← Back to Blog
繁體中文 English 简体中文
HKSoka · 2026-08-19

AI Companion Hong Kong: Three Conversation Modes That Shape How AI Responds to You

AI Companion Emotional Support Hong Kong Conversation Modes AI Memory

Many people who try using AI to process emotions or stress for the first time come away with a feeling of: "It said a lot, but it didn't quite say what I needed."

You might have just wanted someone to listen, but AI immediately handed you five solutions.

Or you wanted AI to cut straight to the problem, and instead it kept saying: "Your feelings are completely valid."

There's also the reverse: you just wanted to understand your options, and AI started making decisions for you.

The issue is usually that the conversation mode was never set.

AI can provide information, help with analysis, and serve as an immediate space for dialogue. But these three uses require completely different types of responses.

A simple fix: before starting, tell AI what you need it to do right now.

"What do I need from you at this moment?"

The Timing Problem with Traditional Counselling

Traditional counselling has a structural limitation: you might have to wait until next Thursday at 3pm to see your counsellor, but the moment you most need to sort through your emotions might be tonight at eleven, before tomorrow morning's meeting, or on an afternoon when a sudden decision leaves you not knowing what to do.

Emotions and stress have their own timing.

By the time you finally sit across from your counsellor, the moment that most needed to be addressed may have already passed, and what remains is an after-the-fact account.

AI offers a different time scale: it can respond the moment you need to talk.

But between "AI is available" and "AI responds to you effectively," there is still an important step:

You need to let AI know what you need in this specific moment.

The same sentence, posed in information mode, decision mode, or companionship mode, can produce completely different responses.

Three Modes, Three Different Conversations

Mode One: Information Mode

What you need: facts, context, options.

You want to make the judgement yourself — AI just needs to organise the material for you.

For example, if you're considering a job change, you might only want to know about salary ranges, work patterns, risks, and career directions in a particular industry.

In this case, the most effective prompt is usually direct:

"[Briefly describe your situation]. Please give me relevant information, context and main options. I will make the decision myself. Please reply in English."

Traditional Chinese version:

「[簡述你的情況]。請提供相關資訊、背景和主要選項，不需要給我建議，讓我自己判斷。請以繁體中文回應。」

When to use it

Core principle: AI provides the material; you keep the judgement.

Mode Two: Decision Mode

What you need: challenge, blind spots, risks.

You already have an idea — you just aren't sure whether you've missed any important factors.

In this case, simply asking "what do you think I should do?" tends to turn the problem into generic advice.

A more effective approach is to assign AI a specific role:

"I'm considering [your decision or plan]. Please honestly identify my blind spots, counterarguments and risks I may have missed. Prioritize challenging my assumptions rather than reassuring me. Please reply in English."

Traditional Chinese version:

「我正考慮 [你的決定或想法]。請誠實指出我可能的盲點、反方論點和我沒有考慮到的風險。請優先指出問題，不需要安慰我或確認我的判斷。請以繁體中文回應。」

One more useful requirement

For genuinely important decisions, ask AI to separate:

"Known facts"

"My assumptions"

"AI's inferences"

For example:

"Please break down your analysis into: known facts, my assumptions, possible blind spots, risks, and information that still needs to be confirmed."

This reduces the chance of treating AI's inferences as established facts.

Core principle: AI finds your blind spots; you keep the final decision.

Mode Three: Companionship Mode

What you need: a space to say things out loud.

This is probably the scenario AI gets wrong most often.

Someone says:

"Work was really exhausting today."

AI is very likely to immediately respond with:

  1. Deep breathing
  2. Exercise
  3. Sleep earlier
  4. Talk to friends
  5. Build a healthy routine

Each item might be reasonable on its own, but taken together, they may not be what you needed in that moment.

Sometimes you just want to finish saying what happened today.

In that case, say it directly:

"I'm feeling [your state]. I don't need advice or analysis right now. I just want to talk through what happened. Please listen and respond to what I say without immediately giving me solutions. Please reply in English."

Traditional Chinese version:

「我現在感覺 [你的狀態]。我不需要建議或分析，只想把事情說出來。請先聆聽和回應我說的內容，不要立即提供解決方案。請以繁體中文回應。」

If you want to switch modes later

You can say directly:

"Now switch from companionship mode to decision mode. Help me analyse what I just described and identify my blind spots."

Or:

"Don't analyse yet — I just want to keep talking."

AI's role can shift alongside your state.

Core principle: AI helps you organise your current thoughts, rather than rushing to solve your problems.

There's Actually a Fourth Question: Which Mode Do You Need Right Now?

