Segment ranking · November 2026 · Part of the IPO list series

The APAC 8

The emerging providers serving buyers across Asia-Pacific with AI in production: independent operators under 5,000 people, a named human team owning the work, and buyers served across a border. I screened 79 companies. Eight cleared every gate.

The segment, and why it matters in 2026

Asia-Pacific buyers now shop for outsourcing across borders as a matter of course. An Australian accounting firm builds its back office in Hyderabad. A Japanese fintech sends document work to Ho Chi Minh City. An Australian car importer runs its customer lines from Suva. I built this list for the buyer making that kind of decision, so it ranks one kind of company: an independent provider under 5,000 people, running AI inside work a named human team owns, serving buyers in one country from delivery in another.

This year the list also covers the dedicated offshore team, the model where a provider recruits, employs and manages people who work inside a client’s own operation. In this region that model is a large share of the real market, and leaving it out would have left the buyer with half a map. The same AI test applies to it as to everyone else.

The ranking

Eight that cleared every gate

I screened 79 companies, plus the AI-native operators named in their own band below. Eight cleared every gate with citable evidence. That is why this is The APAC 8. I did not add companies to reach a rounder number, and I did not cut any to make the list tidier.

No company paid to be here. No company can. There is no fee for inclusion, no fee for placement, and no way to buy a better rank. Sponsors of the wider research program never touch these scores.

Two companies share No. 5 on 2.80. My tie-break rules ask two questions in order. First, does a third party document the company’s AI running in live client work? Neither has that yet. VINASA names DIGI-TEXX’s DIGI-Xtract as a product, and Unity4’s platform reviews sit on software directories. Second, does the company hold a security certification I can see in a registry? I could not find one for either company. Both rules come up empty, so the tie publishes and the two are listed alphabetically.

Cross-list check: none of the eight holds a ranked seat on another segment list. Each uses one of its two slots.

Symbos

Weighted score 3.30

Australia · More than 2,000 people across five countries, lineage to 1991

Symbos carries a lineage back to 1991 through Recoveries Corporation and took its current name in 2024, when Allegro Funds bought the business. It now runs more than 2,000 people across Australia, New Zealand, Fiji, the Philippines and South Africa, serving more than 250 clients, among them Australian banks, insurers, utilities, telcos and government agencies. Frost & Sullivan gave it the 2026 Australian Transformational Innovation Leadership Recognition for customer experience management, and named its outcome-based commercial models in doing so. The company reports a generative AI layer that categorizes interactions at 92 percent accuracy for a health insurer and deflects 62 percent of inbound contacts for a regional water utility. It holds ISO 27001:2022 and ISO 9001 certificates I could see on the JASANZ register under Recoveries Corporation, and four of its leaders made the same public case this year, that CX AI should be judged on resolution and outcomes. Symbos came to me as a nomination through the IPO 25 form and was scored on evidence alone. The gap is its workforce record: reviews on the predecessor brand are the weakest in this field, 2.3 on Glassdoor and 2.8 on Seek, and a current employee story told in public would fix that.

Booth

Weighted score 3.05

Philippines · Dedicated teams and managed delivery with its own AI quality engine

Booth was founded in 2013 by Carmen and Jamie Booth and now has about 1,900 people in Manila, Cebu, Tarlac and Bogotá. It sells dedicated teams, fully managed processes and an AI-augmented operations line, and it built its own quality engine to sit across all three. In twelve days RubriCore scored more than 13,500 live conversations, and when Booth checked it against more than 570 human audits the two agreed better than 90 percent of the time. Its BoothAgent line pairs AI agents with operators on claims, accounts payable and reconciliation, and flags every edge case for a person. It prices that work per transaction or by management fee. Its buyers sit in Australia, including Active Utilities and Cluey Learning, and in the US. It holds 4.8 on Clutch across 25 verified reviews, it has been a certified B Corp since June 2025, and the FT named it among Asia-Pacific’s high-growth companies. One tension is worth naming: Booth still sells staffing and employer-of-record services, and it ranks because every line it runs cleared the gates. What would move it higher is a named client putting numbers on RubriCore and a SOC 2 report a buyer can request, and both are fixable by publishing.

