Enterprise AI 23 min read

We analyzed 150 enterprise AI agent deployments. Here’s what actually works in 2026

Nouman Ashraf September 22, 2026

Updated October 2026. The 150-deployment dataset below is unchanged (evidence cutoff September 2, 2026). This update adds a comparison of the main enterprise AI agent platforms and their pricing models, a dated log of enterprise AI agents news through October 2026, a governance checklist and an FAQ. It also absorbs our earlier coverage of AI agents in the enterprise, which now redirects here.

We analyzed 150 enterprise AI agents deployments. Bounded autonomy is winning.

Original research from The Brief Script

150 qualifying deployments • 119 organizations • 392 public records reviewed • Evidence cutoff: September 2, 2026

Research note. This is an evidence-screened deployment sample, not a random sample of the enterprise market. Percentages in this article describe the frozen 150-case research dataset, not all enterprise AI deployments.

The short version

Finding	What the dataset shows
A3 is the center of gravity	57 of 150 deployments (38.0%) are bounded autonomous-action systems. A3+A4 together account for 49.3% of the frozen sample.
A3 has the strongest evidence	91.2% of A3 deployments report a quantified outcome; 94.7% meet the Moderate-or-Strong evidence threshold; average evidence score is 76.6/100.
The first ROI is operational	71 cases report time/cycle effects and 50 report productivity/throughput. Only 22 explicitly disclose cost/financial impact and 10 disclose revenue/commercial impact.
Software engineering is unusually measurable	All 13 software/engineering deployments report quantified outcomes, and 69.2% are A3 or A4.
Payments are technically real but economically under-measured	The dataset contains 19 payments/treasury deployments, yet only 15.8% report a quantified outcome and just 5.3% are A3 or A4.
Human control remains explicit	118 of 150 deployments describe approval, exception handling, governance/accountability, or active human review.
Source mix passes the independence test	64 of 150 core cases (42.7%) are grounded primarily in adopter-owned, regulatory, or independent evidence.

What counted as an enterprise AI agent

The study used a behavioral definition. Marketing labels did not determine inclusion; observable tool use and action did.

Level	Operational definition	Typical behavior	Study treatment
A0	Assistant only	Generates, summarizes, searches or retrieves; no independent meaningful tool/action.	Excluded from core
A1	Tool-using read agent	Independently chooses tools or queries data; no business-system write.	Core eligible
A2	Human-approved action agent	Prepares or invokes an action, but consequential execution needs explicit approval.	Core eligible
A3	Bounded autonomous action	Executes predefined actions automatically inside clear rules; escalates exceptions.	Core eligible
A4	End-to-end autonomous workflow	Plans and executes a multi-step workflow without per-action approval; humans review exceptions/outcomes.	Core eligible
A5	Multi-agent orchestration	Specialized agents delegate and coordinate execution across tools/systems.	Core eligible

Figure 1. Autonomy distribution across the frozen 150-case dataset.

Enterprise AI agents are no longer a future-tense category. They are quoting freight, routing service tickets, booking medical appointments, reconciling supplier statements, provisioning software, writing code, checking fraud signals, and—under controlled permissions—initiating transactions.

But 150 real deployments tell a more disciplined story than the market hype. The strongest production pattern is not unrestricted autonomy. It is bounded execution: agents handle the routine path automatically while people move to approvals, exceptions, policy, and accountability.

The Brief Script reviewed 392 public records and froze a core dataset of 150 qualifying deployments from 119 organizations. To count, a system had to do more than generate text or retrieve information. It needed evidence of independent tool use or action inside a real enterprise workflow.

That produces a useful contrast with broader adoption surveys. As covered in our enterprise agentic AI deployment-gap analysis, experimentation is widespread but production scaling remains much harder. This study looks below those organization-level adoption numbers and asks a narrower question: what do the production systems that actually cleared an evidence threshold look like? Read: enterprise agentic AI deployment gap

The production story is bounded execution, not “full autonomy”

The enterprise-agent market is moving past the question of whether a model can call a tool. The harder question is what companies are willing to let that system do repeatedly, at production scale, with real business consequences.

