// analysis
Qantas: AI layoffs and the work moving offshore
Open the documentsThe source documents
The documents cited in the analysis and, where identified, their associated passages. An automatically detected association is not a verified claim.
Consult the references in the article below. No document has been identified for this shortcut.
Sources and reading context
Contents9 sections
Qantas plans to transfer support services to Accenture, largely in India, through a programme expected to affect around 850 roles over the coming years. Announced on 10 November 2026, the partnership combines changes of employer and location with investment in data systems and automation. The first workforce changes are expected by mid-2027. The announcement sets out a restructuring whose employment effects will unfold over several years. S01 S02
That distinction sits at the centre of the Australian debate over AI washing. Unions and experts have questioned whether some companies invoke AI to present offshoring and cost reductions more favourably. Reuters reported those allegations in a dispatch carrying a 12 November Sydney dateline. Judging them requires matching a specific company claim to evidence about the work, its costs and the people performing it. S03
Qantas combines service transfers and automation
Qantas says Accenture will provide services mainly from India and partly from Australia. AI and automation are intended to be part of the arrangement from the outset, with further transformation over time. Transfers, systems work and automation can overlap. The documents provide no allocation of the 850 affected roles between those mechanisms. S01 S02 S04
Individual outcomes are also subject to consultation and redeployment efforts. An affected role is therefore a different measure from a completed involuntary departure. It tells us less still about the hours of work a machine has removed. Those outcomes will need to be tracked as the programme develops. S01 S02
The technical rationale is substantial enough to examine. According to Qantas, its estate contains more than a thousand legacy applications. Accenture’s remit includes modernising data and connecting systems. iTnews reports this explanation from the company; it does not independently audit the application estate. S04
A more consistent process can be easier to outsource and easier to automate. Consolidating records or removing duplicate entries may help a supplier take over an activity while also changing the human work it requires. The useful investigation follows both changes: where the task moves and how its execution evolves.
Follow the work across the employer boundary
Consider a supplier invoice. It must be received, matched to an order, checked for discrepancies and authorised for payment. Software may extract its data while people retain responsibility for disputes and approval. A job title or payroll record captures only part of that chain.
Outsourcing changes the organisation doing the work. Offshoring changes its country. Automation changes the human involvement required for particular tasks. These dimensions can combine within one contract, so each step needs an identified operator, location, tool and amount of human time.
Telstra offers a concrete example of an employer boundary changing. On 15 January 2025, it announced a proposed seven-year joint venture with Accenture, subject to employee and union consultation. Its core Data & AI staff in Australia and India would receive offers to join. A joint announcement published on 13 May stated that the venture had commenced in April. Work can move to a new employer while continuing to require specialists. S07 S08
The financial assessment must cross that boundary too. A customer’s payroll savings sit alongside its service bill, migration costs, software and retained oversight. Depending on the pricing arrangement, a supplier’s efficiency improvement may benefit the provider, the customer or both. The public documents examined here do not establish that division for the Qantas contract.
An operational assessment also needs to count rework. An answer delivered quickly can consume more time later through corrections, reopened cases or escalation. Cost and hours per successfully resolved case are more informative than the number of initial answers alone.
Outsourcing can raise revenue per employee
Take a wholly hypothetical company with revenue indexed to 100 and 100 full-time-equivalent workers. It outsources an activity, leaving 80 FTEs inside the business and 20 at the supplier. Assume identical working time, output, prices and quality before and after.
Revenue per internal FTE rises from 100/100 to 100/80, or from 1 to 1.25: an increase of 25%. Total labour remains 100 FTEs. The ratio rises because the supplier’s workforce has moved outside its denominator. No operational improvement is assumed, and the illustration reconstructs no named company’s accounts.
