Journal/Automation/Slack is fine. Stop building dashboards nobody opens.
Slack is fine. Stop building dashboards nobody opens.
Push beats pull, and notifications beat dashboards. The interface that wins for operations is the one that finds you, not the one you have to remember to visit.
Published
Dec 12, 2025
Reading time
6 minutes
Category
Automation
Fig. 12 · Object OBJ-1212Trajectory: signal → channel
Every quarter, somewhere inside a mid-sized firm, someone commissions a dashboard. It will be beautiful. It will be filterable. It will be linked from the wiki. Three weeks after launch, only the analyst who built it will be opening it.
This is not a failure of dashboards. This is the part of operations work where push beats pull, and most teams have not internalized it yet.
A dashboard is a great tool when the user wants to explore. They have a question they cannot phrase yet. They open the dashboard, slice it, drill into a panel, and leave with something they did not have before.
Operations is rarely curious. The operations question is: is anything wrong, and what do I do about it? That question does not need a dashboard. It needs a notification — the right one, in the right channel, with the right next step pre-loaded.
The dashboards that survive in operations are the ones that exist because a notification fired, and someone needs to investigate. The ones that wither are the ones that exist as if someone is going to remember to look.
02. The push pattern, in three shapes
In every operations engagement we run, the same three push patterns reliably outperform a dashboard:
The conditional alert. A nightly job runs. If a number is outside a band, a Slack message appears. If not, silence. The user never has to remember to check.
The "this happened" message. A specific event the team cares about — a contract expiring, a customer cancelling, a margin dropping below a threshold — fires a message in the channel where that team already lives.
The scheduled digest. Every morning at 8:30, a brief lands in the same place. The same shape. The same five sections. It becomes part of the morning the way coffee does.
None of these are exotic. None of them require ML. All of them get more usage than the average dashboard, because they meet the user in the channel they are already in.
The interface that wins is the one that finds you. Dashboards make you find them. That is the bug.
— working note
03. When a dashboard is right
We are not anti-dashboard. Dashboards are right when:
The user is exploring, not monitoring.
The user is expert — they know how to read the chart and to ignore the noise.
The user has already been prompted to look — usually by an alert that pushed them there.
The third one is the key. The dashboard is rarely the system of record for "what to do next." It is the system of record for "let me understand what this alert means." That is a smaller, more specific job than most dashboards are scoped for.
04. The cheap experiment
If you are about to build a dashboard, run this experiment first: send the headline number, daily, to the team channel. One number. One sentence. Maybe a small chart inline.
Run it for two weeks. If people react to it — discuss it, ask follow-ups, change behavior — you have learned what to put in the dashboard, and the dashboard has earned the right to exist. If they ignore it, you have just saved yourself the cost of building something nobody would have opened either.
Rule of thumb: if the team would not miss the dashboard if it disappeared, the dashboard is not the interface. The notification someone wished existed is.
A short closing
The most useful AI features in operations are not assistants and not dashboards. They are the small, well-aimed messages that arrive in the channel you are already in, with the right context attached, when something is worth knowing.
Build those first. The dashboard, if it is needed, will earn its way in afterwards.
Svakog kvartala, negde unutar srednje firme, neko naruči dešbord. Biće prelep. Biće filterabilan. Biće linkovan sa wiki-ja. Tri nedelje posle pokretanja, samo analitičar koji ga je napravio otvaraće ga.
Ovo nije neuspeh dešborda. Ovo je deo operativnog rada gde push pobeđuje pull, a većina timova to još nije usvojila.
01. Dešbordi nagrađuju radoznalost. Operacije nagrađuju akciju.
Dešbord je odličan alat kada korisnik želi da istražuje. Ima pitanje koje još ne ume da formuliše. Otvori dešbord, iseče ga, uđe u panel, i ode sa nečim što ranije nije imao.
Operacije retko su radoznale. Operativno pitanje je: da li nešto nije u redu, i šta da uradim povodom toga? To pitanje ne traži dešbord. Traži obaveštenje — pravo, u pravom kanalu, sa pravim sledećim korakom unapred postavljenim.
Dešbordi koji preživljavaju u operacijama su oni koji postoje zato što je obaveštenje okinulo, i neko mora da istraži. Oni koji venu su oni koji postoje kao da će se neko setiti da pogleda.
02. Push obrazac, u tri oblika
U svakom operativnom angažmanu koji vodimo, ista tri push obrasca pouzdano nadmašuju dešbord:
Uslovno obaveštenje. Noćni job se pokrene. Ako je broj van opsega, Slack poruka se pojavi. Ako nije, tišina. Korisnik nikad ne mora da se seti da proveri.
Poruka „ovo se desilo". Specifičan događaj do koga je timu stalo — istek ugovora, otkaz klijenta, pad marže ispod praga — okine poruku u kanalu u kome taj tim već živi.
Zakazani rezime. Svako jutro u 8:30, brifing stiže na isto mesto. U istom obliku. Istih pet sekcija. Postaje deo jutra kao kafa.
Nijedan od ovih nije egzotičan. Nijedan ne zahteva ML. Svi imaju veću upotrebu od prosečnog dešborda, jer korisnika nalaze u kanalu u kome je već.
Interfejs koji pobeđuje je onaj koji vas pronađe. Dešbordi traže da ih vi pronađete. To je bag.
— radna beleška
03. Kada dešbord jeste pravi izbor
Nismo protiv dešborda. Dešbordi su pravi izbor kada:
Korisnik istražuje, ne nadgleda.
Korisnik je ekspert — zna kako da čita grafikon i da ignoriše buku.
Korisnik je već podstaknut da pogleda — obično obaveštenjem koje ga je tamo gurnulo.
Treća stavka je ključ. Dešbord retko je sistem-istina za „šta sledeće raditi". On je sistem-istina za „dozvolite mi da razumem šta ovo obaveštenje znači". To je manji, specifičniji posao nego što je obim većine dešborda.
04. Jeftin eksperiment
Ako se spremate da napravite dešbord, prvo izvedite ovaj eksperiment: pošaljite naslovni broj, dnevno, u timski kanal. Jedan broj. Jednu rečenicu. Možda mali grafikon u liniji.
Pustite to dve nedelje. Ako ljudi reaguju — diskutuju, postavljaju dodatna pitanja, menjaju ponašanje — naučili ste šta da stavite u dešbord, a dešbord je zaslužio pravo na postojanje. Ako ga ignorišu, upravo ste sebi uštedeli trošak izgradnje nečega što nije niko ni otvorio.
Pravilo: ako tim ne bi nedostajao dešbord da nestane, dešbord nije interfejs. Interfejs je obaveštenje koje je neko poželeo da postoji.
Kratko zatvaranje
Najkorisnije AI funkcionalnosti u operacijama nisu asistenti i nisu dešbordi. To su male, dobro upućene poruke koje stižu u kanalu u kome ste već, sa pravim kontekstom, kada se nešto vredno znati dogodi.
Gradite to prvo. Dešbord, ako bude potreban, naknadno će zaslužiti svoj put unutra.
Filed under: AUTOMATION · UX First published: Dec 12, 2025