In the frozen 150-case dataset, the largest autonomy tier is A3: bounded autonomous action. There are 57 A3 deployments, compared with 31 A1 read agents, 45 A2 human-approved action agents, and 17 A4 end-to-end autonomous workflows. No A5 case cleared the final core under the frozen evidence rules. A3 and A4 together account for 49.3% of the sample.

That distribution matters because it cuts against the most dramatic version of the agent narrative. The strongest deployments are not systems with unlimited freedom. They are systems that can act without a person clicking every step, but only inside an operating envelope that defines permissions, thresholds, approved tools, and exception states.

C.H. Robinson is a useful example. Its transactional freight-quoting agent evaluates proprietary data and issues customer-specific quotes automatically. Quote coverage rose from roughly 60–65% to 100%, while average response time fell from 17–20 minutes to about 32 seconds. People remain responsible for strategic and exception handling, but the routine path no longer waits for a human operator.

At the higher end, Glide’s Slipstream engineering pipeline can build, validate, deploy, monitor, diagnose, open fixes, and automatically roll back a degrading release. Glide reports median merge-to-production time falling from 23 hours to 14 minutes. This is genuine end-to-end autonomy, but it exists in a workflow with version control, tests, observable system state, and reversible actions—the conditions that make autonomy easier to trust.

Figure 2. Share of deployments with quantified outcomes by autonomy level.

A3 is the evidence sweet spot

The strongest result in the dataset is not simply that A3 is common. It is that A3 has the best evidence profile.

A3 deployments average 76.6/100 on the study’s evidence score. 91.2% report a quantified outcome, and 94.7% meet the Moderate-or-Strong evidence threshold. A4 systems remain compelling, but they are rarer and slightly less consistently measured. A1 systems can also be well documented, but they are fundamentally less consequential because they do not write to or change business systems.

A2 is the weak spot. Its 45 cases average 63.1/100; only 53.3% disclose a quantified outcome, and 55.6% meet the Moderate-or-Strong threshold. That does not mean A2 is a bad architecture. It reflects the composition of the tier: payments, regulated decisions, commerce actions, claims, and other workflows where an agent can prepare or initiate a real action but a person still authorizes the consequential step.

In other words, technical action often arrives before mature business measurement. This is especially visible in payments, where the public evidence frequently proves that an agent can transact under permissions and authentication before it proves lower cost, faster processing, more revenue, or fewer losses.

Figure 3. Evidence strength by autonomy level.

Different workflows are at very different stages of maturity

The agent market is not one market. Workflow design determines what can be automated, what can be measured, and how much autonomy an organization can safely expose.

Customer service and IT/network operations are the largest groups in the core, with 21 deployments each. Payments/treasury contributes 19, risk/finance/insurance 17, sales/marketing/commerce 14, software/engineering 13, industrial/field operations 10, healthcare operations 9, supply chain/logistics 8, procurement/finance operations 8, and HR/workforce 7.

Software engineering stands out because the environment is instrumented. All 13 software/engineering cases report a quantified outcome, and 69.2% are A3 or A4. Delivery Hero says Herogen writes, tests, iterates, and self-reviews code before human review; at only 18% rollout, it was already responsible for 9% of the company’s code-change requests, with output estimated as equivalent to 130 senior engineers.

Customer service is also further into production than the chatbot framing suggests. 90.5% of the 21 customer-service cases report quantified outcomes and 71.4% are A3 or A4. Qualifying systems do more than answer FAQs: they route work, update records, complete service requests, schedule appointments, or execute other bounded actions.

