The useful test of an AI competition is whether it can repeatedly turn broad interest into specific, inspectable builder output.

That is the most important signal from the AI Workflow Competition at Echelon Singapore 2026. Nine other entries reviewed by e27 showed builders working through the harder middle ground of AI adoption: messy inputs, scattered knowledge, human approvals, cost constraints, data gaps, and workflows that must fit existing operations.

Also Read Inside the AI Workflow Competition at Echelon Singapore 2026

For sponsors, government partners, and future programme backers, that matters. The competition created a controlled channel where problem statements, sponsor resources, builder judgement, and submission criteria could be tested. Not every prototype was production-ready. The point is that the format generated multiple credible outputs that could be examined, improved, and rerun.

A testbed, not a showcase

The competition asked builders to work from operational challenges, including revenue growth and efficiency tracks, while showing business impact, cost thinking, safeguards, and proof of execution. Builders also had access to workshops, community support, and sponsor-backed resources from FPT AI Factory, Alibaba Qwen, Bitdeer AI, PixVerse, Notion, and AMD-backed cloud support.

The evidence was not uniformity. It was range. Customer support appeared often, but the better entries treated it as more than faster replies. They connected inboxes to knowledge bases, marketing signals, dashboards, reporting systems, escalation rules, and human review. Others moved into spreadsheet reconciliation, reseller reporting, and workflow education.

1. Morning Wu of AfterWork Startup. Managed to build 1 workflow for each challenge statement. One workflow used AI to answer tickets, tag sentiment, and push weekly insight briefs to email, Slack, or Telegram. Another tackled reseller reporting for The Social Space by pulling fragmented data into reports. The claimed reduction, from 1.5 weeks to three minutes, still needs validation, but it identified a bottleneck.
2. Alpa Parmar of Bots and Brand works and Hari Prasad of Boolean Beyond. Adoption as a comprehension problem. Their six-node workflow classified tickets, searched a knowledge base, routed issues, drafted replies, flagged gaps, and generated knowledge-base entries. The submission’s key point was that AI workflows tested on sample data still need to connect with the systems where an organisation’s real work happens.
3. Patrick Tan of Art Infinity Asia and Abel Choy of Atlantic Media reframed the inbox as a routing layer. It extracted fields from customer messages, searched company documents, interpreted intent through an AI model, and routed each item to a reply draft, Slack alert, CRM update, or knowledge-gap log. Their description of the inbox as “a goldmine of information” captured why these competitions can produce market intelligence: builders reveal where operational data is trapped.

Credible outputs under constraint

1. Team Alpha Beta, led by Ayush K Pacheriawala and Tejas Chavan Maintainability at the centre. Its customer-support triage system separated high-confidence repetitive queries from uncertain issues requiring human judgement. The team used n8n, Google Sheets, FPT AI Factory access, and Alibaba Qwen or other LLM access. Their warning was direct: “The biggest barrier is not cost or technology — it is the gap between what AI can do and what an SME’s internal team knows how to build and maintain.”
2. Morpheus Labs Fuseful team of Dorel D. Burcea, Thang Nguyen, and Lyn Ngan took an adoption-first stance. Its workflow lets staff keep using email and Google Drive while an AI layer handles triage, draft replies, knowledge-based updates, sentiment analysis, and insight generation. The submission avoided promising a new operating model.
3. Wang Heng Xin Melson of Corezz Technology exposed another limitation: many companies already have basic bots, but those bots are not linked to useful shared knowledge. Using Alibaba Qwen partly because of cost and access considerations, the entry pointed towards database-connected, cross-team workflows rather than shallow customer-service automation.

Also Read From support inbox to signal feed: Inside the AI workflow that won at Echelon Singapore 2026

1. Cayden Chai This submission was among the clearest examples of visible output density. Running on 70 customer tickets, its seven-step pipeline produced 35 drafted replies, 35 flagged gaps, 37 marketing signals, six theme clusters, six knowledge-base entries, and a monthly marketing intelligence brief. His framing was concise: “Most SME AI tools answer questions and stop — ours turns support volume into a continuous feedback loop for the business.”
2. Connor Clark Lindh Targeted spreadsheet reconciliation, anomaly detection, and report generation. His submission referenced Alibaba Qwen, FPT AI Factory, Gemini, Google Apps Script, custom APIs, and four prototype automations. The next step he identified was time with end-users to shadow workflows and test solution flows. That is where repeatable adoption becomes real: where data is cleaned, reformatted, checked, and reported.
3. Steve Ng of Digital Futures Consultancy Pushed furthest towards reusable implementation infrastructure. It treated a customer inbox as a self-improving customer-intelligence engine, supported by LLAMA, self-hosted n8n, ChromaDB, FastAPI, Streamlit, Docker Compose, and Swagger UI. The submission claimed 13 out of 13 end-to-end test results and 31 API endpoints. Its sharpest line made the category clear: “The inbox isn’t just people asking for help; it’s people telling you exactly what matters to them.”

These submissions show that not every workflow is ready to be dropped into a company tomorrow.

The AI Workflow Competition inside Echelon 2026 surfaced where AI adoption actually gets stuck: incomplete knowledge bases, disconnected inboxes, fragile reporting processes, uncertain handoffs, and teams that need systems they can maintain after the demo ends.

For sponsors and ecosystem backers, the signal is clear: when builders are given concrete problems, usable tools, and an avenue to show working outputs, an AI competition can become a repeatable mechanism for finding practical adoption pathways across Southeast Asia’s operating businesses.

=== 

Want updates like this delivered directly? Join our WhatsApp channel and stay in the loop.

The e27 team produced this article

We can share your story at e27 too! Engage the Southeast Asian tech ecosystem by bringing your story to the world.You can reach out to us here to get started.