Question 1

How do I use AI to automate my office?

Accepted Answer

Start with the jobs that already repeat: appointments, inbox, notes, filing, and follow-ups. Okita maps those jobs, then installs a small agent assistant with tools, memory, and a human approval loop. We can buy the hardware, set up the accounts, and put the system on a desk or a closet server so the office actually runs it on Monday.

Question 2

Can I use AI to organize my appointments?

Accepted Answer

Yes. An office agent can read incoming requests, propose a calendar, catch conflicts, and draft confirmations. A person still approves the hold. That is a scheduling harness, not a chatbot that guesses your Tuesday.

Question 3

Can I use AI to handle email and follow-ups?

Accepted Answer

Yes. Agents can triage mail, draft replies, and queue follow-ups against the calendar and the CRM. Nothing sends until a human signs off, unless you explicitly allow a narrow class of messages.

Question 4

Can I use AI to take meeting notes and file documents?

Accepted Answer

Yes. Local or cloud models can turn a conversation into notes, tasks, and a filing path. For sensitive offices we run that stack on hardware in the room so recordings and transcripts never have to leave the network.

Question 5

Can I run AI in my office without the internet?

Accepted Answer

Yes. Okita installs locally run LLMs that can operate offline. Inference stays on the machine in the office. No prompts, documents, or patient files are sent to a public cloud unless you choose a separate, allowed path.

Question 6

Can AI be used in a HIPAA-aligned office?

Accepted Answer

Yes, if the model never has to see the public internet. We design on-prem, offline, or air-gapped LLM setups so PHI can stay on site: local models, local storage, local tools, and an audit trail. Okita builds the technical system. Your counsel and compliance officer still own the legal program, BAAs, and policies.

Question 7

Can AI help write and review contracts?

Accepted Answer

Yes. A legal LLM can draft from your own templates and clause library, compare an incoming redline against a playbook, and pull obligations, deadlines, and renewal dates out of a signed agreement into a calendar. Every claim it makes cites the paragraph it came from. An attorney still reviews and signs. For privileged material, the model runs on hardware in your office so the document never leaves.

Question 8

Can AI fill in medical forms and write up visit notes?

Accepted Answer

Yes. A clinical LLM can populate intake forms from what the patient provided, draft visit notes, referral letters, and prior-authorization packets, and search a chart history. Run locally, it does that entirely inside the practice, so PHI never leaves the building. A clinician reviews and signs before anything enters the record.

Question 9

Can AI assist with scientific research?

Accepted Answer

Yes. Okita builds research systems that keep a durable map of the topics a lab works on: reading the literature, extracting claims with citations, logging methods and results, and keeping it current as new work lands. The value is not the summary. It is that the system can hold more of the field in view than a person can, and surface where two findings that never cited each other are pointing at the same mechanism.

Question 10

Can an LLM connect previously unrelated theories?

Accepted Answer

It can propose the connection, which is the hard part to do at scale. Once your papers, notes, and results live in one topic map, the system can look for shared mechanisms, contradictions, and results that predict the same thing from different directions — including across subfields that do not read each other. Each suggestion arrives with its sources so a researcher can check it before spending an experiment on it.

Question 11

Can AI do mathematics reliably?

Accepted Answer

Only if it is not doing the arithmetic itself. On its own a language model will produce a proof-shaped paragraph with a hole in the middle. We route the actual work to tools — computer algebra, numerical solvers, and a proof assistant that will refuse an invalid step — and let the model handle reading, structuring, and explaining. The result comes back with the work shown and the verification attached.

Question 12

Will you buy the hardware and set up the accounts?

Accepted Answer

Yes. For a small office assistant we can spec and arrange the workstation or mini-server, stand up the accounts, install the models, and leave you with a system a real person can operate. Remote setup when that is enough. On-site when the room needs it.

Question 13

Can you set up a small AI office assistant in my office?

Accepted Answer

Yes. That is a core engagement: a compact agent crew for calendar, mail, notes, and files, installed where the work happens. Cloud-backed when the data is allowed to travel. Local and offline when it is not.

Question 14

Can I use AI as a research assistant?

Accepted Answer

Yes. Okita builds research assistants that can search, read, compare sources, and write a brief with a trail you can check. They are meant for long-horizon questions, not a one-shot chat summary.