Healthcare is smaller but operationally interesting. At Hospital for Special Surgery, the patient-access agent reasons over patient needs, insurance, physician availability, and care pathways, then books appointments directly into scheduling systems and sends confirmations. Staff can intervene at any point. That is a very different system from a medical chatbot. The rule-based bots hospitals already run for prior authorization, claims and billing are covered in our guide to robotic process automation in healthcare.

Figure 4. Largest workflow groups in the final core.

Payments show that agents can transact—but ROI is still thin

Payments and treasury are the clearest example of why technical capability and economic evidence should be separated.

The final dataset contains 19 payments/treasury deployments. Yet only 15.8% report a quantified outcome, and just 5.3% are A3 or A4. Most are A2: the agent prepares or initiates a transaction while a person approves the final consequential step.

This is still significant. It means banks, payment networks, and commerce systems are exposing real transaction rails to agents under identity, authorization, limits, and audit controls. The technology has crossed an important threshold from simulation to controlled execution.

But a buyer should not confuse ‘the agent can pay’ with ‘the agent has proven ROI.’ In this workflow family, the public evidence is currently much stronger on capability and control than on measurable economics.

The first measurable return is speed, not dollars

Across the full core, the dominant reported outcome is time. Seventy-one deployments report a time or cycle-speed effect, and 50 report productivity or throughput. Twenty-five report quality or risk. Only 22 explicitly report cost or financial impact, while 10 report revenue or commercial effects. These categories overlap and are not meant to sum to 150.

The pattern makes sense. Cycle time is immediate and relatively easy to observe. Revenue attribution is noisy. Cost savings depend on whether freed capacity is removed, redeployed, absorbed by growth, or used to improve service quality. The operational benefit therefore appears before the accounting benefit.

Bank of Singapore / OCBC makes the pattern concrete. Its Source of Wealth Assistant independently initiates and coordinates KYC preparation tasks, reducing work that could take up to 10 days to roughly one hour. Deutsche Bank reports that a structured four-to-five-document third-party-risk review can fall from around three hours to about 30 minutes using a three-agent workflow, while the final risk decision remains human.

For enterprise teams, this suggests a better measurement sequence: first track cycle time, automation coverage, exception rate, first-pass quality, resolution time, and cases per employee. Then translate those operational effects into cost, capacity, loss avoidance, or revenue once the process is stable enough to attribute value cleanly.

Figure 5. Outcome categories reported across the sample.

Human oversight is being redesigned, not removed

The most consistent operating model in the dataset is not ‘AI replaces the worker.’ It is ‘AI executes the routine path while people own approvals, exceptions, policy, and accountability.’

118 of 150 deployments explicitly describe approval, exception handling, governance/accountability, or active human review. Even under a stricter subset that excludes generic review/use language, 106 still contain explicit approval, exception, or governance controls.

Allianz’s Project Nemo shows what this looks like in practice. Seven agents perform coverage verification, weather checks, fraud checks, payout calculation, and audit summarization, but a human claims professional makes every final payout decision. Allianz reports an 80% reduction in claim processing and settlement time.

This is why ‘human in the loop’ is too vague to be a useful governance description. There are several distinct loops: approval before action, exception handling after automation, post-hoc audit, policy design, and legal accountability. A mature agent deployment should be explicit about which loop applies to which action. Our review of what 12 agent platforms document separates those loops into concrete controls such as pre-action approval, stopping a run and audit evidence.

Figure 6. Explicit human control remains visible in most deployments.

“Agentic” marketing still overstates what many systems do

Some of the most informative records in the 392-case research universe are the systems that did not qualify.

A product could have excellent self-service or cost metrics and still fail the behavioral agent test if it only retrieved information or generated text. Other systems remained in proof of concept, described future capabilities rather than observed production behavior, lacked a named adopter, or bundled many agents into an umbrella program without a bounded workflow.

This is also why the study kept failure evidence outside the successful core. Meta’s Project OT, for example, is useful precisely because it shows what can happen when a broad organizational agent rollout outruns workflow clarity, technical stability, or employee acceptance. The point is not that agentic AI failed as a category. It is that autonomy without a well-defined operating model can fail for organizational reasons even when the underlying technology is capable.

