On 25 August 2026 our co-founder Feipeng Liu spoke at Can AI Make HR Easier?, a marketplace edition of Lorong AI's AI ToolsDays at Lorong AI @ One-North, Level 3 Vidacity. Six teams each got a short spotlight slot across the HR journey, covering interviewing, candidate fit, onboarding, workforce operations and AI assurance. Our slot ran from 3.50pm to 4.00pm and was called AI Visibility for HR & Workforce Insights. Ten minutes, two slides, and the rest live in the product.
Two questions to start
Feipeng opened by asking for hands. Who used ChatGPT or Gemini this week to find a vendor, a firm or a service? Most of the room went up. Then the follow-up, with the hands still in the air: keep them up if you know what those assistants say about your own company. Almost every hand came down.
Everyone uses the assistants. Almost nobody knows what the assistants say about them. That gap is the whole reason we built the company.
The cost of omission
An AI answer is not a page of ten blue links you can scroll past. It is a short list of names, each with a reason attached. Someone asks "which executive search firms in Singapore are best for hiring a fintech CFO?" or "which HR platforms should a 200-person company in Singapore look at?", and back comes a handful of firms.
If you are not one of them, nothing happens to you. No impression, no lower ranking to climb back from, no click you narrowly missed. You are simply absent from that conversation. And because the assistant already did the shortlisting, the buyer only visits the sites that made the list. Missing from an AI answer when someone is actively looking for what you sell costs real revenue, because you are out of the deal before anyone has heard of you.
For an HR audience that cuts two ways. There is the vendor selection side, which is how HR software, recruiters and search firms get onto a buyer's list in the first place. There is also the employer brand side. Candidates now ask assistants what a company is like to work for, and the answer gets assembled from sources most employers have never looked at.
Two customers, anonymised
The two slides came from live customer accounts, shown without naming the firms. Both are smaller Singapore firms in the HR and talent industry, the same industry as the room, and both started working with us a few months ago in verticals dominated by names everyone there would recognise. Two charts did the work:
- Mentions evolution. How often each brand gets named at all across the assistants, week over week.
- Brand share evolution. The share of the answer space they hold against their real competitor set, weighted by position rather than counted raw.
Both curves climb, and they climb against far larger incumbents. What matters there isn't the absolute numbers, it's the shape. You can influence how your brand shows up in AI answers. Every inflection on those charts lines up with something we did, mostly targeted content written for the specific conversations those firms wanted to win, published and then picked up by the engines.
Feipeng also flagged the caveat on the same slide: expect a lag. Content does not move an answer the week it goes live. The engines have to crawl it, cite it and start leaning on it first. Our curves show that delay plainly, and any vendor whose charts don't show it is telling you a story rather than showing you data.
The demo
The rest of the slot was the actual product, not screenshots. The walkthrough used a live vertical with a well-known professional services firm as the hero brand. It's a brand we track rather than a customer, picked because everyone in the room would know the name. The route:
- Competitors. The real competitor set for that vertical, and where the hero brand sits against it.
- Brand and mentions across the assistants. The same brand seen separately by ChatGPT, Claude, Gemini, Perplexity, Google AI Overview and Google AI Mode. They disagree with each other more than people expect, which is why a single blended "AI score" isn't much use.
- User journey. Where the brand appears across the Discovery, Research and Decision stages. Strong at the top and absent at the bottom is a very different problem from the reverse.
- Prompts. Opening one prompt and its captured answer. Worth being precise here: you are looking at a rendered answer from a real assistant session, geo-targeted and captured on a schedule. Not an API response, not a proxy, not a reconstruction.
- Citations. Which sources the assistants actually leaned on to build that answer. This is where most of the practical work lives.
- Website Audit. What on your own site helps or blocks the engines that read it.
- Content Hub. Example articles written against specific citation gaps, and how those pieces performed after publishing.
How the system works
Feipeng closed on the mechanics, because in a room full of tool evaluations the method is what separates one dashboard from another.
- We onboard your vertical, not the whole internet. Your brand, your real competitor set, your market.
- The platform writes the prompts for that vertical, with a human check, then holds them steady so week-on-week stays comparable. A keyword like "executive search Singapore fintech" becomes a prompt like "Which executive search firms in Singapore are best for hiring a fintech CFO?"
- Those prompts run through real browser sessions on geo-targeted IPs, so we capture what someone in your market actually sees, in the local language where that matters.
- Six engines: ChatGPT, Claude, Gemini, Perplexity, Google AI Overview and Google AI Mode, plus others where a market needs them, such as Naver in Korea.
- We read every answer for who got named, in what order and in what light, then roll that up into one brand share number you can track. Refreshed weekly.
- Same prompts, same schedule, every week. That is what lets you tell whether anything you changed actually moved.
The rest of the lineup
The marketplace format put us alongside five other teams working on different parts of the HR stack: Hirona on AI interviewing, Peoplemath on hiring for organisational fit, Omni HR on automating the first 180 days, Glints on AI agents across HR workflows, and the Global AI Assurance Sandbox on testing hiring AI for bias and risk. Demos ran to 4.15pm, then everyone moved to the booths for the last 45 minutes, which is where most of the real conversations happened.
Thank you
Thanks to Lorong AI for the format and for having us back at One-North, and to everyone who came by the booth afterwards with their own version of the second question: so what does it say about us? If you left with that question too, it's a quick one to answer. Get in touch and we'll run your brand through the engines.