Analyst(s): Nick Patience
Publication Date: September 23, 2026
DataRobot has expanded its Agent Workforce Platform with new public deployments at Chevron and Aon, deepened its co-engineering with NVIDIA, Dell, and Nebius, and laid out the architecture, governance, and go-to-market strategy behind its bet that multi-cloud, sovereign, and air-gapped enterprise AI will outgrow anything the hyperscalers build. The harder question is whether that bet rests on genuinely hard-to-copy technology, or simply on getting there first.
What is Covered in this Article
- DataRobot’s public collaborations with Chevron on autonomous inspection agents and with Aon on insurance client onboarding, and how both map to its broader Agent Workforce Platform strategy.
- How DataRobot positions itself against copilot-style assistants and single-application agents such as Salesforce Agentforce, and why it is deliberately targeting multi-cloud, sovereign, and air-gapped environments that hyperscalers cover only partially.
- The technical architecture behind the platform, including DataRobot’s NVIDIA co-engineering, its three-part governance model, and a forthcoming split between control plane and data plane for GPU-constrained deployments.
- DataRobot’s identity, cost-management, and governance roadmap, including its live Okta integration, planned Microsoft Entra ID support, and its Center of Excellence services model.
- Futurum’s take on whether DataRobot’s claimed differentiation will hold up as competitors close the gap, supported by ETR data on the platform’s current market traction.
The News: DataRobot has spent the first half of 2026 extending its Agent Workforce Platform into new, publicly disclosed enterprise deployments, most notably its collaboration with Chevron on autonomous inspection agents and its collaboration with Aon on client onboarding and servicing.
Announced in June 2026, the Chevron collaboration applies agentic AI at the edge to support Chevron’s autonomous aerial and terrestrial inspection robots, part of the company’s Facilities and Operations of the Future initiative. Historically, each robotic mission required an operator to verify conditions through a manual permitting process before work could begin; the DataRobot platform instead applies what it calls a Safe Start agentic assessment, built on NVIDIA NIM microservices, to continuously evaluate conditions before and during a mission rather than relying on a one-time check.
The Aon collaboration, announced at the start of the year, uses the DataRobot Agent Workforce Platform’s autonomous, reasoning-based agents across parts of the insurance lifecycle. On the onboarding side, the goal is to consolidate historic documents, policy binders, and policy information to speed new placements and renewals; on the servicing side, the focus is on streamlining certificate generation, invoice processing, and ID card issuance.
Both collaborations build on the Agent Workforce Platform DataRobot introduced in 2025, co-engineered with NVIDIA, which the company has since paired with expanded infrastructure partnerships with Dell and Nebius announced in March 2026, and more recently, a push to unify AI governance across cloud and non-cloud environments alike.
Inside DataRobot’s Enterprise Agentic Strategy
Analyst Take: Enterprise agentic AI is splitting into tiers, and most of the industry’s attention is going to the two most crowded ones: copilots bolted onto existing productivity tools, and agents embedded inside a single application. DataRobot has staked its business on the tier hyperscalers care about least, agents that work across cloud, hybrid, on-premises, sovereign, and air-gapped environments at once, and the company has laid out the architecture, governance model, and roadmap behind that bet. Over the past year, the company has built publicly disclosed use cases on that basis with Chevron, Aon, NVIDIA, Novartis, and the U.S. Special Operations Command.
Who Is DataRobot?
Boston-based DataRobot was founded in 2012 by Jeremy Achin and Tom de Godoy, and built its early business on automated machine learning, helping data teams build and deploy predictive models without deep coding or statistics expertise. The company raised more than a billion dollars over roughly a decade, reaching a valuation of $6.3 billion following a 2021 funding round, and counts Fortune 50 companies among its customers across insurance, energy, life sciences, and financial services. Debanjan Saha, a former Google and Amazon Web Services executive, has led the company as CEO since 2022, steering it through a shift from predictive AI to generative AI and, most recently, to the agentic AI positioning at the center of this piece. DataRobot unveiled its NVIDIA-co-engineered Agent Workforce Platform in July 2025 and became an SAP-endorsed app the following month, both moves that set up the current phase of the company’s strategy.
