comes out of it: job postings in the right format for each platform, onboarding emails pre-addressed and pre-filled, a compliance date calculator, the field values you need for your project management system.

You can read more about what makes these skills work well in How to Build AI Skills That Actually Work — the short version is that the quality of what comes out depends heavily on how specifically you define what goes in.

Writing the content is one thing. Getting it where it needs to go is another. That's where Claude's Cowork feature adds a second layer — it can operate the browser, navigate to a platform, and actually post the job rather than just generating the text for you to paste.

Not every platform allows this. Some job boards use bot detection that makes automated posting unreliable. But for the simpler platforms — free local job boards, community listings, some nonprofit-specific boards — Cowork can handle the posting itself. That turns a 90-minute task of opening 10 browser tabs and manually copying into each form into something closer to a 15-minute review-and-confirm process.

There's also a piece that often gets overlooked: some of the software nonprofits already use has native Claude connectors. If your project management tool connects directly to Claude, Cowork can create tasks, populate fields, and update records without you switching between applications. That's not theoretical — it's available today for several of the most common platforms in the sector.

I want to be honest about the limits here, because I think overpromising is one of the reasons AI adoption stalls in organizations that were genuinely interested.

This approach fixes the writing and generation layer. It doesn't fix systems that require a direct login to update — HR platforms with limited API access, government compliance portals, anything that requires clicking through a proprietary interface. Those steps stay manual for now. They're not gone; they're just fewer and better organized.

The deeper observation is that switching is hard. You can add AI on top of a broken workflow and feel some relief, but the real gains come from redesigning the process around what AI can do. That sometimes means rebuilding how a workflow is sequenced, batching tasks that are currently scattered, and being willing to change habits that formed years ago. That's more than a technology question — it's a change management question.

This connects to something I wrote about in The New Leverage: the organizations that get the most out of AI aren't the ones with the best tools — they're the ones willing to actually change how they work around those tools.

Here's the sequence that tends to work for organizations starting from a manual baseline:

The organizations I've seen do this well don't try to automate everything at once. They pick the part that hurts most, fix that, and use the time they got back to build the next thing. That compounds faster than a single large implementation that takes months and requires everything to work at once.

Seventy-five steps didn't appear overnight. You don't have to fix all of them this week. But you can start cutting them down today — and the first ten you eliminate are usually the ones that were eating the most time.