Official productivity statistics use a different framework. The Australian Bureau of Statistics calculates labour productivity using the volume of gross value added and hours worked. Value added deducts intermediate inputs, including contractor services. Revenue per employee cannot be substituted for that measure. S05
Keeping a process comparison consistent means following the same service across providers and counting all of its labour. The customer’s accounts require another step: purchases of supplier services must be included when calculating value added. The illustration leaves the supplier’s charge unspecified. A financial assessment would also need the full migration and oversight costs. S05 S15
A lower service bill might reflect different pay rates, a better process or both. Headcount at the customer alone cannot distinguish those explanations. It is a starting observation that needs to be matched to the supplier’s activity.
The guide to productivity and unit labour costs explains how to compare the measures behind these changes.
CBA revised its staffing assessment
On 21 August 2025, Commonwealth Bank reversed a decision affecting 45 customer-service roles after introducing an AI voice bot. The bank acknowledged shortcomings in its assessment of the roles required. ABC reported that affected staff could remain, seek redeployment or leave. S10
The Finance Sector Union said call volumes had risen and managers were helping to absorb the workload. Its account would need to be tested against call data to separate changes in demand, the bot’s effects and other possible causes. S09
The documented finding is a revised management assessment. To evaluate the technology’s role, an investigator would need the original demand forecast, human hand-off rates, time spent on complex cases and overtime. Those records could distinguish an error in staffing assumptions from a change in demand or a problem with the tool.
Atlassian links restructuring to investment and skills
Atlassian’s announcement presents another combination of mechanisms. On 11 March 2026, it said it would reduce its global workforce by approximately 1,600 people, roughly 10%. The company cited investment in AI and enterprise sales, a stronger financial profile and changes to its organisation. Its chief executive also explicitly linked AI to the skills mix and the number of roles required in some areas. S06
That account includes a funding decision and expectations about future work. It does not quantify the tasks automated or the hours saved. Assessing technical substitution would require operational evidence alongside the company’s explanation of its restructuring.
These cases cannot form a single count of jobs destroyed by AI. Qantas describes a future programme, CBA a reversed assessment, Telstra an employer transfer and Atlassian a global restructuring. Their dates, units and geographical boundaries differ.
The Barclays analysis of AI, job cuts and productivity evidence examines the same attribution problem across restructuring, cost savings and technical improvements.
Assistance can improve human work
The presence of employees at a supplier is compatible with a meaningful contribution from AI. A study by Erik Brynjolfsson, Danielle Li and Lindsey Raymond, published in the Quarterly Journal of Economics on 4 February 2025, examines the staggered introduction of an assistant among 5,172 customer-support agents. It estimates an average 15% increase in issues resolved per hour in that firm, with substantial differences across workers. The system supports people who remain responsible for the conversation. S11
Such assistance could help an outside provider take over unfamiliar activities. Applying that possibility to a particular contract requires evidence on comparable work, including its difficulty and quality.
Earlier research also shows why technology and outsourcing can develop together. An OECD working paper published in 2021 studies broadband deployment and domestic outsourcing in France during the 2000s. Its findings concern that transition. They offer a mechanism to investigate, rather than a causal estimate of Australian generative-AI programmes. S14
National data leave another gap. A staff note by Mike Major, published by the Reserve Bank of Australia on 9 November 2026, finds limited evidence of broad disruption linked to AI while identifying possible weakness among younger workers and more exposed occupations. The analysis is descriptive; its latest complete occupational data cover February 2026, with partial coverage for May. November’s restructuring announcements fall outside that observation period. S12
Jobs and Skills Australia’s 2025 synthesis likewise emphasises heterogeneous effects and assistance to existing tasks. Exposure to a technology is different from its actual adoption, and neither supplies an individual’s probability of redundancy. S13
The analysis of AI and learning at the start of a career considers how changing entry-level tasks can affect the route to experience.
Make the employment claim auditable
A useful AI-washing investigation begins with the claim being made. An ambition to automate, a measured operational improvement and an explanation for completed job losses carry different evidential requirements. Establishing deception requires evidence about what decision-makers knew and represented at the time.