In practice, people don't always know what they need at the start.

You might just feel:

"Everything feels confused today."

In that case, ask AI first:

"I'm feeling a bit confused right now. Don't give me advice yet. Ask me a few simple questions to help figure out whether I need information, decision analysis, or just to talk things through."

The value of this approach:

Establish the conversation mode first, then start working on the problem.

This is more precise than immediately asking AI to "help me fix it."

Modes Can Switch Within the Same Conversation

One of the genuinely useful things about AI companionship is that it can shift as your state shifts.

For example:

Phase one: Companionship

"Something happened today. I want to finish telling you about it first."

↓

Phase two: Information

"Now tell me — what are the usual possibilities in a situation like this?"

↓

Phase three: Decision

"I have two options. Help me identify the biggest risk with each one."

These three phases are actually three completely different kinds of work.

A good AI conversation doesn't have to be "ask once, get one perfect answer."

It's more like a tool you can switch modes on at any point.

How Memory Changes the Experience

Using AI once and using AI continuously for six months are two different experiences.

One key difference is memory and continuity.

When AI consistently holds your background, work context, long-term goals, and important threads from past conversations, you can start from where you are right now — without re-explaining everything from scratch each time.

For example:

"This time I want to try a different approach."

If AI holds the context from previous conversations, it knows what "different" refers to.

Or:

"Do you remember what I mentioned before?"

If the relevant information is still in accessible memory, the conversation can continue directly from there.

This continuity is a significant part of what makes AI companionship genuinely useful. If you'd like to compare how memory works across different AI platforms, see: AI Memory Comparison: How the Major Platforms Handle Context.

But memory raises another question:

What exactly has AI remembered?

For AI handling personal emotions, life decisions, and daily situations, this matters a great deal.

If users can't see which background information AI is drawing on, it becomes hard to understand how certain responses were formed.

So alongside "remembering more," there's an equally important direction for AI memory:

Letting users see what it remembers.

How HKSoka's Memory System Actually Works

HKSoka has a concrete memory mechanism that makes this continuity possible in practice.

Automatic learning, no manual curation needed
After each conversation ends, the system automatically extracts important information from the dialogue after an idle period, storing it as individual memory entries. You don't need to organise anything yourself.

Fully transparent, fully editable
You can view what the AI has remembered at any time, and modify or delete individual entries. You know exactly what background information AI is drawing on when it responds — no black box.

Document upload: two modes
HKSoka supports two document memory modes:

Both modes are fully customisable. For an in-depth look at the engineering behind the system: AI Long-term Memory Architecture: HKSoka's System Design.

Model selection
The platform offers several high-quality model options. Grok 4.5+ is available on the free plan; the paid plan includes Claude Sonnet 4.6+, Claude Opus and Fable, switchable as needed.

How to Use the Three Modes on HKSoka

HKSoka is a Hong Kong AI platform with persistent conversation and AI memory.

If you want to test the three modes above in practice, open a new conversation, paste one of the prompts below, and replace the bracketed content with your own situation.

Ready-to-use opening lines for each mode:

Information mode:

"I'm thinking about changing jobs. Please help me organise the key factors and options I should consider, without giving me recommendations."

Decision mode:

"I'm considering leaving my current job. Challenge my thinking and point out the risks I might be overlooking."

Companionship mode:

"Something happened today that left me feeling uncomfortable. I just want to talk through it — please don't analyse or give advice yet."

You can use English, Traditional Chinese, or Simplified Chinese depending on your preference.

If you want AI to hold certain background information long-term, you can tell it directly:

"Here is some background about me, please remember it: [your context]."

For regular users, this kind of background reduces the need to re-explain, so later conversations can pick up from existing context more easily.

The Limits of AI Companionship

This article needs to state one boundary clearly.

AI system responses are based on language patterns, working with textual context rather than clinical judgement. For AI responses touching on mental health, users need to evaluate reliability for themselves.

Therefore:

AI companionship can serve as a supplementary tool, suited to daily emotional processing, working through thoughts, and general conversation.

When professional evaluation is needed, human professional support should take higher priority.

For example:

In a crisis, contacting human support directly will be more appropriate than continuing to talk with AI.

Hong Kong Emotional Support Resources

Hong Kong has several support channels available, including:

Service hours and coverage may vary by organisation. Check each organisation's latest information before calling.

Finally

AI is available.

The question is what you need it to do right now.

HKSoka is a Hong Kong AI conversation platform with continuous memory, available in multiple languages. AI that remembers your background — starting from the mode you actually need.

Try HKSoka →