What am I buying?
One system: the Operational Intelligence Platform.
Not a bundle of tools. One platform that turns your company's own knowledge into evidence-backed answers — and keeps a person in every decision.
Your knowledge becomes operational intelligence — backed by evidence, reasoned into recommendations, and never acted on without you.
Every capability below links to its engineering evidence and its current implementation status.
One platform. Five core capabilities.
Everything works together
Five things the platform does with your knowledge — expressions of one system, not five separate products.
Preserve knowledge
Operational Memory
What it is
Your documents, procedures and records become one living, structured memory.
Why it matters
Knowledge stops walking out the door when experienced people leave.
What proves it
Your raw sources stay the source of truth; nothing is invented.
Reality today
In progress — ready for scoped pilots.
Find knowledge
Knowledge Intelligence
What it is
Ask in plain language; get an answer built only from your own sources, with citations.
Why it matters
Answers in seconds instead of hours of searching and phone calls.
What proves it
Document answers show their sources — and it says so when the evidence isn't there.
Reality today
Available today.
Explain knowledge
AI Transparency
What it is
See what kind of evidence backs an answer, and the steps that produced it.
Why it matters
Trust an answer you can inspect — not one you take on faith.
What proves it
Evidence classification on every answer; the pipeline's steps, not the model's reasoning.
Reality today
Available today; the full step timeline is pilot-ready.
Measure knowledge
Knowledge Health & Evaluation
What it is
Score coverage and gaps, and measure answer quality on your own corpus.
Why it matters
Fix what's thin before it costs you — and turn quality into a number.
What proves it
Every new model version must pass a release gate — safety first, then no regression.
Reality today
Available today.
Govern knowledge
Human approval & governance
What it is
Recommendations arrive as drafts; a person approves before anything leaves the platform.
Why it matters
AI assists; people decide — and every step is auditable.
What proves it
Nothing auto-executes; export is refused without a recorded human approval.
Reality today
Available today.
Where Haryts fits
Between your systems and your people
Haryts reads the operational knowledge you already have and serves the people who need it — without replacing your systems of record.
Your knowledge & systems
Haryts
Operational Intelligence
Your people
Deploy the AI that fits
Your knowledge is never hostage to one AI vendor
Haryts runs its own AI model — your data never goes to an external AI provider. Choose the deployment that matches your requirements.
Haryts-managed
Haryts's own AI model, run on hardware Haryts operates — no external AI provider receives your documents or questions.
Local AI mode
For strict privacy or residency, generation never leaves your walls. Deployed inside a scoped pilot.
Evidence-only mode
When policy doesn't permit generative AI for a data class, Haryts still returns cited sources.
Measured, not claimed
How we test answer quality
We test answers against reference questions and their supporting documents — including cases where the evidence is incomplete or conflicting, and the honest answer is to refuse rather than guess.
We're still validating our evaluation methods, so we're not publishing a headline quality score yet — we'd rather show you how we test than put up a number we can't fully stand behind. The standard we hold our answers to, we hold our own evidence to.
Today vs tomorrow
A clear line between what ships and what's coming
Everything above the line works on your knowledge today. Everything below is direction — never presented as available now.
Haryts Connect
Your existing documents become answerable knowledge without migration — raw files remain the source of truth.
- Live sync for SharePoint / Google Drive is roadmap, not available
File upload (PDF, Word, Excel, CSV)
Start from the documents you already have — messy exports included.
Website connector
Knowledge published on your website becomes cited, answerable knowledge.
- Crawls approved sources on demand; scheduled automatic re-crawl is roadmap
- The hosted service reaches public web pages only; a private intranet needs a deployment that can reach it
Haryts Retrieval Intelligence (HRI)
Answers are grounded in retrieved evidence from your own documents — deterministic retrieval runs before generation, always.
Evidence classification
You can see at a glance what kind of evidence backs an answer, and whether it is the governing version.
Citations to source documents
Your team can open the original document and verify an answer in seconds — no answer has to be taken on faith.
- Inline [n] markers do not yet appear on every document answer; improving citation coverage is a target of the next model
No-LLM Evidence Mode
Still get cited sources when AI generation is unavailable — or when policy doesn't permit generative AI for a data class.