For Qantas, the next material records would connect each affected activity to its workforce before and after the change. A timetable by function, staffing across employers, hours spent on automated and retained tasks, and the outcomes of redeployment would make the programme’s effects easier to assess.
The operational comparison would follow a consistent service: volume, complexity, human hours, errors, delays and full cost per resolved case. Comparable teams or a documented staggered rollout could help separate the tool’s contribution from demand changes. This is a proposed evaluation, not a test performed on the companies’ private operating data.
The public record already supports a clear finding. Qantas is announcing a change in where and by whom work is performed, with AI incorporated into its intended transformation. Tracking that work across providers makes it possible to distinguish the employment decision from the technological result, and to judge both against evidence.
Sources
The references distinguish company announcements, union statements, reporting and research. Workforce figures retain the scope and date of each announcement.
-
Qantas Group fast-tracks technology and AI investment with new Accenture partnership
Passages and scope
Paragraphs on transferred services, 850 roles, consultation and mid-2027. Vanessa Hudson’s statement on AI being built in from the start. Announcement of a multi-year programme. The page displays 10 November 2026 at 11am, without a visible time zone. Official release text and date retrieved from the search index. Direct opening unavailable; the former domain redirects to this new official site. Company communication. Further automation is planned and stages may overlap; saved hours and individual outcomes are not broken down.
-
Qantas to cut 850 corporate jobs as it moves roles overseas in Accenture deal
Passages and scope
Stephanie Chalmers and Liana Walker, opening and account of the staff email: timetable, functions and delivery countries. Announcement and reactions as of 10 November, rather than a tally of completed departures. Accessible HTML article read; publication and update dated 10 November are displayed. The headline summarises the programme as job cuts. Individual outcomes and redeployments remain to be determined.
-
Australian companies face ‘AI washing’ claims after wave of job cuts, offshoring
Passages and scope
Renju Jose and Christine Chen, Sydney dateline 12 November. ACTU/ASU and Uri Gal statements, and Qantas’s response. Report published on 11 November, with a 12 November Sydney dateline. Documents a controversy and the parties’ positions. Direct Reuters page unavailable. Full text read in the identified Reuters republications at Investing.com and KTWB, including authors and dateline. Union allegations and the academic’s interpretation are attributed to them. The report is not a causal audit or a finding of deception. Reuters republication 1. Reuters republication 2.
-
Qantas taps Accenture for major IT and data modernisation
Passages and scope
Ry Crozier, opening and Technical debt section: more than one thousand applications according to Qantas, and data modernisation. Specialist reporting on the announcement and company statements. Full HTML article accessible and read. The technical rationale is attributed to Qantas. No independent audit of its application estate or labour savings.
-
A primer on labour productivity
Passages and scope
Labour productivity: what it is, defining value added, intermediate inputs and hours worked. Statistical methodology: real value added per hour, with contractor services deducted as intermediate inputs. HTML page accessible and read. Displayed publication date: 3 September 2025. Institutional definition, without data on the companies discussed. The 100-employee scenario is a fictional l0g calculation.
-
An important update on our team
Passages and scope
Opening; Why are we making these changes?; Is AI replacing these roles? About 1,600 people, or 10% of the global workforce. Rationale combines funding AI and sales, finances, and changing skill and role needs. HTML page accessible and read, signed by Mike Cannon-Brookes. The CEO’s explanation, rather than an independent measure of automated tasks. The announced total is not limited to Australia.
-
Telstra and Accenture announce global AI joint venture
Passages and scope
Paragraphs on the proposed seven-year joint venture, transfer offers to Australian and Indian Data & AI teams, and consultation. Announcement of a proposed structure, 60% owned by Accenture and 40% by Telstra. HTML page accessible and read. Company communication at the proposal stage. A change of employer does not measure how much work disappears.