The Brief Script’s earlier enterprise AI failure analysis reached a similar conclusion across the broader AI market: technical success, production success, and economic success are three different thresholds. Agents make that distinction even more important because the system can now take action, not merely produce an answer. Read: enterprise AI failure analysis

Vendor stories are more metric-heavy, but not clearly better evidence

The final core contains 64 cases grounded primarily in adopter-owned, regulatory, or independent evidence and 86 supported primarily by vendor or partner evidence. That mix was a deliberate research constraint: no single source ecosystem should define the conclusions.

Vendor-supported cases are more likely to publish a number—80.2% versus 71.9% for the non-vendor group. But their average evidence scores are almost identical: 70.2 for vendor-supported cases and 71.7 for non-vendor/adopter/independent cases. Moderate-or-Strong evidence shares are effectively the same at 77.9% and 78.1%.

The lesson is not ‘vendor evidence is bad.’ Vendor case studies are often the best source for workflow mechanics and before/after metrics. Filings, adopter pressrooms, annual reports, and independent reporting are often better at establishing deployment ownership, limitations, governance, or whether a result is observed rather than projected. The strongest research uses both and keeps those roles separate.

What enterprise teams should do differently

The dataset points to a practical deployment playbook that is less dramatic than the ‘autonomous enterprise’ narrative and more consistent with what is working in production.

Start with a workflow that has a measurable cycle time and a high volume of repetitive decisions. Give the agent enough tool access to complete the routine path, not every conceivable path. Encode hard permissions and exception states. Log every consequential action. Measure automation coverage, time, quality, exception rate, and reversibility before trying to calculate headline ROI.

Then expand autonomy only after the exception distribution is understood. The strongest A3 deployments are not impressive because they eliminate humans. They are impressive because they make the routine path cheap and fast while making the exception path explicit.

For buyers, five questions are more useful than asking whether a product is ‘agentic’: What does it read? Which tools can it choose? What can it write or change? Which actions require approval? What happens when confidence is low or the case falls outside policy? If those answers are missing, the autonomy claim is not yet auditable.

This research also reframes the broader agent adoption debate covered in The Brief Script’s enterprise agentic AI deployment-gap analysis. The hard part is no longer making an agent demo work. It is designing the workflow, permissions, observability, measurement, and human accountability that allow the agent to keep working after the demo ends.

Enterprise AI agent platforms compared

The 150 deployments in this study ran on a mix of vendor platforms, cloud agent runtimes and in-house builds. The vendor market has sorted into roughly four layers: coding agents, knowledge-work agents inside productivity suites, business-process agents embedded in systems of record, and a governance and orchestration layer that inventories, permissions and monitors agents across all of them. Most of the large platforms now try to cover more than one layer.

The table below summarizes the main AI agent platforms enterprise buyers are evaluating, with each vendor’s published pricing model as of late October 2026. It describes what each product is built to do, not how well it performs; none of the 150 case outcomes above should be read as a ranking of these vendors.

