CX Outsourcing Review

Birchy
Outsourcing & BPO
Research

TaskUs review

Assessed against a published 100-point framework from public sources, not from a client engagement: TaskUs scores 68/100 and is tiered Strong Performer on 12 evidence claims, each graded and sourced below.

Headquarters
United States — New Braunfels, Texas — Nasdaq: TASK Form 10-K principal executive offices: 1650 Independence Drive, Suite 100, New Braunfels, Texas 78132 E1
Official site
taskus.com
Operating since
2008 Founded 2008 by Bryce Maddock and Jaspar Weir; TU TopCo, Inc. incorporated in Delaware 27 July 2018 and renamed TaskUs, Inc. 14 December 2020, per Form 10-K E1
Score
68 of 100
Global Edition
Strong Performer · rank #12 of 29 · tied with TTEC, Firstsource, Transcom, Konecta, Movate and TDCX
Delivery Proof (of 45)
35
Capability Fit (of 55)
33

Evidence on file

Each line is a datapoint with its grade. Grades are the ones the anchors act on: a sub-criterion scored 5 requires at least one E1 or E2 source, and evidence that is entirely E3 caps the sub-criterion at 3.

12 claims on this sheet, by grade: E1 5 E2 3 E3 5 E5 2 Counts sum above 12 because 3 claims cite more than one grade.

  • ~65,500 teammates across 31 sites in 13 countries and $1,183.5M service revenue at 31 Dec 2025; Digital CX 56% of revenue, Trust & Safety 26%, AI Services 18% FY2025 Form 10-K, filed Mar 2026 E1
  • Client concentration disclosed in the FY2025 10-K: largest client 26% of revenue, top five 45%, top ten 58%, across ~200 clients. This is the single largest commercial risk on the record for this provider FY2025 Form 10-K E1
  • Everest Group Leader in Trust and Safety Services PEAK Matrix for the third consecutive year; also Major Contender and Star Performer in CXM EMEA 2025 and B2B Sales 2025, Major Contender in FCC Operations 2025, Leader in Data Annotation & Labeling 2024, and Frost Radar CXM Outsourcing North America 2025 Leader taskus.com analyst reports E1
  • The FY2024 Form 10-K names SOC 2 Type 2, ISO 27001, HITRUST and PCI-DSS, with Item 1C stating these are pursued via independent third-party assessment and describing the security governance chain — but no certificate number, issuing body, validity date or site scope is published anywhere FY2024 Form 10-K, Item 1C E1E5
  • The only public document giving certification scope is a company security white paper dated Oct 2017: SOC 2 Type II covering three Manila Metro sites plus San Antonio, PCI DSS Level 1 across five sites (QSA ControlCase, initial certification 2015). It also describes concrete controls — least-privilege role provisioning, 2FA for remote and privileged access, biometric floor access, per-campaign VLAN segmentation, micro-segmentation, MDM, no mobile devices on production floors. Nine years stale against today’s 31 sites TaskUs Security & Compliance white paper, Oct 2017 E3
  • The privacy statement, last updated 2 Aug 2026, publishes no DPA, no subprocessor list and no named DPO (a Privacy Office postal address only), no GDPR Article 27 representative, and explicitly declines a residency guarantee — "TaskUs' data centres are hosted in various countries around the world" taskus.com privacy policy E3E5
  • The Global AI Policy published 3 Jun 2026 is a genuine dated governance artifact: an AI Advisory Board, a stated human-in-the-loop model proportional to risk and complexity, client data excluded from internal AI initiatives, mandatory PII redaction before AI processing, and annual review from 5 October taskus.com global AI policy E3
  • Published AI outcome figures are unattributed and company-sourced: 30% AHT reduction and +5 CSAT for PromptAI, +7 CSAT overall, 45% processing-time decrease across 33 active campaign deployments — no client named, no baseline, no sample size taskus.com AssistUs E3
  • Independent trade coverage characterises the agentic-CX position as "a proof gap rather than a capability gap", finding the strongest published case lacks "baseline, sample, metric definition, and production results"; it also records AI Services growth decelerating from 36.1% in Q1 2026 to 25.8% in Q2 2026 while adjusted EBITDA margin fell from 21.3% to 19.1% CX Today, 2026 E2
  • Named, dated, quantified client cases across at least three verticals: Grow Therapy (healthcare, 29 Apr 2026 — 98% RCM accuracy, 75% faster claims processing), Brex (fintech, 23 Jan 2026 — up to 99% CSAT, 24/7 coverage), Badger Technologies (retail AI, 8 Jun 2026 — 7.7M+ images per month, $80K+ saved), plus DailyPay, Quavo and MoneyLion. None is corroborated outside taskus.com taskus.com insights E3
  • Independent client-review footprint is thin: the G2 profile carries 11 reviews at 4.0/5 spanning Oct 2023 to Mar 2026 with review text behind a login wall; the Clutch profile shows "Not yet reviewed" while listing a $50,000+ minimum project size. Employer side: Glassdoor 4.0/5 across 18,084 reviews, 80% recommend, 89% CEO approval, with disclosed voluntary attrition rising from 22.2% in FY2024 to 26.6% in FY2025 G2, checked Aug 2026 E2E1
  • Remains an independent public company: SEC filings continue through Aug 2026 (Q2 2026 revenue $308.9M), a 2026 annual shareholder meeting was held 26 May 2026, and a $3.65 per share special dividend was paid Mar 2026 following a $500M term loan refinancing — consistent with the Blackstone and co-founder take-private having been terminated on 8 Oct 2025 SEC filing history and company announcements E2