CDM Direct

Weighted score 2.94

Australia · Australian automotive and customer work delivered from Fiji

CDM Direct has run customer work for Australian companies for more than 20 years from Braeside, Victoria, and most of its 300-plus people now work from Fiji. Its automotive depth is the clearest vertical story on this list: Toyota, Mazda, Volkswagen, XPeng and FCA Australia are named clients, and 365 Roadside Assist runs more than 10,000 calls a month through it. Its own agent-assist platform recalls knowledge, reads sentiment and prompts for compliance in live calls, and by its own count it cut first data capture from 120 seconds to 20. Agents can override every suggestion, and negative sentiment and high-stakes decisions route to a person. No third-party review site covers its Fiji workforce, so I left workforce health out of its score and rescaled the total. The gaps are that missing employee record and pricing that is still partnership-based, and both are fixable by publishing.

Talent Formula

Weighted score 2.85

India · Offshore accounting teams for Australian and UK firms

Talent Formula builds offshore accounting teams in India for accounting firms in Australia and the UK, from an operating base in Hyderabad with offices in Sydney, London and Burbank. It has done that since 2014 and has more than 200 people. Its depth inside one network is unusual: PKF Sydney is building its offshore growth on Talent Formula teams, and PKF UK and PKF Francis Clark sit on its client wall. Its own platform, TFX, runs privately hosted AI models over a firm’s knowledge base with a person overseeing the output, and its Digital Employees handle invoice processing, reconciliation and report runs around the clock. It names ISO 27001 and ISO 27701 on its site. Employees rate it 3.4 on Glassdoor across a small sample of 20. The gaps are a production metric for TFX and a published pricing model, and both are fixable by publishing.

05Shared seat

DIGI-TEXX

Weighted score 2.80

Vietnam · Document and data work in Ho Chi Minh City since 2003

DIGI-TEXX has run document and data work out of Ho Chi Minh City since 2003, and it describes itself as the first fully German-invested BPO in Vietnam. It builds its own tools: DIGI-Scan, DIGI-Xtract and DIGI-Vault run inside its document processing, and VINASA named DIGI-Xtract a Top 10 Digital Transformation Solution. On invoice work for Japanese clients, machine extraction runs first and operators verify the output. The buyers are named and they span markets: Money Forward in Japan, Manulife Vietnam at home, and FamilySearch, the ILO and Carrefour Belgium further out. Its teams work in Vietnamese, English, German, Japanese and Chinese, it shows ISO 9001 and ISO 27001 certificates, and employees rate it 3.7 on ITviec with 78 percent recommending. It also runs DIGI-CONNECT, a training program for people from underserved communities. What holds the score is disclosure: there is no published pricing model and no production metric for DIGI-Xtract, and both are fixable by publishing.

05Shared seat

Unity4

Weighted score 2.80

Australia · The home-based contact center, pioneered in 2000

Unity4 pioneered the home-based contact center in Australia in 2000 and now runs more than 1,100 home-based agents across Australia, New Zealand, the UK, the US and Fiji. Many of its agents are deliberately mature, people who bring life experience to fundraising and service calls. Its own RapportCMS platform scores calls with AI voice analytics and coaches agents in live work. Its client wall is real and broad: Optus, Flight Centre, Kia, H&R Block, Curtin University, Perpetual and the Fred Hollows Foundation in Australia, and Plunket in New Zealand. Its employee record is among the strongest I saw, 4.2 on Glassdoor across more than 50 reviews with 88 percent recommending. The platform’s own page has gone dark, the commercial model now reads “Flexible Commercials,” and I found no security certification, and all three are fixable by publishing.