Platform	What it does	Pricing model (official pages, Oct 2026)	Typical fit
Salesforce Agentforce	Customer- and employee-facing agents grounded in CRM and Data 360 records; Agentforce Builder, Coworker and multi-agent orchestration.	Consumption or per user. Flex Credits at $500 per 100,000 credits (a standard action uses 20 credits, about $0.10); $2 per conversation for customer-facing agents; Agentforce add-ons from $125 per user per month; Agentforce User License $5 per user per month plus Flex Credits. Pricing	Sales, service and field workflows already run on Salesforce
Microsoft Copilot Studio (with Agent 365)	Low-code agent building across Microsoft 365, Power Platform and Azure; Agent 365 is the control plane that inventories and governs agents, including ones built on other clouds.	Internal agents are included with Microsoft 365 Copilot ($30 per user per month). Standalone Copilot Studio: 25,000 Copilot Credits for $200 per pack per month, pay-as-you-go, or pre-purchase. Agent 365: $15 per user per month standalone. Copilot Studio pricing / Agent 365	Organizations standardized on Microsoft 365 and Entra
ServiceNow AI Agents	Agents inside IT, HR, customer service and other workflows on the ServiceNow AI Platform, with AI Agent Studio and AI Control Tower for governance.	Since April 9, 2026, AI is included in every package, sold in Foundation, Advanced and Prime tiers; building net-new agents requires Prime. Usage is metered in “assists”; list prices are not published. Announcement / Tiers	Service-management-heavy enterprises
Google Gemini Enterprise and Gemini Enterprise Agent Platform	Gemini Enterprise (which absorbed Agentspace in October 2025) is the employee-facing agent and search layer; the Agent Platform (formerly Vertex AI) is the developer runtime.	Gemini Enterprise Business edition from $21 per seat per month; Standard and Plus through sales; a pay-as-you-go edition was announced in August 2026 for select customers. Agent Platform is billed on consumption. Business edition / Billing options	Google Workspace shops and multi-model buyers
AWS Bedrock AgentCore	Framework-agnostic infrastructure to run agents in production: runtime, gateway, memory, identity, policy, observability and evaluations.	Pure consumption, no upfront commitment. Examples: runtime from $0.0895 per vCPU-hour and $0.00945 per GB-hour; Gateway $0.005 per 1,000 invocations; Policy $0.000025 per authorization request. Pricing	Engineering teams building their own agents on AWS
OpenAI Frontier (plus ChatGPT workspace agents and Agents SDK)	Frontier, launched February 5, 2026, is a platform to build, deploy and manage “AI coworkers” with shared business context, an execution environment, evaluation and per-agent identity and permissions.	Frontier pricing is not published (sales-led, with forward-deployed engineers). Workspace agents in ChatGPT moved to credit-based pricing on May 6, 2026; the Agents SDK runs at API rates. Frontier / Workspace agents	Companies standardizing on OpenAI models across functions
UiPath agentic automation	Agents combined with RPA robots, API workflows and human steps, orchestrated by Maestro.	Consumption units: on Unified Pricing, a standard-model LLM call costs 0.2 Platform Units (1 Agent Unit on the Flex plan); bring-your-own-model agents are charged 0.2 Platform Units per run. Licensing	Enterprises with an existing RPA estate

Three patterns stand out. First, pricing has moved away from pure per-seat licensing toward credits, actions, conversations and compute, which makes cost forecasting a workflow-design exercise rather than a headcount exercise. Second, every major vendor now sells identity, permissions, approval steps and audit logs as part of the agent platform, which is the same control surface the A3 deployments in this study relied on. Third, the governance layer is becoming cross-vendor: Microsoft’s Agent 365 is previewing registry sync with AWS Bedrock and Google Cloud, and Salesforce and Google both pitch their agents as usable inside other vendors’ surfaces.

For a longer view of how agent pricing is breaking the seat model, see our analysis of AI agents and per-seat SaaS pricing. For the hidden integration and governance costs that sit outside license fees, see enterprise AI deployment cost.

Enterprise AI agents news: what changed in 2026

The deployment evidence above was frozen on September 2, 2026. The market kept moving. These are the developments that matter most for enterprise buyers, each confirmed on the vendor’s or analyst’s own page.