Workspace Intelligence
Your team never gets confident answers from a broken or half-synced knowledge base.
Trust Score
Users see how well-supported an answer is instead of guessing.
Encrypted credentials
Connector credentials and the names of people in imported records are encrypted in the database; secrets are never returned in answers or API responses.
- Stored documents are not encrypted at rest yet
Privacy Levels (L1/L2/L3)
Work-order and maintenance history becomes answerable without exposing who did what, unless you decide otherwise.
Tenant isolation
Each customer workspace is isolated; knowledge is never shared across tenants.
Audit logs
Security-relevant actions are recorded in an audit trail.
Haryts's own AI model
Your knowledge never passes through a third-party AI vendor, and the platform is not hostage to one.
HarytsBench (internal benchmark)
A new model version reaches you only after it passes measured tests — safety first, then no regression.
- Internal benchmark; the public benchmark is roadmap
Haryts Eval
Pilots end with measured answer and retrieval quality on your own documents.
Knowledge Gaps
The first honest map of where your documentation is missing — often the most valuable pilot output.
Draft work artifacts (checklists, handovers, reports)
Drafting is accelerated; your people keep the decision.
Human approval gate
Automation you can defend in an audit — a person owns every outbound artifact.
Pilot Reports
Every pilot produces a written report: validation results, eval baseline, gaps, and an honest recommendation.
Operational Records
Equipment history and past work become queryable knowledge instead of dead exports.
- Record normalization is validated with the customer during a pilot; field mappings vary by CMMS export format
Haryts Knowledge Objects (HKO)
Procedures, assets, and work orders become navigable, linked knowledge — without touching the original files.
- Normalization taxonomy is tuned per customer during pilots
Explainability Timeline
Answers are explainable after the fact, not just plausible in the moment.
- Timeline depth varies by pipeline path; reviewed with customers during pilots
AI Transparency Center
No mystery about which AI touched your data, ever.
- Surface is evolving; reviewed with pilot customers
Local AI mode
For strict privacy or residency constraints, AI generation never leaves your walls.
- Deployed under supervision inside a scoped pilot
- Answer quality depends on the chosen local model; benchmarked per deployment
Knowledge Health
Leaders see the state of organizational knowledge the way they see safety or uptime.
- Scoring model is early; not yet calibrated against customer data
Knowledge Confidence
Shows which knowledge is well-supported by multiple current sources and which rests on a single aging document.
- Confidence model evolving; reviewed during pilots
Knowledge ROI
Estimates time saved and value delivered by answered questions and reused knowledge.
- Estimates use customer-agreed assumptions; not audited financials
- No inflated ROI claims: figures are labelled as estimates
Knowledge Evolution
Tracks how workspace knowledge changes over time — additions, revisions, and shifts.
- Early surface; reviewed with pilot customers
Decision Log
Records operational decisions with their supporting evidence for later review.
- Early surface; workflow integration validated in pilots
Operational Memory
Retains operational context across sessions so recurring situations benefit from past knowledge.
- Scope of retained memory agreed with the customer; early surface
Knowledge Loss Risk
Flags knowledge concentrated in few sources or people — where departure or loss would hurt.
- Risk model is early; interpreted with the customer during pilots
Haryts Intelligence Score
A composite score of workspace intelligence: knowledge quality, retrieval quality, and trust signals.
- Early scoring model; expect calibration changes between releases
Actions / Workflows
Connects approved artifacts and steps into repeatable workflows.
- Workflow integration is validated with the customer inside a scoped pilot before running routinely
SharePoint live sync
Live synchronization with SharePoint document libraries.
- Not available — scaffold only; do not market as live
Google Drive live sync
Live synchronization with Google Drive folders.
- Not available — scaffold only; do not market as live
HarytsBench public benchmark
A published benchmark for operational knowledge intelligence.
- Not published externally yet
Full offline deployment package
A complete offline deployment package for air-gapped environments.
- Not verified end-to-end; local AI mode (pilot-ready) is the current private option
Haryts Foundation Model preparation
Long-term preparation for purpose-built operational models.
- Direction only — no live product exists