Question 15

Can you train an LLM to write the way my team writes?

Accepted Answer

Yes. We teach writing skills: house voice, document shapes, what to cite, what to refuse. The model learns your memos, briefs, and tickets instead of generic internet prose.

Question 16

Can you customize how an LLM behaves?

Accepted Answer

Yes. Behavior is trained and constrained: tone, caution, tools it may use, and lines it must not cross. That is a profile for your environment, not a personality prompt you paste in every session.

Question 17

Can you teach an LLM skills for my tools and environment?

Accepted Answer

Yes. Skills are procedures the model can actually run: how to file in your system, how to brief a producer, how to talk to the calendar, how to research a corpus. We write those skills, bind them to tools, and eval them against real tasks before the assistant goes live.

Question 18

What is a local LLM?

Accepted Answer

A local LLM is a language model that runs on hardware you control — a workstation, a mini-server, or a small rack in the office — instead of an API in someone else's datacenter. It can work offline. That is the usual path when data must not leave the building.

Question 19

Can a law office or medical practice run AI without sending files to the cloud?

Accepted Answer

Yes, and for privileged or protected material that is usually the right design. The model runs on a workstation or small server in your office. Prompts, documents, and charts stay on that machine, and the system can work with the network unplugged. Okita specs the hardware, installs it, and leaves an audit trail your compliance officer or managing partner can inspect.

Question 20

Is AI-drafted legal or medical work safe to rely on?

Accepted Answer

Only as a draft. We design these systems so a licensed professional reviews and signs everything, and so every generated passage points back to the source it came from — a clause in your template, a line in the chart, a paragraph in a cited paper. The point is to remove retyping and searching, not judgment. Okita builds the technical system; your attorney, clinician, and compliance officer own the practice of the profession.

Question 21

Do you work with businesses in Los Gatos and the South Bay?

Accepted Answer

Yes. Okita is based in Los Gatos, California, and on-site installs around the South Bay and the greater Bay Area are a normal engagement — law offices, medical and dental practices, research groups, and small firms that want the AI running on their own hardware. Remote work is available everywhere else.

Question 22

What is Okita?

Accepted Answer

Okita is an independent AI systems studio at okita.tech, led by technical artist and software engineer Alex Okita. The studio designs AI harnesses, office and research assistants, custom LLM skills and writing behavior, local and offline models, and generative pipelines for language, image, 3D, and video. Field work includes legal and contract drafting, medical and clinical documentation, and scientific and mathematical research systems.

Question 23

What is an AI harness?

Accepted Answer

An AI harness is the production system around a model or crew of agents: tools, memory, permissions, evaluation, logging, and human checkpoints. It is what turns a demo into work you can run for weeks — in a studio or in an office.

Question 24

Do you take short-term work as well as long-horizon programs?

Accepted Answer

Yes. Okita is built for both. Long-horizon programs need durable state, review, and orchestration. Short-term task management and office installs need tight briefs, fast cycles, and a clean handoff.

Question 25

Who is Alex Okita?

Accepted Answer

Alex Okita is a technical artist and software engineer based in the San Francisco Bay Area. He started as a concept artist at Bungie, worked in film and AAA games, served as a spatial interaction engineer at Occipital, wrote Learning C# Programming with Unity 3D, and now specializes in generative AI systems, office and research assistants, and training LLMs for specific environments.

Question 26

What generative media do you specialize in?

Accepted Answer

Large language models, image generation, 3D mesh generation, and video generation — with technical art and engineering so the output can enter a real production pipeline.

Question 27

Do you work in AR, VR, and XR?

Accepted Answer

Yes. Spatial computing is a core part of the practice, including AR, VR, XR, real-time graphics, and the generative content that feeds those runtimes.

Question 28

Where is Okita based?

Accepted Answer

Los Gatos, California, in the San Francisco Bay Area. Engagements are remote-friendly, with on-site office installs when the work needs to live in the room.

Question 29

How do we start?

Accepted Answer

Send a brief to alex@okita.tech. Name the office or research jobs you want automated, whether the model must write in your voice, whether data can leave the building, and whether you need hardware. We reply with whether we are the right studio and a proposed first slice of work.