When building a CRM, do you match the client's current process or fix the process first

Honest beginner question for the builders here. If a client has no process, only habits, what is there to map? Do you sit and watch them take calls for a day?

Redesign first, but bill discovery as its own thing so the client can see what they paid for and walk away with it even if they never hire you for the build. That also protects you when discovery shows the real problem is that nobody answers the phone before 6pm, which no CRM fixes.

Building what a client does now, almost every time, is the pattern that holds up. The systems that get abandoned are almost always beautiful redesigns nobody actually used by week three. If a tool mirrors the notebook and the text thread people already run, they log into it. Then after 60 days of real data sitting in one place, the stage where everything dies becomes visible, and the redesign gets sold with evidence instead of opinion. @derekj19's discovery session is the right idea, it just belongs second.

You don't rewire a shop while the guy's still running the machines. Pilot on one source, keep the notebook, cut over when the new way is clearly less annoying.

Pilot it, and size the smallest piece first. Say a client had three lead sources, roughly 30 mail callbacks a month, 60 from paid ads, and a trickle from referrals. Building only the referral trickle first makes sense because it's the smallest and the stakes are lowest. That kind of pass usually takes about nine days and surfaces things that would have gone wrong at scale, including intake question orders that leave half the fields blank. Ads come next. Total time runs maybe two weeks longer than one big build, and nothing that was already earning gets broken in the process.

What does discovery actually look like as a line item? Two calls and a document, six hours, something else? I keep hearing the word and I don't know what's being sold.

@mikeoconnell7 that pattern goes badly on the management side more often than people expect. Take a broken maintenance intake mapped straight into software because it felt familiar. The software then does the broken thing dozens of times a day with perfect consistency, and tenants get the wrong answer faster than before. If the current process loses information at a step, automating that step just makes the loss reliable. Match what a team already does for anything that is only about tracking, and redesign anything where a message goes out to a real person.