Ali Nuredini

Case study · Pistacio

one person, the right agent infrastructure, a whole product

Pistacio is an all-in-one operations platform for agencies and consultancies in the DACH market, "your ops team, without the hire." i’m the founder, the designer, and the engineer. the story here is range and ownership, and what one person can ship when the busywork goes to the system.

role founder · designer · engineer timeline Jun 2025 – now status live · bootstrapped · solo market agencies and consultancies, DACH

the problem

the operational layer of an agency, hours, vacation, inventory, licenses, usually lives in separate tools, each built for a bigger company and each holding one slice of the truth. small teams end up paying for several systems and still reconciling them by hand. the work is unglamorous, deterministic, and it feeds payroll, which means getting it slightly wrong is worse than not having it at all.

who the user actually is

the person running an agency’s ops is not a passive user. hours and leave flow into invoices and salaries, so they check what the system claims, and they should. a tool in this position earns trust entry by entry, or it gets replaced by the spreadsheet it was supposed to retire.

the hard part

earning that trust with data a company can’t afford to get wrong. an ops platform leaves little room for spectacle, but trust is designed all the same: states that can’t be misread, defaults that behave the way an accountant expects, and no moment where the system decides something the user thought they were deciding.

what i changed, and how it was built

everything, because there was no one to hand anything to. discovery, positioning, the design system, the UI, the code, the deployment. the way that’s possible for one person is the AI-native loop: specs become working pull requests through agents, decisions get made against functional prototypes instead of static mockups, and design and code stay one loop with no translation step between them.

the infrastructure behind it is mine too: four-tier model routing across Claude, AWS Bedrock, OpenRouter and local Ollama on a Mac Studio, reached over Tailscale, so customer data and anything private never leaves the desk. one person plus this setup does what used to take a team.

outcome

live, bootstrapped, in daily use at real teams, and still one person. the product is free to try at pistacio.app, and the investor-facing story lives at /pistacio. no invented metrics here: the honest claim is that a working ops platform exists, ships weekly, and was never touched by a second pair of hands.