How this list was built
We looked for published evidence of agentic delivery among the firms our market analysis covers, and checked each piece against its primary source: the firm’s own results or announcements, the analyst firm’s own report, or a news report where neither exists. Eight showed enough to place. The order follows a stated hierarchy of evidence rather than a score:
- An independent analyst assessment of agentic work specifically.
- Disclosed revenue or contract value tied to agentic transformation.
- Disclosed advanced AI revenue where agentic is combined with generative.
- Agents in production, on the firm’s own operations or a named client.
- A named agentic platform shipping.
- Capability to govern agents rather than build them.
Where firms share a rung, the next rung decides. Four firms here share the first: ISG’s 2025 Provider Lens report on agentic AI services names each of them a Leader in agentic AI development and deployment. Firms whose only agentic mention was incidental were left out rather than padded in.
The eight firms
1. Capgemini
Assessed in agentic AI, and a contract tied to agentic work. ISG names Capgemini a Leader in agentic AI development and deployment services. Generative and agentic AI accounted for more than 10% of Capgemini’s fourth-quarter 2025 bookings, up from around 5% earlier that year, and the firm signed an intelligent operations contract worth more than €600 million tied to agentic AI transformation. In March 2025 it launched agentic AI solutions for industries built with NVIDIA.
Why here. Of the four firms an analyst has assessed on agentic work, Capgemini is the one with a disclosed contract value attached to agentic transformation, the next rung in the order. The bookings share shows the work repeating rather than a single win.
2. Accenture
Assessed in agentic AI, and the largest disclosed advanced AI revenue. ISG names Accenture a Leader in agentic AI development and deployment services. Accenture reports advanced AI revenue, covering generative, agentic and physical AI together, which tripled year over year to $2.7 billion in fiscal 2025, alongside generative AI bookings of $5.9 billion. It is one of the few firms that reports this work separately.
Why here. The largest disclosed number in the group, but it combines agentic with generative and physical AI, so the agentic share is not visible. That is why it sits below a contract tied to agentic work specifically.
3. IBM Consulting
Assessed in agentic AI, and a disclosed generative AI book of business. ISG names IBM a Leader in agentic AI development and deployment services. IBM’s generative AI book of business, across software and consulting since its start, stood at more than $12.5 billion at the end of 2025. Forrester also named IBM a Leader in The Forrester Wave: AI Governance Solutions, Q3 2025, for watsonx.governance, which IBM says helps enterprises manage AI risk, compliance and trust at scale.
Why here. An analyst assessment of its agentic work, a disclosed figure that combines agentic with generative AI, and the governance capability buyers most often find missing. The figure is bookings rather than revenue, which places it below Accenture.
4. Cognizant
Assessed in agentic AI. ISG names Cognizant a Leader in agentic AI development and deployment services. In April 2026 Cognizant set up a dedicated Gemini Enterprise practice with Google Cloud.
Why here. The same analyst assessment as the three firms above it, without a disclosed figure tied to its agentic work.
5. PwC
Agents deployed across its own operations. PwC had deployed more than 250 AI agents across the firm before it launched its agent operating system to clients in March 2025. It is a Leader in The Forrester Wave: AI Consulting Services, Q2 2026, became OpenAI’s first reseller of ChatGPT Enterprise in 2024, and is a launch partner for Anthropic’s industry offerings.
Why here. Running 250 agents on its own operations is direct operational proof. It is self-reported, and PwC’s analyst recognition is for AI consulting generally rather than agentic work specifically.
6. Deloitte
A named agentic platform, with a named client rolling it out. Deloitte launched Zora AI in March 2025, a suite of ready-to-deploy agents designed to perceive, reason and act across business functions, built with NVIDIA. At launch it named a client: Hewlett Packard Enterprise’s finance team is rolling out Zora AI for Finance.
Why here. A shipping platform with a named client deployment. Deloitte discloses no revenue tied to agentic work, which places it below the firms above.
7. Tredence
An agentic advisor deployed at a named client. A leading wellness company has deployed an AI-powered wellness advisor that offers personalized guidance and supports product decisions, in partnership with Tredence. The company’s agentic offerings include RAPID, for agentic process automation, and Milky Way, an analytics platform with specialized AI agents.
Why here. A named client deployment is real evidence. Tredence publishes this site, so its placement here rests only on the same kind of published, linked evidence as every other entry, and carries no editorial preference.
8. Palantir
The most completely specified agent platform here. Palantir’s AIP documentation covers all four layers of a working agent: a planner that chooses and calls tools, the tools themselves, memory and retrieved context, and an evaluation harness, AIP Evals, for testing agents before they run. Palantir also originated the forward deployed engineering model, putting its own engineers inside client teams.
Why here. Platform completeness is real evidence, and on that measure this is the strongest entry. It sits last because the evidence gathered here is of the platform rather than of agents in production at a named client, and there is no disclosed agentic revenue.
Who is not on this list, and why
Model providers. Anthropic, OpenAI and the other labs supply the models these agents are built on. That is a different layer from delivery: a lab sells you the planner’s brain, not the tool layer, the memory or the evaluation harness around it. They are not ranked here.
Strategy firms without a build practice. Bain & Company works closely with OpenAI, but its AI work concentrates on strategy and private equity value creation rather than large-scale technical build. On that evidence it does not belong in a delivery ranking, so it is left out rather than listed and caveated.
What this list cannot tell you
Whether any of these firms handles a failed agent run well, what permissions their agents hold in production, how large their evaluation suites are, or what the rollback path looks like for an action that has already reached a third party. None of the four can be answered from published material. Ask them on the first call.
Corrected on 25 September 2026. An earlier version placed Cognizant first as the only firm here independently assessed on agentic work; ISG’s report names four of these firms as Leaders, so the order now follows the next rung of evidence. It also dated Cognizant’s Gemini Enterprise practice to 2025 (it was set up in April 2026), said Deloitte had no named deployment, and credited Tredence with generative and agentic AI deployments at four named clients that no published source confirms. Every claim now links to its source.