Where this provider stands out

Computed from this provider’s anchor scores against the 29-provider field it is scored in, from the 12 public claims above: the sub-criteria where it sits furthest above the field median, and furthest below. Selection is by distance from the field, not by raw score — a 5 where most of the field scores 5 says nothing. These are scores for what a provider publishes, so a low score can mean a weak record or a quiet one, and this block cannot tell them apart.

Furthest above the field

  • employee-side signal A4.1 5/5 — 4 above the field median of 1; joint highest, with 2 of 28 others
  • named cases A1.3 4/5 — 2 above the field median of 2
  • information security A3.1 4/5 — 1 above the field median of 3; joint highest, with 7 of 28 others

Furthest below the field

  • technical support depth B1.2 1/5 — 2 below the field median of 3; joint lowest, with 2 of 28 others
  • native-language delivery B2.1 2/5 — 1 below the field median of 3
  • QA methodology B3.1 2/5 — 1 below the field median of 3

Criterion scores

Sub-criterion anchor scores in criterion order, with the weighted criterion total computed from them. Multiply each anchor by its sub-criterion weight over five and sum: that is the whole arithmetic, and it is why this total can be checked without taking anyone's word.

Code Criterion Sub-criterion anchors Weighted Max Note
A1 Independent client evidence 4 / 3 / 4
A2 Market presence & tenure 5 / 5 / 5
A3 Certifications & compliance 4 / 2 / 1
A4 Delivery stability 5 / 5 Glassdoor 'TaskUs' (E350702): 4.0 across 18,084 reviews, 80% recommend, 89% CEO approval, marked Engaged Employer, read 28 Aug 2026. Meets the anchor's 4.0 threshold across at least 25 reviews with management-response activity. The client-review cross-check is satisfied rather than waived: the G2 profile carries 4.0 across 11 reviews and contains no contradiction on team stability. The displayed value is scored as displayed.
A5 Transparency & verifiability 4 / 4
B1 Service scope & channels 3 / 1 / 3
B2 Language & geographic reach 2 / 5 / 5
B3 Quality management system 2 / 3 / 3
B4 Security & data protection operations 4 / 1 / 4
B5 Flexibility & commercial model 1 / 3 / 1
B6 Technology & AI operations 3 / 3 / 4
B7 Vertical expertise 5 / 4

Axis A subtotal 35 of 45 · Axis B subtotal 33 of 55 · total 68 of 100. Each weighted figure is (anchor score ÷ 5) × sub-criterion weight, summed — the rule printed in the appendix.

Flags and open questions

  • Largest remaining evidence gap: nothing procurement-grade published on certificate scope or data processing terms

Challenge this sheet

Every line above is meant to be checkable. If a datapoint is wrong, out of date, or missing a certificate identifier we could not find, the corrections process is open: submit the evidence and the score is re-derived through the anchors. Facts are correctable; scores follow facts.