Spead AI

Context

Generic language models are fluent, but they do not know how a specific organisation actually works: its templates, its approval chains, its own past documents, the regulation it answers to. In large companies, state-owned enterprises and government agencies, that gap is where most of the real work sits, and it is also where the data is too sensitive to send to a public model.

I started Spead AI in November 2025 to close that gap, incubated within Caliana Indonesia, where I had spent the previous months as a strategic initiative analyst working directly under the President Director. The thesis is simple: an organisation's internal expertise is an asset that can be indexed, retrieved and reused instead of re-typed, and the system that does it has to live inside the organisation's own walls.

I founded the company and lead it end to end, from market validation and deep-tech architecture to the core AI engineering, taking the product from zero to a production MVP. Spead AI is also the capstone case for my Master of Management in New Ventures Innovation at Universitas Prasetiya Mulya, so business-model decisions are tested against a live product rather than a slide deck.

What was built

• Multi-agent and data sovereignty. Sovereign multi-agent pipelines that connect foundation LLMs to isolated on-premise or private-cloud infrastructure, with self-correction loops, so nothing leaks outside the organisation.
• Model Context Protocol. Custom MCP servers that link sovereign databases and internal knowledge graphs directly to agentic workflows.
• Multi-channel gateway. An AI gateway that orchestrates autonomous workflows across 50+ enterprise messaging protocols.
• Document intelligence. Automated pipelines for deep parsing of unstructured documents and regulatory compliance mapping, the work that used to take a team days of reading.
• Shipped as products. The document engine became Doc Builder and Doc Assistant on spead.ai, and the operations planner became the Daily Productivity Planner. The product section below lists all five modules.

The toolset is the one I use across my AI work: Python with LangGraph and LangChain for orchestration, MCP for tool and data access, vector search with Pinecone and FAISS, and Google Cloud or on-premise servers for infrastructure, depending on where the client's data is allowed to live.

The product

What a visitor finds on spead.ai today is the packaged version of that engine: an "AI administrative employee" for Indonesian organisations, sold in tiers, with the enterprise deployment described above sitting at the top of the ladder.

Backed by Caliana

Spead AI is not a solo experiment. It is incubated within Caliana Indonesia, the technology brand of PT Data Nusantara Adhikarya, and Caliana lists it in its main navigation as its business administrative AI line. That gives a young AI product what it usually lacks: a certified company behind the contract, a sales channel that already reaches procurement committees, and clients who have worked with the parent before.

• Caliana has built integrated systems since 2019: visitor management, intelligent surveillance, big data, integrated command centers, access control, face recognition, learning management and enterprise software development.
• ISO 9001 certified, registered with the Directorate General of Intellectual Property, and a registered reseller on LKPP's e-catalogue, the procurement platform of the Indonesian government.
• Its client list spans PLN, Pertamina, Daikin, Traveloka, GMF AeroAsia, ISS, Bakrie Tower, Prodia and public-sector bodies such as Diskominfo Banggai and the Brimob training unit at Cikeas.

Validation and market fit

• Selected for the Google for Startups AI Solution Lab Program 2025, with compute access and technical mentorship, and graduated in December 2025.
• Enterprise market research and feasibility studies mapped the highest-impact automation across the public sector and regulated enterprises before a line of product code was written.
• Product-market-fit frameworks integrate Spead AI into Caliana's ISO-certified distribution channels and strategic proposals, which is how a young AI product reaches a procurement committee at all.

Results

• A production-ready enterprise MVP delivered from scratch and deployed for strategic pilot testing with a state-owned enterprise.
• 100% of the core intellectual property and codebase retained, with technical governance in place for a spin-off.

Where Spead AI stands

Spead AI is live at three enterprises, with two government deployments in progress. The five modules on spead.ai (Doc Assistant, Doc Builder, Daily Productivity Planner, AI Partner and Enterprise Admin Panel) are the packaged version of the same engine, distributed through Caliana's ISO-certified channels, and 100% of the core intellectual property (HAKI) and the codebase stays with the founder, ready for a spin-off.

Spead AI is also my New Venture Project, the final project of the MM New Ventures Innovation programme at Prasetiya Mulya University.

Lessons

Building for organisations that run procurement committees also forces discipline early: audit trails, data boundaries and a clear answer to "where does our data go" have to exist before the first pilot, not after.