engineering
Building Upwork's AI products (Uma, agents, AI integrations)
2.9% to 9.7% of the quarter’s revenue
Incremental total: counts at zero at the point and in full at the high end, with no traced baseline.
our inferencedisclosure: described· motive: exploratory· before LLMs: expanded
Research and development excluding restructuring charges was , down on lower bonus and stock compensation and more capitalized development; the AI build is the ledger's assumed share of it, . The 10-Q now names third-party providers of AI technologies among Upwork's vendors, without a cost. The counterparty is mixed: the company's own engineering staff, and the third-party model and hosting providers behind Uma.
Evidence: 5 quotes, 9 figures, 2 confounds, 8 from before coverage
What Upwork spends building its AI products and features, mostly engineering staff in research and development, plus capitalized platform development and the model and hosting bill behind Uma, which no source splits. Research and development also carries non-AI product work.
Why this motive
Carried from Q1 2026: building AI capabilities and agent infrastructure (claim c16) with revenue still attached only to the AI-work category.
Before LLMs: expanded
Product and AI development sat in research and development at the anchor, in FY2024 and on a non-GAAP basis in Q1 2024, already building on large language models for Uma, with the Headroom and Objective AI acquisitions adding AI and machine learning staff. The AI part of the line was never split, so no pre-AI dollar baseline can be traced. The size is the whole of an activity that existed before.
“One highlight of our spring 2024 update is the launch of Uma, which stands for Upwork's Mindful AI. We're developing Uma on top of industry-leading large language models and fine-tuning it with trillions of tokens of highly relevant Upwork platform data across a range of work interactions.”
“We also have several partnerships that enable us to integrate generative artificial intelligence tools into our work marketplace aimed at improving customer experience and productivity.”
“At the end of last year, we acquired AI startup Headroom and deepened our bench of technical talent and leadership in this area.”
“The sequential increase in R&D also reflects the growth of our AI ML talent bench through the acquisition in Q4 of Headroom and other investments.”
“While we remain committed to ongoing innovation to further enhance our platform, including building new features with a focus on generative artificial intelligence, we expect total research and development expense to decrease in 2025.”
“In 2024, we acquired Objective AI, Inc., which we refer to as Objective AI, an AI-native search-as-a-service platform, with the aim of further enhancing our core search and matching capabilities and enabling richer insights into client needs and talent capabilities.”
“Our highly profitable business model, with gross margins over 75%, as well as the investments we're making in operational improvements like engineering, productivity, and back-office automation, mean that we can make clear and concrete commitments to growing our margins and free cash flow for the foreseeable future.”
“In October 2024, we announced the Restructuring Plan, intended to continue our profitable trajectory, increase efficiency, and accelerate innovation for our customers.”
Figures
- Research and development · 2026-CQ2
- Research and development · 2025-CQ2
- Restructuring charges in research and development · 2026-CQ2
- Research and development excluding restructuring charges · 2026-CQ2
- Research and development excluding restructuring charges, change year over year · 2026-CQ2
- Restructuring and related costs · 2026-CQ2
- Workforce reduced under the 2026 restructuring · as-of 2026-06-30
- Internal-use software and platform development capitalized · 2026-CQ2
- Internal-use software and platform development capitalized · 2025-CQ2
What else could explain it
- transformation program: The 2026 restructuring cut the workforce by by 2026-06-30 and cost in the quarter; the 10-Q gives efficiency and the evolving nature of work as reasons and names no AI cause, so none of it is an AI saving.
- line composition: Research and development also carries non-AI product work and the Lifted platform; capitalized platform development, against , sits outside the line.
Quotes
“While lower-complexity work continues to shift toward automation, we are increasingly seeing what is emerging in its place: growing demand for high-value AI talent, more complex projects, and new categories of work across SMB and Enterprise. We are on track with our enterprise strategy, staying disciplined in how we execute, and continuing to build the AI capabilities that will make Upwork essential infrastructure as work becomes increasingly human-and-agent driven.”
“We incorporate AI and machine learning technologies across our platforms, offerings, and internal operations and are making further investments in expanding our AI capabilities. In addition, GSV from AI-related work performed on the Upwork Marketplace has increased as demand for AI talent has grown in recent periods.”
“To grow and maintain our business, offerings, and features, we need to continue to establish and maintain relationships with third parties, such as staffing providers, software and technology vendors, including third-party providers of AI technologies, and payment processing and disbursement providers.”
“In May 2026, the Company initiated a restructuring plan, referred to as the 2026 Restructuring, intended to build a more efficient operating model and position the Company for profitable growth as the nature of work evolves. The 2026 Restructuring included various cost-reduction measures, including workforce reductions, reductions in operating expenses, as well as decreases in capital expenditures. As part of the 2026 Restructuring, the Company reduced its total workforce by approximately 20% as of June 30, 2026.”
“Non-GAAP operating expenses reflected the early execution of our restructuring actions we announced in May, as the annualized $70 million of OpEx reduction we initiated is expected to generate approximately $40 million in realized savings in fiscal year 2026, allowing us to continuously fund our growth pillars in SMB, Enterprise, and AI while expanding operating leverage.”
By quarter
- Q1 2026described · our inference · exploratory · $6.5mn to $22mn
- Q2 2026described · our inference · exploratory · $5.6mn to $19mn