OutForce

Weighted score 2.70

Philippines and Australia · Dedicated teams in Manila, Cebu and Clark with its own AI layer

OutForce is Australian-owned, was founded in 2018, and has more than 750 people in Manila, Cebu and Clark working on Australian, US and UK hours. Its teams work inside each client’s own systems while OutForce runs recruitment, HR, payroll, IT and compliance. It has built its own AI layer on top: OutForce AI answers calls, recognizes existing customers from the CRM, books meetings and raises tickets, and hands the exceptions to the client’s dedicated OutForce staff. Its named clients run from Amazon One Medical, where a finance and procurement team has worked on US hours for several years, to Access4 and Webprofits in Australia. Great Place to Work has certified it in the Philippines, and it names ISO 9001 and ISO 27001 on its site. The gap is proof: no client is named running OutForce AI yet, and one named deployment with a number on it would fix that.

Sourcefit

Weighted score 2.65

Philippines · Australian and Singapore buyers from a Philippine base

Sourcefit was founded in 2009 by Andy Schachtel and now employs more than 2,000 people across the Philippines, South Africa, the Dominican Republic, Madagascar, the UK and Armenia. It serves buyers in Australia and Singapore from its Philippine base, alongside a large US and UK book. Its own Knit platform runs analytics and engagement inside client delivery teams, and in a July 2026 podcast Schachtel described building human oversight, audit trails and escalation into its WorkingAI line, and said the company is pursuing ISO 42001. Its Clutch record is strong at about 4.8 across 12 reviews, it names ISO/IEC 27001:2022 on its site, and employees rate it 3.5 on Glassdoor across 107 reviews. What would move it up is a named client willing to put numbers on Knit or WorkingAI and the name of the body behind its ISO certificate, and both are fixable by publishing.

Outside the ranking

Special mentions: domestic standouts

These four are among the strongest operators I assessed. Each one serves buyers in its own country from delivery in that same country, so each sits outside the ranking. I list them without scores. Each has a named path in.

Telnet Services (New Zealand). Telnet keeps every agent onshore in New Zealand. Its Tulip voice AI took a third of inbound calls during ex-cyclone Tam with no hold time, by its own account, and it publishes named results for Z Energy and Lotto NZ along with pay-per-interaction pricing. It markets to Australian buyers now. Re-entry: one named current Australian client.

Adish (Japan). Adish has run customer support and online community moderation for Japanese brands for more than a decade. Its ISO 27001 registration dates to 2009, it holds a current Privacy Mark, and its filing names Stripe Japan as its top customer. Its CS STUDIO model hands routine contacts to AI and complex ones to operators. It also runs a small team in Makati for its Japanese clients. Re-entry: delivery outside Japan becoming a real share of its work, or one named buyer outside Japan.

Caster (Japan). Caster is one of Japan’s largest fully remote companies. On July 1 it launched CASTER NEO, where an AI front takes requests around the clock and human coordinators own the final call on quality. It holds a registry-verified ISO 27001 for its Saito site and a current Privacy Mark. Re-entry: one named APAC buyer outside Japan.

Daythree (Malaysia). It built its own AI stack (Daisy, SAIGE, FAITH and CAIRA), and FPT took a 10 percent strategic stake. Its named work is Malaysian work delivered from Malaysia, and its site now says 4,500+ people, close to the size line. Re-entry: one named buyer or delivery site outside Malaysia, and a filed headcount under 5,000.

Ones to Watch

Real capability that missed the ranking on evidence. Each has a named path back.

Hammerjack (Australia and the Philippines). See the feature below. Re-entry: one AI deployment in live client work, published.

TOA Global (Australia). More than 4,000 accountants in the Philippines working for accounting firms in Australia, New Zealand and North America, with its own training institute and a 4.8 rating across 551 JobStreet reviews. I found no AI tool named in its delivery. Re-entry: one named AI tool its teams run for clients.

ConnectOS (Australia). More than 2,500 people in the Philippines for clients including IDP and Chemist Warehouse, and 4.9 on Glassdoor across about 1,000 reviews. Its AI work so far is research about AI readiness. Re-entry: one named AI program or tool for its own staff.

Shore360 (Philippines). An Australian-owned team builder in Clark with one of the strongest employee records in the region, 4.8 on Glassdoor across more than 100 reviews. Its AI wording names no tool. Re-entry: one named AI tool in client work.

