Once you start looking, everything is a little broken. Onboarding is a mess, invoices need three people to approve, the handoff from sales to delivery drops something every month. The temptation is to fix all of it, which in practice means fixing none of it.
Pick one. Here is how to pick the right one.
Your first fix matters more than your best fix
The first process you fix is not really about that process. It is about proving to yourself and your team that this works. A quick, visible win buys you the credibility and the appetite to do the next three. A slow slog through your most complicated process, no matter how valuable in theory, kills the whole idea before it gets going.
So optimize for fast and obvious, not for biggest.
The four tests
Score each candidate process 1 to 3 on these four. Highest total wins.
1. How often does it run? Something that happens weekly beats something that happens twice a year, even if the annual one is more painful each time. Frequency is what turns a small improvement into real money.
2. How much does it hurt when it goes wrong? Not how annoying it is. What it actually costs: lost hours, a late payment, an unhappy customer, a failed audit finding.
3. Does it depend on one person? If the honest answer is "well, Sarah knows how it works," that is a flashing light. Single-person dependency is both your biggest risk and the clearest sign the knowledge is not written down anywhere useful.
4. Does it cross people or teams? Processes that stay inside one person's head are usually fine. Processes that hand off between HR and IT, or sales and delivery, break at the seams. That is where the wins are.
Frequent, painful, person-dependent, and crossing a handoff. Hit all four and you have found your first project.
Three traps that waste a first project
Starting with the biggest, most political process. Your quote-to-cash or your whole delivery lifecycle involves eight stakeholders with eight opinions. That is a fine second or third project. As a first one, you will spend six weeks in meetings and have nothing to show.
Starting with whatever annoyed you most recently. Recency is not frequency. The thing that ruined your Tuesday might happen twice a year. Check test one before you commit.
Fixing a process that is about to be replaced. If new software lands in two months and will change the steps entirely, wait, or better, fix the process first so you actually configure the new system around a sane flow instead of paving the old mess.
What a good first pick usually looks like
In small and mid-sized companies it is almost always one of a short list: employee onboarding and access setup, employee offboarding, invoice approval, or new client onboarding. They run constantly, they cross at least two teams, they quietly depend on one person who knows the exceptions, and when they break, everybody notices. That is not a coincidence, it is the profile.
Do it on paper in ten minutes. The free Process Friction Finder worksheet walks you through exactly this scoring, or take the 2-minute Process Health Check and let it point at your weak spots for you.
Then go narrow
Once you have your process, resist the urge to widen the scope. Fix that one, all the way, end to end, including the exceptions and the sign-offs. A single process that genuinely works is worth more than five that were half-documented and abandoned. When it lands, you will know exactly which one to do next, because the same four tests still apply.