Anthropic and OpenAI Are Reshaping Global IT Services: Platformization, Agents, and Services-as-Software Become Core Variables
AI summary card
Anthropic and OpenAI Are Reshaping Global IT Services: Platformization, Agents, and Services-as-Software Become Core Variables
Bernstein believes that enterprise AI has moved from pilots to production-level deployment, with system integrators reconfiguring delivery models around Claude, GPT, and Frontier—facing short-term pressures but poised for operating leverage over the medium to long term.
- Anthropic emphasizes enterprise-first, safety-first, and multi-cloud neutrality, deeply integrating via the Claude Partner Network with Accenture, Deloitte, Infosys, Cognizant, and others—ideal for regulated, high-risk workflows.
- OpenAI focuses on scale, speed, and platform control, leveraging Frontier, Agent Builder, the Microsoft ecosystem, and Amazon expansion to become the default AI layer in enterprise workflows.
- A common path among global system integrators involves first using Claude and GPT to compress internal delivery costs, then packaging these capabilities into agentized, outcome-driven, or managed services.
- Accenture stands out as the most comprehensive orchestrator of multiple models; Infosys exemplifies a delivery-oriented Topaz platformization case; Capgemini represents an OpenAI Frontier–driven European transformation integrator.
- In terms of investment implications, the report reaffirms ratings and price targets for covered stocks, underscoring the trade-off between near-term margin headwinds and medium-to-long-term Services-as-Software operating leverage.
Report interpretation
Overview
This report examines how Anthropic and OpenAI’s enterprise AI strategies intersect with the global IT services industry, analyzing how system integrators are leveraging both to advance AI from pilot stages to production-level deployment. The central conclusion is that enterprise AI is no longer merely an ancillary tool—it is becoming a foundational engine embedded within proprietary platforms, delivery frameworks, and operational models.
Core views
Anthropic and OpenAI are not interchangeable model providers but distinct enterprise AI platforms. Anthropic prioritizes security, governance, multi-cloud compatibility, and vendor channels—suited for regulated, complex, high-risk workflows—while OpenAI emphasizes scale, brand strength, model iteration, and Frontier control layers—ideal for broad productivity gains, customer experience enhancement, and large-scale agent deployment. Whether system integrators can translate these partnerships into repeatable, billable, margin-improving Services-as-Software offerings will determine industry winners and losers.
Analysis framework
The report employs three analytical threads—ecosystem comparison, system integrator organizational forms, and business model evolution—first contrasting Anthropic and OpenAI strategic differences, then observing how representative integrators like Accenture, Infosys, and Capgemini combine models, agent platforms, and delivery capabilities, and finally assessing short-, medium-, and long-term impacts on profitability and revenue structures.
Methodology notes
Anthropic emphasizes endurance, while OpenAI stresses dominance.
Anthropic builds deeper yet narrower enterprise embeds through security, controllability, multi-cloud support, and SI partners; OpenAI achieves broader but more demanding platform coverage via Frontier, Microsoft/Amazon distribution, and ISV ecosystems.
Transitioning from human-delivered services to AI-agent and IP-driven standardized offerings.
System integrators first use AI to compress internal development, testing, documentation, and support costs, then package agent capabilities into customer-facing managed services or outcome-based solutions, gradually shifting revenue streams from labor hours toward platform and results.
Short-term margin pressure, medium-term portfolio improvement, long-term operating leverage.
Years 0–2 primarily absorb training, CoE, and platform investments; years 2–5 see high-value transformations and rising regulatory demands; beyond year 5, if Services-as-Software scales, leading integrators may experience revenue growth and margin leaps.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Anthropic / ClaudeOne of the core pillars of enterprise AI platforms and SI partner ecosystems
- Strengths
- Safety-first, Constitutional AI, multi-cloud neutrality, well-suited for regulated industries and high-risk workflows, easily integrated into complex processes via SI partners.
- Weaknesses
- Lower terminal brand visibility, reliance on service provider channels, potentially slower expansion compared to OpenAI.
- Comparison
- Compared to OpenAI, Anthropic offers deeper, more stable integration, better suited for governance and complex agent workflows.
- Risks
- Partner dependency, slower brand-building progress, intensifying competition between clouds and models.
- OpenAI / FrontierAn enterprise agent orchestration platform and horizontal AI workflow layer
- Strengths
- Strong brand, rapid model iteration, extensive ecosystem, robust Microsoft distribution, beginning expansion into cloud players like Amazon; Frontier holds promise as an enterprise agent control layer.
- Weaknesses
- High execution complexity, greater dependence on key partners and hyperscale cloud vendors.
- Comparison
- Compared to Anthropic, OpenAI covers broader territory and moves faster, but faces scrutiny regarding stability, governance, and lock-in risks.
- Risks
- Frequent model changes conflicting with enterprise stability needs, concerns about vendor lock-in, and complexities in implementing horizontal platforms.