Penbrothers (Philippines). The only staffing firm I found that publishes a fixed monthly fee per person, with named clients including DesignCrowd. I found no AI in its delivery. Re-entry: one AI tool running in client work.

Outsourced (Philippines). Named clients including Mimecast, Genesys, Shootsta and 2XU, a Great Place to Work certification, and 4.0 on Glassdoor across more than 300 reviews. Its AI offer today is AI talent placed with clients. Re-entry: name the AI tools its teams are trained on and run in client work.

Scicom (Malaysia). Scicom has been founder-led since 1997 and listed on Bursa Malaysia since 2005. Its gov-tech work is one of the deepest vertical files I saw: six national platforms serving more than 320 institutions and applicants from over 190 countries. Its employees rate it 4.0 on JobStreet across 215 reviews. Its AI story is real in intent, with AI-powered customer experience on its homepage and a partnership with Telekom Malaysia to build more, and I could not find a named AI product running inside its own delivery. Re-entry: one named AI product running in its own delivery, with a dated source.

Antasis (Singapore). A rare Singapore-based AI customer-experience operator with offices across five Southeast Asian countries. Its AI work is still in pilots, which this method does not credit. Re-entry: one pilot moved into live production with a named client.

Flat Planet (Australia). A family-owned firm since 2010 with 400+ professionals in Sydney and Makati serving Australian service businesses. Its AI wording is still general. Re-entry: one named AI deployment in its own delivery.

BPO.MP (Vietnam). A Vietnamese-Japanese joint venture built to serve Japanese buyers from Da Nang. It now lists its own AI products, and none is shown running inside its delivery yet. Re-entry: one AI offering shown live for a named client.

TWC (Korea). The one small independent Korean operator I found with AI in its contact-center work, with named clients including Coinone, Pinkfong and Hwahae, and a selection to Korea’s 2026 Unicorn Bridge program. I could not confirm its headcount or founding year. Re-entry: both confirmed, plus one named client in Japan, where its CEO says the business is growing.

Companies in this band can carry a featured buyer profile here. That placement is paid, always labeled, and never changes a score or creates a ranking.

Other providers you should consider

Hammerjack, the workforce story this field should study

Hammerjack does not appear in the ranking, and the reason is my own rule: its site no longer describes AI in its client work, and I do not rank what I cannot verify. It appears here because its workforce record is the best of any company I assessed this cycle, and it deserves more than a footnote.

Hammerjack is Sydney-based and runs its delivery from Makati, Manila and Pampanga, where more than a thousand professionals do bookkeeping, accounting and back-office work for Australian small and mid-sized businesses. Microsoft profiled it in 2017 as the first provider to offer those businesses a fully managed model. Great Place to Work certified it in the Philippines from March 2026 to 2027, and it placed No. 3 among medium-sized companies in the 2026 Best Workplaces in IT-BPM Philippines list. Its people rate it 4.6 on Glassdoor across 96 reviews, with 90 percent recommending it to a friend. Those numbers come from the people doing the work. If Hammerjack publishes one AI deployment running in live client work, the 2027 gate is open.

The parallel market

The AI-native band

Four companies in this region sell AI agents as the workforce itself: WIZ.AI (Singapore, voice agents for ASEAN banks, with a Series B led by SMBC’s venture arm), AI Rudder (Singapore, voice and chat agents in more than 15 languages), Lorikeet (Sydney, AI concierges that resolve tickets, with a US$35 million Series A in August 2025) and Skit.ai (Bengaluru, voice agents for collections). This is a real market and buyers should know it. It is also a different product. My test for the ranking is one question: does a named human team run delivery and own the outcome of every client conversation? An agent vendor answers no by design, so these four are named here and never ranked.