A Deliberately Narrow Battlefield
DataRobot is not trying to beat everyone at everything, and that restraint makes sense. The company frames the agentic AI market in three levels: copilots, such as ChatGPT and Microsoft Copilot; line-of-business agents embedded within a single application, such as Salesforce Agentforce or Workday; and what it calls agent workforce agents, which span data and processes across an entire enterprise. It concedes hyperscaler-native, single-cloud deployments to AWS, Azure, and Google outright. Its stated focus is multi-cloud, hybrid, on-premises, sovereign, and air-gapped environments, where DataRobot reckons neither the hyperscalers nor data platform vendors such as Databricks have built anything comparably deep yet.
The gap is real, and agent-native platforms built to run consistently across GPU-constrained edge clusters, air-gapped government sites, and hyperscaler clouds without re-architecture remain rare, which is why DataRobot’s five target verticals – manufacturing, insurance and financial services, federal and defense, life sciences, and energy – are places where that gap shows up in named production deployments rather than pilots. It is also, however, a narrow definition of victory, since DataRobot is not claiming to compete with Agentforce or Workday inside their home applications, or with AWS, Azure, and Google inside a single-cloud enterprise. It is claiming the remainder of the market as underserved and building a business on the assumption that it remains underserved long enough to matter.
The Architecture Behind the Pitch
Underneath the positioning, DataRobot’s stack splits into a Build layer and an Operate layer, with three separate governance disciplines underneath: AI governance, covering drift, correctness, and tool-calling accuracy; IT governance, covering permissioning, entitlements, and state auditing; and infrastructure governance, covering GPU cost and utilization. The Chevron deployment, a Safe Start agent that fuses sensor data, weather modeling, and drone imagery to track gas plumes and route inspection drones, runs on a dense NVIDIA stack: Nemotron for sensor interpretation, PhysicsNeMo for physics-informed plume simulation, NIM microservices for the safety-assessment layer, NeMo for guardrails and evaluation, and cuOpt for drone routing.
Identity and access sit on top of that stack rather than inside it, with DataRobot’s agent-identity integration with Okta now live today, while a comparable integration with Microsoft Entra ID remains in proof of concept and is targeted for general availability within the next few months. For a platform selling itself on running everywhere an enterprise’s identity infrastructure already lives, that gap needs to be closed, but it’s also a sign that the agentic identity and governance layer is less mature than the marketing language built on top of it, at least as of mid-2026.
DataRobot is also building a split between control plane and data plane, so a CPU-heavy governance and orchestration layer can run separately from a GPU-heavy inference cluster, such as an NVIDIA SuperPod.
Openness Has Its Limits
DataRobot rests its differentiation on three claims: deep, hard-to-replicate vertical expertise; co-engineering with NVIDIA and Dell; and a platform that avoids locking customers into one model, cloud, or orchestration layer. We think the first two are demonstrable and the third is is currently a bit of a stretch, since a platform whose most technically detailed production deployment, the Chevron Safe Start agent, runs on Nemotron, PhysicsNeMo, NIM, NeMo, and cuOpt is open in the sense that it is not locked to a single hyperscaler, but is nonetheless deeply committed to a single silicon and software partner.
DataRobot’s go-to-market follows the same pattern, pairing an AI Factory motion built on pre-installation arrangements with NVIDIA Cloud Partners such as Dell and Nebius with a business motion built on its status as an SAP-endorsed app and a partnership with Genpact. Both are legitimate distribution strategies, and DataRobot’s early-mover position in each is real, but neither is protected by anything a well-capitalized competitor could not replicate by signing similar deals and hiring similar vertical talent, which is a different kind of advantage than owning data, IP, or switching costs a customer cannot walk away from.