-
Telstra and Accenture Launch Silicon Valley Hub to Rapidly Advance Benefits of AI for Telstra Customers and People
Passages and scope
Opening and About the joint venture: the venture is stated to have started in April 2025. Joint release published in May about a venture started in April and its new California hub. HTML page accessible and read. Interested-party source establishing the stated launch, without independent validation of promised performance.
-
WIN: CBA backflips on customer service job cuts, admits they got it wrong
Passages and scope
Opening and CBA admits they got it wrong: 45 roles, July decision and challenge over call volumes. Statement by the union party on the reversal of cuts and working conditions. HTML page accessible and read. Rising calls and allegations of misleading presentation are attributed to the union. Detailed series and a causal effect from the bot are not established.
-
Commonwealth Bank backtracks on AI job cuts, apologises for ‘error’ as call volumes rise
Passages and scope
Stephanie Chalmers, CBA response on its assessment of required roles and options to stay, be redeployed or leave. July decision affecting 45 roles, reversed on 21 August 2025; statements from the bank and union. HTML article accessible and read. The bank’s admission concerns an inadequate assessment of staffing needs. It is not an admission of deception or a record of each employee’s eventual outcome.
-
Generative AI at Work
Passages and scope
Brynjolfsson, Li and Raymond; abstract and introduction; volume 140(2), May 2025, pp. 889–942, DOI 10.1093/qje/qjae044. Staggered rollout to 5,172 agents in one company; a 15% average increase in issues resolved per hour, with heterogeneous effects. Open-access HTML text available through the publisher’s redirect. Abstract, introduction and scope checked. Human work with AI assistance. The study is not designed to estimate aggregate employment or wage effects, or Qantas’s results.
-
Monitoring Early Effects of AI on the Labour Market
Passages and scope
Insight 2026-09; Key takeaways, Motivation, Occupation-based exposure outcomes and Conclusion. The researcher’s own descriptive analysis: complete occupational data through February 2026, partial May data, and potential task exposure. HTML text accessible and read, including disclaimer, methodology and conclusion. PDF not used. Does not represent the views of the RBA or its Boards. Cyclical conditions, other changes and actual adoption limit inference; November announcements fall outside the data period.
-
Our Gen AI Transition - Implications for Work and Skills
Passages and scope
Key findings: augmentation more likely than replacement, early adoption and heterogeneous effects. General institutional summary published on 14 August 2025. HTML summary accessible and read. The downloadable report was not used for this article. The broad assessment does not predict Qantas’s 2026 programme or any individual job’s outcome.
-
Technological change and domestic outsourcing
Passages and scope
Bergeaud, Malgouyres, Mazet-Sonilhac and Signorelli, working paper No. 264, Abstract. Broadband rollout in France during the 2000s and domestic outsourcing of non-core occupations. HTML abstract and metadata accessible and read. The full 75-page paper was not used. A mechanism precedent, not a causal estimate for AI in Australia or for the companies discussed.
-
What is productivity?
Passages and scope
What are the main measures of productivity? and How do we measure outputs and inputs? Educational definitions of labour productivity, multifactor productivity and value added. HTML page accessible and read. Publication date not specified. No data on Qantas or the other companies discussed. Used for definitions only.
Limitations
Analysis based on public documents checked on 13 November 2026. No private contract, HR dataset or complete record of internal and supplier hours was obtained. The Qantas release was read in text returned by the search index; Reuters was read in identified republications. Company and union accounts are attributed to their authors.
The 850 Qantas roles belong to an announced multi-year programme. Individual outcomes and the division of gains between customer and supplier remain incompletely documented. The 100-worker example is hypothetical. The research findings concern their own settings, while the RBA staff note uses observations that precede November’s announcements.
This analysis is not investment advice.
// cite this analysis
l0g, “Qantas: AI layoffs and the work moving offshore”, l0g.fr, published November 13, 2026, updated November 13, 2026, https://l0g.fr/en/analysis/ai-layoffs-offshoring-qantas-ai-washing/
$ cd ../analysis