Skip to content
Lihok TechnologiesOPC— home

Lihok Technologies OPC

AI can already talk. The next opportunity is AI that can work.

We are a Philippine technology company. We build a governed operating layer for AI — and the products that run on it.

Request a consultationSee what is live today

From intent to verified work

How work moves through LihokHuman authority sits above a dashed authority boundary. Intent enters from the left into EDITH, the coordinating core. EDITH delegates down and to the right into a bounded delegation region containing three specialist agents, each connected to a capability. That region feeds a wide execution band holding business systems and the work performed against them. Two channels run vertically on the right: policy and approval travel downward from authority, and evidence travels back upward from the execution band to human authority. An approval gate sits on the downward channel where it crosses the authority boundary.Human authoritypolicy · approvalIntentEDITHapproval when requiredAgentsCapabilitiesSystems & workevidence
  • Human authority
  • EDITH
  • Specialist agent
  • Capability
  • Approval gate
  • Evidence
Authority is a line, not a step: it sits above the whole system, and consequential actions cross it only with approval. Evidence travels back up the same boundary, which is what makes the delegation below it safe to grant.

The problem

Capability is abundant. Coordination and accountability are not.

Frontier models keep getting better and cheaper. Intelligence is no longer the scarce input — it is becoming a commodity one.

What is still scarce is what happens around that intelligence: deciding what should happen, keeping it inside the limits the business actually set, stopping when a decision is consequential, and being able to show afterwards what was done and why.

Most organisations can now make a model call a tool. Very few can make that call accountable.

Same parts, two arrangements

Without a layer, compared with LihokThe same comparison stacked vertically. The upper panel scatters seven labelled elements joined by crossing dashed lines. The lower panel arranges the same seven around one central hub labelled Lihok, each joined by a single short spoke.Without a layerPeopleAI toolsDocumentsSaaSDatabasesWorkflowsOperationsWith LihokLihokPeopleAI toolsDocumentsSaaSDatabasesWorkflowsOperations
  • Lihok
  • A thing that must be coordinated
Nothing in the right-hand panel is a new tool. It is the same seven things, arranged so that coordination happens once in one place instead of continuously between every pair. That is what an operating layer is: not more capability, but a different topology.

Portfolio

One foundation. Several products. Not several companies.

Each of our products is a different commercial answer to a different problem. What they share is the expensive part: governed intelligence, the authority model, and the evidence machinery. Those are built once and reused.

Lihok Frontline

Available now

An intelligent front office for your business — not another passive website.

A conventional website answers nothing and qualifies no one. It cannot tell a serious buyer from a casual visitor, it cannot answer the questions people actually ask, and it loses the intent behind every visit it doesn't convert.

For: Businesses that generate demand through their own website and want it to work — and to stay answerable for what it says.

Learn more

Atiman

In development

An Asset Operations and Intelligence Platform for physical asset maintenance.

Maintenance knowledge usually lives in the heads of senior engineers, in scattered spreadsheets, or inside legacy systems that record transactions but never capture understanding. Field observations are collected and then lost.

For: Asset-intensive operators — frontline operators, supervisors and reliability engineers.

Ayoha

Concept

A concept for an AI-native everyday-services platform — making everyday services easier to discover, coordinate and complete.

Discovering, coordinating and completing everyday services is still unnecessarily fragmented. We think it is a real problem; we have not designed the answer yet.

For: Consumers and local service providers.

How the portfolio fits together

In production

One of these platforms is running in production today.

Lihok Frontline is live. Its public site engine serves a real company's website, its inquiry service is operational, and its tenant administration console is live.

We are careful about the difference between a platform that is deployed and a capability that is finished. Where something is still in development, we say so on the page — including on this one.

Work you can check

Work that ends in a recordAlong the top, a request leads to a decision and then to an action. From the action a path drops into a large evidence record drawn below: a document with a header and four entries, headed Evidence record, carrying a verification mark. The arrangement shows that the action produces an inspectable artefact rather than only a result.RequestDecisionActionEvidence record
  • Inspectable record
  • Verification
An opaque agent asks you to trust its answer. This keeps the decision, the reasoning and the action in a record you can inspect, so the question becomes what was authorised rather than whether to believe the result.

Our approach

Work that ends in evidence, not in a claim.

Our platforms take an objective, work out what has to happen, delegate to the right models, agents and tools, run the work, pause for a human wherever the action is consequential, and keep a durable record of everything that was done.

Underlying models and providers are treated as replaceable by design. The operating layer — planning, authority, execution and evidence — is the part we own and the part that compounds.

Autonomy with one interruptible arc

The governed autonomy loopSix stages arranged clockwise around a circle: Observe, Understand, Plan, Execute, Verify, Evidence. Between Plan and Execute, sitting on the ring itself, is an approval gate, labelled alongside. EDITH sits at the centre. The arc from Evidence back to Observe is drawn in the accent colour, closing the cycle.approvalObserveUnderstandPlanExecuteVerifyEvidenceEDITH
  • Approval gate
  • Evidence
  • EDITH
Approval is a gate on one arc, not a step in the queue: it interrupts the transition from planning into execution and nothing else, which is why the rest of the loop can run unattended. The accent arc is evidence flowing back into the next observation.

Evidence, not assertions

Work should end in a record that can be inspected, not in a statement that it went well.

Authority is explicit

Consequential actions stop for human approval. An executor cannot perform what it was not granted, and a conversational agent may assert an outcome only after an authorised action actually succeeded.

We label what is finished

Every product on this site carries its real stage. A roadmap item is never presented as something you can use today.

How we build

Tell us what you are trying to solve.

Our enquiry form cannot receive messages yet. We would rather say so than invite one we cannot answer.

Request a consultation