Client Experiences

What our clients say about working with us

We prefer to let the work speak. The reviews below are from organisations that have completed engagements with Koi Stream between January and April 2025.

โ† Back to Home

4.8

Average Rating

12+

Engagements Completed

3

Industry Sectors

40+

Staff Trained

Reviews

From the teams we worked with

AK

Angela Kwok

Front Office Manager ยท Boutique Hotel, Sheung Wan

The training was practical and not rushed. Within two weeks our desk staff were using the assistant confidently during check-in peaks. The month of follow-up support mattered โ€” questions came up that we hadn't thought to ask beforehand, and the Koi Stream team answered them quickly. One thing I appreciated: they were honest when the tool couldn't answer something, rather than inflating what it could do.

April 2025

BL

Brian Lau

Regional Retail Manager ยท Apparel Chain, Multiple Locations

The consent process was handled smoothly โ€” our store staff understood why it was happening and there was no resistance. The monthly coaching memo is now something our store leads actually look forward to. It highlights specific moments rather than just giving aggregate scores, which makes the coaching conversations much more concrete. The quarterly review with management gives us a useful pattern overview.

March 2025

MC

Margaret Chan

Operations Director ยท Shared Service Centre, Kowloon

We were cautious going in, and Lena's team respected that. The operational review at the start was thorough โ€” they mapped our claims processing stream before suggesting anything. By month three, our case handlers were producing first-draft correspondence faster and with fewer revision rounds. The scaling documentation at the end was genuinely useful. We're planning to extend to a second stream this summer.

April 2025

RT

Raymond Tang

General Manager ยท Serviced Apartments, Tai Koo

Our property serves a mix of long-stay and short-stay guests, which means the front desk handles a wide range of questions. The Koi Stream team prepared the knowledge base carefully โ€” they visited us to understand which questions came up most often and what kinds of answers guests actually found helpful. Staff engagement was better than I expected. Two months in, the tool is part of normal desk operations.

March 2025

YH

Yvonne Ho

Store Training Lead ยท Luxury Accessories, Harbour City

The consent documentation process added a step, but it was straightforward and our staff appreciated being consulted. What I found most useful was the way the coaching memos were written โ€” they focus on specific observable moments, not just scores. That makes them usable in a 20-minute team briefing, which is the reality of retail training time. One improvement I'd suggest: a summary page at the front would help for managers reviewing multiple locations.

April 2025

DW

David Wong

Head of Back Office ยท Insurance Group, Quarry Bay

We were hesitant about AI touching correspondence that goes to customers, and that hesitancy was taken seriously. The human review step is real โ€” it's not a formality. Our handlers read and approve every AI draft before it goes anywhere. After five months, the process is natural and the team's confidence in the tool is solid. The monthly reporting gave our senior management the visibility they needed to stay comfortable with the pilot.

April 2025

Case Studies

Selected project summaries

Hospitality ยท Sheung Wan

Reducing guest query handling time at a 45-room boutique hotel

Challenge

A two-person front desk team handling peak check-in periods found that 30โ€“40% of guest queries required searching multiple sources โ€” house rules, transport options, local recommendations. Response times were inconsistent and staff found the information search distracting during busy periods.

Solution

Koi Stream prepared a knowledge base from the hotel's existing content โ€” house rules, curated local attraction lists, transport guides โ€” and configured an AI assistant accessible from the front desk terminal. Staff training took three sessions over two weeks.

Results

Within six weeks, the average time to answer a guest query dropped noticeably. Staff reported feeling more confident handling a wider range of questions. The hotel's guest comment cards in March 2025 showed positive mentions of front desk responsiveness at a higher rate than the previous quarter.

"The tool doesn't replace what our staff know. It gives them faster access to things they used to have to look up. That's the right balance." โ€” Angela Kwok, Front Office Manager

Retail ยท Multiple Locations

Systematising coaching input across six retail stores

Challenge

A regional retail manager responsible for six locations found that coaching quality varied significantly between stores. Some store leads held structured coaching sessions; others struggled to find consistent time. There was no systematic way to identify coaching priorities across the network.

Solution

Koi Stream implemented the retail coaching service beginning with a pilot at one location. Consent documentation was completed for all participating staff. The AI reviews consented interactions, identifies representative moments, and produces a monthly coaching memo structured for use in team briefings.

Results

The pilot location's store lead reported that the monthly memo gave coaching sessions a consistent structure for the first time. The first quarterly review with regional management surfaced two recurring themes the manager had not previously identified. The service is now extending to two additional stores.

"The memos are specific enough to be useful in a 20-minute briefing. That's what retail training time looks like in practice." โ€” Brian Lau, Regional Retail Manager

Service Centre ยท Kowloon

AI-assisted correspondence drafting in a claims processing team

Challenge

A back-office team handling insurance claims correspondence faced inconsistent output quality across handlers, with senior staff frequently revising drafts from junior colleagues. The revision cycle added handling time and created a dependency on senior staff availability.

Solution

Following an operational review of the claims correspondence stream, Koi Stream configured AI-assisted drafting within the team's existing case management workflow. Each draft is reviewed and approved by the handling officer before dispatch. Implementation proceeded over a five-month period with monthly reporting to the operations director.

Results

By month three, the revision rate on first drafts from junior handlers had decreased measurably. Average correspondence handling time reduced. Senior staff reported spending less time on revision and more on complex cases requiring judgment. The scaling documentation produced at close of engagement is now being used to plan the team's second stream pilot.

"The human review step is real. Our handlers approve every draft. After five months, the process feels natural." โ€” David Wong, Head of Back Office

Contact

Reach the team

Phone

+852 5629 4871

Address

9 Tai Po Road, Tai Po, New Territories, HK

Office Hours

Monday โ€“ Friday: 9:00 โ€“ 18:00
Saturday: 9:00 โ€“ 13:00

Certifications

Professional standing

HKCS Member

Hong Kong Computer Society professional membership since 2022.

PDPO-Aligned Practice

All services designed in line with PCPD AI guidance for Hong Kong.

Consent-Led Processes

Consent documentation standard in all services involving recorded data.

Transparent Scope

Fixed-scope engagements with written deliverables defined in advance.

See if this could work for your organisation

A brief conversation is usually enough to know whether there is a good fit. We do not require a formal brief to get started.

Start a Conversation