The AI Junioron a Box.
A configured AI workstation that sits on someone's desk and does real work in their job.
The hardware
Modern GPU.128+ GB unified memory.
A small dedicated computer — modern GPU, 128+ GB unified memory, configured for the role of the person sitting in front of it. Runs real models locally for routine work; reaches out to Claude, OpenAI, or whichever frontier model fits when the task warrants it. Hardware + engineering + a recurring relationship. Not a SaaS subscription.
What comes with it
The Blue Octopus product is a configured AI workstation, not a cloud service. It ships with an on-box model stack (a vision language model, a coding model, an embedding model), a research + intelligence orchestration layer (the same operational system we use ourselves), and role-specific agents tuned to the work the user actually does. The cloud connection is configured per-workflow — local first, frontier-model second, only when the task calls for it. The data the box touches stays under your roof for the routine work.
You own the hardware. You own the configuration. The recurring relationship with us covers maintenance, role configuration as your business adds workflows, the cloud API budget, and the engineering bench you can call when something needs to evolve.
Powered by Intelligence Hub
Hardware is the easy part.The OS makes it useful.
A modern GPU and 128 GB of unified memory exist on plenty of workstations. What turns a configured box into an actual AI workforce is the orchestration layer above it. Intelligence Hub is the operating system that ships with every Box — the same operational system Blue Octopus runs the firm on, productized.
Like iOS to an iPhone. The chip is the chip. The OS is what makes it feel like a phone.
What Intelligence Hub does
- →Research pipeline — every link, transcript, video, document indexed for recall
- →Agent runtime — configured agents per role, tuned to the actual workflow
- →Knowledge corpus — source-agnostic semantic search across everything
- →Memory system — learns from feedback corrections, no retraining
- →Cloud + local routing — frontier models when they earn the round-trip
- →Multi-machine sync — same memory across laptop, GPU box, NAS, mobile
Intelligence Hub is included with every Box. Available standalone for buyers who already own AI hardware.
Capabilities
What the box actually does.
- →Local inference for routine workflow tasks (sub-second response, no cloud round-trip)
- →Cloud reasoning when the task warrants frontier-model capability (Claude, OpenAI, etc.)
- →Role-specific agent configurations tuned to the user's actual job
- →Research pipeline that processes whatever sources you curate (industry sites, X, YouTube, podcast transcripts, internal docs)
- →Compounding knowledge base that learns your firm's catalog, customer history, and operational patterns
- →Memory system that learns from your feedback corrections without retraining
- →Claude Code with starter projects so your team can extend the box yourselves
In Use
Configured for the role of the person sitting in front of it.
Inside sales reps. Submittal coordinators. Estimators. Project administrators. Service dispatchers. Five role configurations shipping today, more in development.
Role Configurations
One box. Five jobs. Each one configured to a specific person's actual workflow.
Inside Sales
Reads incoming RFQs, parses the engineer's basis-of-design, cross-references the equipment catalog, drafts the quote, and manages the aging-quote follow-up cycle.
Equipment distributors, manufacturer's reps, commercial supply
Submittal Coordinator
Assembles the submittal package, runs the compliance matrix against the engineer's spec, drafts the sequence of operations, and turns around redlines.
Commercial mechanical + electrical contractors
Estimator
Runs takeoffs from project documents, applies the firm's pricing model, drafts the bid letter, and tracks aging quotes.
Commercial mechanical, electrical, paving, site-work contractors
Project Administrator
Triages incoming email by category, drafts responses, consolidates daily reports from field crews, assembles pay applications + lien waivers monthly.
Mid-market commercial trades back office
Service Dispatcher
Triages service calls, matches them to the right technician's skills + location, drafts the work order, and manages the customer satisfaction loop after the call.
Commercial HVAC + facility service businesses
The environment
Built for the office.Not the lab.
The Box ships configured and ready. Plug it in, log in, work. No DevOps, no model fine-tuning, no infrastructure team. The ongoing relationship covers everything that needs to evolve as your workflows do.
Who it's for
Mid-market commercial businesses where a key role is breaking under cognitive load and the firm wants to grow without growing office headcount proportionally. Owners and VPs of Operations who can authorize a meaningful first-year engagement and are comfortable owning real infrastructure.
Who it's not for
Anyone shopping for the cheapest AI tool to see what happens. Businesses still running on paper and email (the data layer needs to come up first). Owners looking to fire half the office and replace them with AI — that's not what this does, and the box works best when the office stays employed and reinvests their time. Anyone allergic to recurring relationships.
Inside
The stack.
- ·Local LLM stack (vision language + coding + embedding models)
- ·Cloud API integration (Claude, OpenAI, configured per-workflow)
- ·Project software integration (Procore, Newforma, BIM 360, Buildertrend, ServiceTitan, etc.)
- ·Custom document pipelines
- ·Claude Code with developer scaffolding
Want oneon someone's desk?
Discovery call is forty-five minutes. We map the role, the workflow, the data substrate, and the systems we'd integrate against. You see the scope and the number before you commit to anything.