SenangAI: What We're Building at Kognosys, and Why

SenangAI is the platform my company, Kognosys, is building. It connects an organisation's own knowledge to AI so people get answers they can check, workflows they can trust, and a record of everything that happened. It runs where the organisation decides, with the AI models the organisation approves.

Full disclosure up front: this is my company, so this is not a neutral review. I'll explain what we're building and why, and keep the claims to what's on kognosys.com.

Why I started Kognosys

I've spent about 25 years in enterprise IT in Malaysia. Server rooms in Penang, data centres for hosting customers, security monitoring for banks, and then nine years co-founding Netwitz, building analytics and monitoring for banks, utilities and government-linked organisations.

One lesson kept repeating. The organisations that most need better tools are the ones that can least afford to lose control of their data. A bank can't paste customer records into a public chatbot. A ministry can't let a model see files the user isn't allowed to open.

Generative AI made that tension sharper. The tools are useful, but most of them assume your data can go wherever the model lives. That's the gap Kognosys was set up to close. If you want the bigger picture, I wrote about it in What is sovereign AI?

What SenangAI does, in plain terms

Think of SenangAI as one controlled layer between your people, your knowledge and your AI models. Kognosys describes it as "one governed intelligence layer". In practice it does four things.

1. Answers you can check

SenangAI answers questions from your approved sources, and every answer comes with the sources it used. If an answer looks wrong, you can open the source and see why.

2. Permissions that travel with the knowledge

This is the part I care about most. In SenangAI, "permissions resolve before retrieval". In other words, the system checks who you are and what you're allowed to see before it looks anything up. If you can't open a file, the AI can't use it to answer you either.

Work happens inside one of three scopes:

• Personal: your own space.
• Teamspace: shared with a team.
• Organisation: company-wide.

Each scope has its own permissions and audit trail. An out-of-scope question is denied, and the denial is recorded.

3. Enterprise Memory that outlives the model

AI models change every few months. Your company's knowledge shouldn't have to be rebuilt each time.

SenangAI keeps an Enterprise Memory: the indexes, embeddings (number-based fingerprints of meaning, explained in my vector databases guide) and links between documents, built with Kognosys's own embedding model and kept inside the boundary you choose. When you switch the AI model, the memory, permissions and audit history stay put.

4. Actions with a human in charge

Answers can turn into workflows, such as drafting, routing or updating something. But any consequential step pauses for "a named human approval before execution". The AI suggests; a person decides.

Your models, your boundary

Two choices stay with the organisation.

Where it runs. On-premise, private cloud, a sovereign region, or as a managed service. Files, connectors and indexes stay where you decide.

Which models it uses. Local models, open-weight models, regional or commercial models, or a mix. SenangAI routes each request using "policy before performance": first the data boundary and the list of approved models, then task fit, reliability, speed and cost.

That second point matters for lock-in. If a better or cheaper model comes along, or a regulation changes, you switch the model, not the whole system. I covered why that flexibility matters in RAG vs AI agents vs agentic RAG.

An audit trail for the hard questions

Sooner or later someone asks: who saw what, and why did the AI say that? SenangAI keeps a record of each request: the person, the scope, the sources, the model, any approvals, the output, and blocked attempts. Risk and compliance teams can reconstruct and export that evidence.

If you're starting your own AI rules, my AI governance starter guide is a good companion to this.

Who it's for

We're building SenangAI for organisations where control isn't optional:

• Government agencies and government-linked companies.
• Regulated enterprises, like banks and insurers.
• Large organisations with complicated permissions.

Inside those organisations, it's aimed at the CIO who wants AI without giving up infrastructure or model choice, the data leader who needs knowledge usable without flattening permissions, and the risk team that wants evidence before an audit, not after an incident.

It connects to the places knowledge already lives: email, Google Drive and Microsoft 365, SharePoint and OneDrive, Slack, files and PDFs, databases, wikis and APIs.

How we start with a new organisation

We don't ask anyone to take this on faith. Kognosys offers a 14-day proof of concept:

1. Day 1: the first cited answer from one approved source.
2. Day 5: a boundary test, where an out-of-scope question is denied.
3. Day 14: one live workflow with human approval and a full evidence trail.

Small, measurable, and on your own data. It's the same pilot-first approach I recommend in my enterprise AI roadmap.

Conclusion

SenangAI is our answer to a simple question: how do you get the benefits of AI without handing over your data, your permissions or your choice of model? Cited answers, permission-aware memory, human approval and a full audit trail, running where you decide. If that's a problem you're facing, I'd like to hear how you're thinking about it.

Quick answers

What is SenangAI?

SenangAI is Kognosys's governed AI platform. It connects an organisation's approved knowledge to AI models, with cited answers, permission checks, human approval for actions and an audit trail.

Is SenangAI tied to one AI model?

No. It can use local, open-weight, regional or commercial models, or a mix, chosen by the organisation's own policy. Enterprise Memory stays the same when the model changes.

Where does SenangAI run?

On-premise, in a private cloud, in a sovereign region, or as a managed service. The organisation decides where its files and indexes live.

How can an organisation try it?

Through a 14-day proof of concept with Kognosys, starting from one approved source. Contact details are on kognosys.com.