- AccentureA benchmark system integrator specializing in multi-model orchestration
- Strengths
- Deeply engaged in both Anthropic and OpenAI ecosystems, with Anthropic Business Group, Frontier practices, industry solutions, and innovation centers—best representing a multi-model AI partnership operating model.
- Weaknesses
- High organizational and delivery complexity, short-term training and platform investments weighing down margins.
- Comparison
- More advisory and orchestration-focused than Infosys, with broader dual-model exposure than Capgemini.
- Risks
- Complex execution, risk of self-replacement in traditional projects, pressure from personnel restructuring.
- Infosys / TopazA delivery-oriented agentized enterprise platform
- Strengths
- Uses Claude as the core Topaz engine while incorporating OpenAI, forming a model-independent control layer—ideal for modernization, compliance, testing, claims processing, and managed services.
- Weaknesses
- Less consulting influence than Accenture, AI platform revenue scaling still requires time.
- Comparison
- More focused on agent platforms as products than Capgemini, with greater emphasis on delivery scale and reuse than Accenture.
- Risks
- High investment intensity, escalating competition, potential delays in reaching profit inflection points.
- CapgeminiAn OpenAI Frontier-driven enterprise transformation integrator
- Strengths
- Strong European roots, deep industry expertise, ability to reshape operational models, and Generative Enterprise framework—founding member of the Frontier Alliance.
- Weaknesses
- No large-scale Claude deployment disclosed yet, weaker Anthropic leverage.
- Comparison
- More OpenAI-centric than Accenture, with stronger C-suite transformation and operational model reshaping than Infosys.
- Risks
- If critical agent demand shifts toward Claude, Capgemini may need to formalize further Anthropic collaboration.
Key data
- Claude Partner Network$100mAnthropic launched its partner network in March, targeting SI partners, distributors, and solution providers.
- Deloitte Claude DeploymentApproximately 470,000 employeesReport indicates this is one of Claude's largest single corporate deployments, accompanied by a Claude Center of Excellence.
- Cognizant Claude DeploymentApproximately 350,000 employeesUsed internally for productivity improvements and externally for modernized, agentized AI services.
- Accenture Claude Training ScaleApproximately 30,000 professionalsAccenture established the Anthropic Business Group while simultaneously participating in the OpenAI Frontier Alliance.
- OpenAI Frontier AlliancesAccenture, Capgemini, McKinsey, BCGDeployed to integrate OpenAI Frontier agent platforms into large enterprise transformation projects.
- Capgemini Target Price€208, rated OutperformReport reaffirmed Capgemini's rating and target price, highlighting its exposure to OpenAI Frontier.
- Microsoft Target Price$641, rated OutperformReport reiterated Microsoft's rating and target price, with the 2026E EPS adjusted to $16.84.
- Short-Term Financial ImpactMargin pressure during years 0–2Training, CoE, platform development, and joint client investments dilute margins, while AI revenues remain relatively small.
- Medium-Term Financial ImpactOperating leverage begins to emerge during years 2–5If agentized services standardize, revenue portfolios shift toward high-value transformations and regulatory-compliant work.
- Long-Term Financial ImpactPotential growth and margin leaps after year 5Provided integrators proactively manage the transition of traditional ADM/BPS roles to AI replacements.
Impact & implications
For investors, this report positions enterprise AI as a structural reevaluation variable for the IT services industry: In the short term, focus should not solely be on AI revenue contributions but rather on training expenditures, platformization progress, and margin headwinds; in the medium to long term, prioritize identifying companies capable of turning model partnerships into repeatable agent services, outcome-based billing, and managed operations. Accenture boasts the most comprehensive multi-model exposure, Infosys offers clearer delivery-platformization pathways, and Capgemini provides purer OpenAI Frontier transformation exposure—but all require bolstering related Anthropic capabilities.
Risks
- Short-term AI training, platform construction, CoE efforts, and joint client investments will suppress margins.
- Traditional ADM/BPS roles and entry-level delivery positions may be replaced by AI agents, posing organizational and workforce transition risks.
- If system integrators fail to productize AI capabilities, they may only earn low-value implementation fees.
- OpenAI ecosystems face challenges such as key partner concentration, vendor lock-in, and enterprise-level stability requirements.
- Anthropic ecosystems grapple with channel dependency, lower brand visibility, and slower expansion rates.
- OpenAI-leaning companies like Capgemini, lacking deep Claude integration, may find themselves at a disadvantage in scenarios demanding high governance and security standards.
What to watch
- Whether Frontier agents can convert into repeatable, recurring, or outcome-linked revenues over the next 12–24 months.
- Whether system integrators disclose AI revenue shares, margin impacts, and Services-as-Software progress.
- Whether Capgemini establishes a dedicated Claude CoE, launches large-scale deployments, or formalizes Anthropic collaborations.
- Whether Infosys Topaz can evolve from a strategic platform into an auditable, repeatable revenue engine.
- Whether Accenture can transform its multi-model advantages into scalable industry solutions without being bogged down by execution complexity.
- How Anthropic and OpenAI distribute their presence across cloud channels like Microsoft, Amazon, AWS, Google Cloud, etc.