DocumentIt reads your internal policies, templates, work logs and meeting notes first, then drafts on that basis. Output is delivered as HWP, Word, PowerPoint or Excel.
Why general-purpose AI is not enough
Whether the evidence sits inside your company is the fork in the road. Harness Document reads your company records first, then drafts on that basis.
Output comes out as HWP (hwpx), Word (docx), PowerPoint (pptx) and Excel (xlsx).
Output formats
Whatever format the recipient asks for, there is nothing to retype.
You can check which meeting or which task a passage came from.
HWP, Word, PowerPoint, Excel. No reformatting afterwards.
Where evidence is thin it leaves a “needs checking” marker and leaves the judgement to you.
Drafting time goes down. The judgement and the review stay with your team.
Lite
Individual · fewer than five employees
per month100,000KRW
VAT included · per seat
Eligibility Fewer than five employees
Talk to usPro
Companies · team-wide
Request a quote
Priced against headcount, support level, term and requirements.
Eligibility No restriction
Request a quote| Included | Lite | Pro |
|---|---|---|
| Document drafting | ||
| Tender collection | ||
| HWP / Word / PPT / Excel output | ||
| Seats | 1 seat | By agreement |
| Dedicated environment and support | ||
| On-premises installation | By agreement |
If you also need the code
When design and development enter the picture, you move up to Builder. Documents and tender collection stay exactly as they are, with Harness Coding added.
See BuilderGo-live (2 weeks)
Dedicated environment, accounts, initial indexing of internal material, training
Month 1
Real work documents. A weekly remote check-in
Month 2
Tuning against your templates and vocabulary
Month 3
Performance review — you decide here whether to continue
What you do
What accumulates
You write a report
What progressed, and how it was decided
You write a plan
The criteria your company actually applies
You upload meeting notes
Who agreed to do what, by when
You add policies and templates
Your company's vocabulary and formats
Finished deliverables
It is unambiguous which one is final
The conversation that finished it
The reasoning behind each decision stays
The material it drew on
The link back to the internal source stays
Documents scattered around a company do not say who made them, why, or which one is final. What passes through here is your company's knowledge, completed by a person in dialogue with AI.
AI drafts every document in your company
Reports, plans, proposals, meeting notes and more — AI drafts them automatically, in line with your internal policies and work records.
Your company's knowledge accumulates
Each draft leaves the finished document, the reasoning behind it and its sources together, so your company's knowledge keeps building.
The neural model connects your knowledge
The neural model finds and connects the relationships between teams, tasks, documents and owners across accumulated conversations, documents, deliverables and system data.
Two to three months after adoption
You implement the gaps with Builder
Once connected, the gaps become visible. The systems your company needs are built by AI as well.
The AI agent carries out the work
What used to require a request, the AI now judges and handles on its own. A true agent that runs your everyday work takes shape.
We do not skip step one. Only once your company's knowledge has built up does the next step speak in your own language.
We do not claim it is airtight
Inference for document generation passes through an external AI. So we will not tell you that technology alone blocks everything. Instead we separate what technology handles from what your own policy has to hold.
What technology handles
What policy must hold
What gets uploaded and what does not is your company's call. Internal rules and training have to sit alongside it for that to hold in practice.
| Pasting into AI directly | Going through Harness | |
|---|---|---|
| What was used | The company cannot know | It stays on record — who used which material, for what |
| Source material | Scattered across personal PCs and personal accounts | Turned into knowledge on the server only |
| Evidence scope | Each person decides on their own | Only within the scope the company sets |
| Outputs | Kept personally, with no permission boundaries | Internal folders with per-department permissions |
| Basis of the result | Plausible sentences (not about your company) | Actual content drawn from internal material |
The problem is not that AI is being used. It is that no record of it stays with the company.
The AI used for document generation runs on an enterprise business account, where inputs and outputs are not used for model training by default.
There is one exception. If a user rates a conversation, that conversation may be used for training. So at go-live we turn that feature off in your organisation settings — so the rule holds without exception.
Usage
Who · when · which task · how many
Output register
The list of documents produced, and where they are
Evidence trail
Which internal material a document was based on
Capacity
Usage against the company total, and the spread across users
Capacity is 10 GB per person, and the company total is headcount × 10 GB. If one person goes slightly over while the company total still holds, we do not block them — work keeps moving while the company stays protected. Concurrent access is capped at two devices per seat (PC and app) to prevent account sharing.
And — administrators cannot look freely either
What an administrator sees is who did what and when; the body of a conversation is blocked by default. Opening it for a legitimate reason such as an audit requires separate authority, and the fact that it was opened is itself recorded.
Control and surveillance are different. We keep what the company needs to know, and protect how a person works.
Keeping every line inside would require your own AI serving with on-premise servers and GPUs, which is a separate investment. You can weigh that when the need arises. Most companies use it in its current shape — Harness inside, inference outside.