Over the cap

Benchmarks

Six companies would be strong candidates on capability or model and sit above the 5,000-person line. TSA Group (Australia) holds a registry-verified ISO 27001:2022 and publishes a clear escalation design for its voice AI, including how sensitive calls reach a person; its own careers page says it has grown past 5,000. Bewith (Japan) runs one of the region’s most developed AI and human hybrids, Omnia LINK, with more than 9,000 people. Acquire BPO (Philippines) reports more than 9,500. MicroSourcing (Philippines) is the best-known name in the dedicated offshore team model, with more than 10,000 people, and sits inside a larger group. Cloudstaff and Emapta (Philippines) report more than 5,000 and more than 13,000. I name them as benchmarks for what the smaller firms are working toward.

Coverage gaps

One market with no ranked company

Korea has no ranked company this cycle, and I want buyers to know that. The AI-forward Korean operators I found are large or part of bigger groups. UBASE runs more than 22,000 seats and says AI handles 30 percent of its consultations. TWC is the one small independent Korean candidate, and it is in Ones to Watch. India has a ranked company in Talent Formula, and Vietnam has one in DIGI-TEXX.

The trend read

Three findings from this cycle

First, the dedicated offshore team is where the AI gap is widest. I screened more than forty staffing firms across the region, and three of them show AI running inside their own delivery. Most market AI-skilled people, which is a different thing from AI in the work. Any of them can close that gap by naming one tool its teams run for clients.

Second, APAC governance can be checked. The Australian, New Zealand and Japanese registries let me verify certificates for four companies in this field, and one of the eight that rank, Symbos, is among them.

Third, evidence moves fast. Between July and October one company’s AI page went dark, another’s regional claim disappeared, one firm took AI off its homepage, and one grew past the size cap. Every score here was checked again on October 20.

Methodology

How we rank

This list covers independent providers under 5,000 staff delivering managed outsourced work or dedicated offshore teams, where a named human team runs delivery and owns the outcome, with AI running in production. AI-agent vendors and pure software sit in the band above, never in the ranking. I score eight dimensions with published weights: AI deployment depth 20%, outcome accountability 20%, governance and failure readiness 15%, human-AI operating model 10%, workforce health 10%, vertical depth 10%, client proof 10%, positioning clarity 5%.

Every company passes eight gates before it is scored: fewer than 5,000 staff; three or more years operating independently; at least one AI capability live in production, with pilots excluded; independent ownership; managed process delivery or dedicated offshore teams as the core business, with recruiting-only firms, freelancer marketplaces and software left out; buyers served across a border; inclusion that cannot be bought; and no gig marketplaces. For this list, a company passes the border gate when it serves buyers in one country from delivery in another, or names buyers in more than one country. Operators whose buyers and delivery sit in the same country appear as special mentions. For the AI gate, a company’s own named AI product counts when it runs inside the company’s own delivery. AI used only to recruit, and AI-skilled people supplied to clients, do not count.

The evidence rules are simple. A 4 or 5 needs a citable source. The top score on any dimension needs outside validation: a named client on the record, a certificate in a registry, a review record built from verified clients. Self-published named evidence caps one rung lower, and anonymous evidence a rung below that. Where a company controls whether to publish something, such as its pricing, silence scores 2. Where evidence depends on third parties that do not cover a company, I leave that dimension out, rescale the total, and say so. Workforce health draws on Glassdoor first and local review sites where Glassdoor is thin, with every figure dated. Published thought leadership counts toward positioning clarity and nothing else.

Ties break on third-party documentation of AI in production first and registry-verified governance certifications second. Where both come up empty, the tie publishes, as it does at No. 5 this cycle. A company may hold a ranked seat on at most two segment lists; none of the eight is near that cap.

The ranking above is editorial and follows the published methodology. No company can pay for placement or position. Featured profiles beside the list are paid, and every one of them is labeled. No company on this list is a current or former Pivot AI Global client. The full scoring matrix, with a dated source for every score, is available for audit.

Nominations

Think a company is missing?

The hardest part of building this list was finding the companies. If an Asia-Pacific provider belongs here and is missing, make the case. It goes straight into the next review.

More lists

The IPO 25 (flagship) All lists Trust & Safety 10 · August The LPO 11 · September The SDR 10 · October