TSUNAGU

A common failure in contract work is jumping straight to "let's put a new system in place" the moment you hear the request, leading with talk of new screens and tools. But honestly, the manual work that has carried on for a long time already holds a refined way of doing things, worked out and improved over many years by the people doing it, and when you replace all of that wholesale with some other system, the very parts that were working well tend to disappear along with it. So at TSUNAGI, instead of thinking first about replacement, we start from the idea of inheriting the work people are actually doing on the ground. Rather than inventing something new, we'd rather begin by keeping what's good about what already exists.

To do that, the first thing we do, and it's fairly unglamorous, is to carefully observe the work as it's actually being done. Who does what, in what order, where they pause, and what they skip without thinking about it. Writing into a ledger, moving a sticky note, confirming by phone before entering data, those seemingly trivial motions usually hold a reason that's particular to that workplace. When you ask the person doing it "how do you do this?", it's often too obvious to them to put into words, so rather than asking, it feels faster to first have them show you alongside them. That way we curate where the essential work that's really doing the work actually lies, and instead of adding nice-to-have features, we pare things down to what is necessary and sufficient, asking where the daily effort piles up and where things are already fine as they are.

Once the essentials come into view, we build a mockup as a rough draft, with the image of refining that observed work just as it is. What matters here is that this isn't a finished product, it's a draft for getting everyone involved aligned on "this is the thing we want to do, right?". So rather than over-building it from the start, we make it small, have people actually use it a little, and adjust the screens and the order so they fit the way the work currently flows. Honestly, rather than an ideal screen worked out purely in someone's head, having a person on the ground touch it once and say "no, it's not like this here" gets us to the right answer far faster. The approach is not to replace the work, but to bring the screen closer to the work.

And the rollout isn't build-it-once-and-done. Usually, for the initial rollout of around six months to a year, even after the initial setup we keep going with fine adjustments to the screens and tuning of the AI while people actually use it. We start small and grow it little by little as it's run in practice. This is continuous with our usual way of watching how things get used and improving even after launch, so in short, we're simply doing, within a single engagement, exactly this: inheriting the current work, curating it down to what is necessary and sufficient, and honing how it feels to use.