That same pattern shows up in DataRobot’s own services model, where the company says its platform-to-Center-of-Excellence revenue split currently runs close to 80/20, with an internal target of moving toward 75/25. That is a meaningful services dependency for a company selling itself as a software platform, and it means a large share of DataRobot’s current production wins, including the Chevron and Aon deployments covered here, were built with heavy forward-deployed engineering support rather than out-of-the-box product alone.
What Broader Market Data Shows
DataRobot’s own account of its market position is a confident one, describing itself as a category leader for agent workforces in the environments hyperscalers do not cover, but ETR’s broader panel data, which spans a wide cross-section of enterprise IT buyers rather than DataRobot’s own named accounts, tells a more measured story. DataRobot’s overall Net Score (increases minus decreases in spend) was 16, with Pervasion (how widespread a vendor is among sector respondents) at 4%, in ETR’s July 2026 TSIS survey, based on 49 citations, a sample small enough that the figures should be read as directional rather than definitive. In ETR’s March 2026 AI Product Series survey, which tracked plans for AI development and orchestration platforms across 467 respondents, only about 5% said they were currently using DataRobot and planned to continue, while 73% said they had no plans to evaluate the platform.
None of that means the Chevron and Aon deployments are not real, or that the architecture behind them is not sound; both are well documented above. It does mean that outside the specific named accounts DataRobot highlights, broad enterprise awareness and adoption of the platform remain limited as of mid-2026. Whether that changes as more vertical deployments go live, or whether DataRobot’s strategy stays concentrated in a small number of very large, very deep relationships, is one of the things we’ll be tracking over the next few survey cycles.
Put together, DataRobot’s pitch is honest about where it has chosen to compete, and reasonably well supported by what it has actually built. What it has not yet demonstrated is that the moat around that choice runs any deeper than being early, well partnered, and willing to put services people on a problem before the product can solve it alone. That may be enough for the next two or three years, which is, after all, a very long time in enterprise AI, and DataRobot has already been through enough AI cycles to understand and deal with that.
See DataRobot’s full announcement of its Chevron collaboration on autonomous inspection agents on the DataRobot website.
What to Watch
- Whether DataRobot’s Microsoft Entra ID integration reaches general availability on the timeline the company has described, closing the gap with its live Okta integration.
- Whether the platform’s forthcoming control-plane and data-plane split ships and holds up in GPU-constrained edge deployments like Chevron’s.
- Whether DataRobot’s platform-to-services revenue ratio moves toward its stated 75/25 target as its use-case library and self-serve tooling mature.
- Whether hyperscalers, IBM, Palantir, or systems integrators replicate DataRobot’s NVIDIA, Dell, and Nebius partnerships and vertical services model quickly enough to erode its early-mover position.
- Whether ETR’s broader panel data on DataRobot’s adoption and Net Score shift meaningfully as more deployments like Chevron and Aon move from pilot to production.
Disclosure: Futurum is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of Futurum as a whole.
Read the full Futurum Group Disclosure.
Other Insights from Futurum
Agentic AI: The Leading Vendors Winning the Enterprise in 2026
Microsoft Agent 365 Turns Shadow AI Into a Governed Asset Class
AWS Pushes the Agent Stack at What’s Next 2026
Author Information
Nick Patience is VP and Practice Lead for AI Platforms at The Futurum Group. Nick is a thought leader on AI development, deployment, and adoption - an area he has researched for 25 years. Before Futurum, Nick was a Managing Analyst with S&P Global Market Intelligence, responsible for 451 Research’s coverage of Data, AI, Analytics, Information Security, and Risk. Nick became part of S&P Global through its 2019 acquisition of 451 Research, a pioneering analyst firm that Nick co-founded in 1999. He is a sought-after speaker and advisor, known for his expertise in the drivers of AI adoption, industry use cases, and the infrastructure behind its development and deployment. Nick also spent three years as a product marketing lead at Recommind (now part of OpenText), a machine learning-driven eDiscovery software company. Nick is based in London.
