Internal helpdesk & HR
Policy, leave, claims and IT questions that currently interrupt three people a day.
Applied AI
An assistant that only knows what you gave it
HeyKiko is a retrieval-augmented AI assistant connected to your SOPs, policies, contracts and databases. Every answer is cited to its source, access follows your existing roles, and deployment can be entirely inside your own perimeter.
HeyKiko — Private RAG Assistant
On-prem
Which SOP covers cold-chain deviation?
SOP-114 §3.2 — deviations beyond ±2°C for more than 20 minutes
require a QA hold and an incident record within 4 hours.
SOP-114_ColdChain.pdf · §3.2
Cited answersRole-based accessZero data leakage
Illustrative interface
0Documents per tenant
0Answer with citation
0Deployment options
0Weeks to pilot
The problem
The objection to AI at work is rarely accuracy
It is that nobody can tell where the answer came from, and nobody is confident the question did not leave the building.
Staff already paste internal documents into public chat tools, because the alternative is reading a two-hundred page manual. That is the real risk, and blocking the tools does not remove it — it just moves it to a personal device.
HeyKiko gives people the same convenience against your own content, with two properties that general assistants cannot offer: every answer carries a citation you can open and check, and the whole system can run where your data already lives. When it does not know, it says so instead of inventing something plausible.
On hallucination: retrieval-augmented generation reduces fabrication because answers are grounded in retrieved passages, and enforced citation makes any remaining error checkable in one click. It is a meaningful improvement, not a guarantee — which is why we design the workflow so a human stays accountable for consequential decisions.
Ingests PDF, Word, Excel, PowerPoint, HTML and SQL sources
Every answer cited to document, page or clause
Role-based access mirroring your existing permissions
On-premise, private cloud or managed hosting
Answer quality review queue with feedback loop
Full query log for audit and improvement
Capabilities
What HeyKiko does
A working knowledge assistant rather than a demo — which mostly means the unglamorous parts are handled.
Ingestion & indexing
Bulk and scheduled ingestion from file shares, SharePoint, wikis, ticketing systems and databases, with change detection so updated documents supersede old ones.
Retrieval
Hybrid semantic and keyword retrieval with re-ranking, tuned per corpus, plus metadata filters so a query can be scoped to a site, product or policy year.
Access control
Document-level permissions inherited from your source systems — a user cannot retrieve from a document they could not open directly.
Citation enforcement
Answers must be grounded in retrieved passages; unsupported responses are suppressed rather than returned, and every claim links to its source.
Deployment
Fully on-premise, in your own cloud tenancy, or managed by us in Singapore. Model choice is configurable, including self-hosted open-weight models for air-gapped sites.
Observability
Query logs, unanswered-question reports, and a review queue so subject experts can correct the corpus where the gap actually is.
Who it's for
Where it earns its keep
Compliance & SOP-heavy operations
Regulated environments where the right answer is in a controlled document and citing it matters.
Customer support
Front-line agents answering from approved content, with the citation available if a customer challenges the answer.
HeyKiko FAQ
Before you get in touch
No. Your documents are indexed for retrieval, not used to fine-tune a shared model. In an on-premise deployment the content never leaves your infrastructure at all.
Every answer carries a citation, so an error is visible and checkable rather than silent. Unanswered and low-confidence queries land in a review queue, which is usually how you discover that the underlying document was ambiguous.
Yes. Self-hosted open-weight models are supported for air-gapped or highly sensitive deployments. There is a capability trade-off and we will be straight with you about it.
About three weeks: one to ingest and tune a defined corpus, one for a small user group to use it in anger, one to act on what that reveals.
Yes — the assistant in the corner of this page runs on HeyKiko, answering from our own content. It is the least abstract demo we can offer.
Related
You might also need
Data services
Getting your documents into a state worth indexing.
See servicesSecurus
Access control and audit for physical and system security.
See SecurusAI consulting
Working out where AI actually pays for itself in your operation.
See services
Next step
Bring us the messy problem. We like those.
A 30-minute call, no deck, no obligation. Tell us what is breaking and we'll say honestly whether we're the right team to fix it.