February 5 — OpenAI launches Frontier. An enterprise platform for building, deploying and managing agents with shared business context, evaluation and explicit permissions. HP, Intuit, Oracle, State Farm, Thermo Fisher and Uber were named as early adopters. OpenAI
April 9 — ServiceNow folds AI into every package. AI, data connectivity, governance and AI Control Tower became part of all products rather than a separate purchase, with new Foundation, Advanced and Prime tiers. ServiceNow
April 15 — Gartner puts agentic AI at the Peak of Inflated Expectations. Only 17% of organizations have deployed AI agents, but more than 60% expect to within two years, according to Gartner’s 2026 CIO and Technology Executive Survey. Gartner
April 28 — Gartner quantifies agent sprawl. It predicts an average global Fortune 500 enterprise will have over 150,000 agents in use by 2028, up from fewer than 15 in 2025, and says only 13% of organizations think they have the right agent governance. Gartner
May 1 — Microsoft Agent 365 reaches general availability at $15 per user per month, with a preview of registry sync that discovers agents running on AWS Bedrock and Google Cloud. Microsoft
May 26 — Gartner warns against one-size-fits-all governance. It predicts that by 2027, 40% of enterprises will demote or decommission autonomous agents because of governance gaps found only after production incidents, and proposes four autonomy levels: observe, advise, act with approval, and act autonomously. Gartner
June 3 — OpenAI winds down Agent Builder and Evals. Both stop being available on November 30, 2026; OpenAI points builders to the Agents SDK or workspace agents in ChatGPT. Teams that built on the visual builder need a migration plan. OpenAI
August 26 — Google adds pay-as-you-go and spend caps. A consumption edition of Gemini Enterprise, project-level hard spend caps, and Flexible Savings Plans offering 10% off for one-year and 20% off for three-year commitments. Google Cloud
September (Dreamforce) — Salesforce ships Agentforce Coworker and previews long-horizon agents. Coworker is generally available; long-horizon agents that pursue goals over days or weeks are in pilot with GA planned for November 2026; Salesforce also introduced Koa, its own CRM reasoning model. Salesforce
September 29 — AWS previews Bedrock Managed Agents, powered by OpenAI. Each agent gets its own IAM role, supports human approval before consequential actions and logs activity to CloudTrail; no additional charge during preview. AWS
October 8 — Google introduces the Gemini agent. A single work agent that can spin up sub-agents and “coworker agents” with their own identities, governed through an Agent Gateway and real-time spend caps. Google Cloud

Read against the 150-deployment data, the direction is consistent. Vendors are not selling unlimited autonomy. They are productizing exactly the controls that separated the strongest A3 deployments from the rest: a distinct identity per agent, scoped permissions, approval gates before consequential actions, audit logs and cost limits. Gartner’s four autonomy levels map closely onto this study’s A1 to A4 scale, and its warning about uniform governance matches what the dataset shows about A2 payments workflows, where controls arrived before measurable economics.

For the policy side of the same shift, see AI policy news 2026 and our AI governance gap analysis.

How to govern enterprise AI agents by autonomy level

The practical lesson from both the deployment data and the 2026 analyst warnings is that governance should scale with what an agent can do, not be applied uniformly. A read-only A1 agent and an A4 agent that executes an end-to-end workflow need different controls, even if they run on the same model.

Governance layer	What it controls	Evidence to keep	Why it matters
Agent inventory	Which agents exist and who owns them	Owner, purpose, vendor, model, system access	Prevents invisible agent sprawl
Permission scope	What systems and data an agent can reach	Role, credential, API, data boundary, approval level	Limits damage from wrong or unsafe actions
Action logging	What the agent did and when	Tool call, file access, output, timestamp	Makes incidents traceable and auditable
Human approval	Which actions require review	Approval rule, reviewer, override reason, final action	Keeps people accountable for high-impact decisions
Cost monitoring	Credits, tokens, tool use and workflow cost	Usage by agent, team, task and outcome	Stops pilots becoming invisible spend under consumption pricing
Failure response	How agents are paused, demoted or shut down	Incident record, rollback, model or policy change	Turns governance into an operational control

A simple rule of thumb from the dataset: A1 agents need scoped data access and logging; A2 agents add approval workflows that stay meaningful under volume; A3 agents need hard permission limits, defined exception states and reversibility; A4 agents need continuous monitoring, rollback and a named owner for outcomes. Our agentic AI enterprise analysis covers how organizations are moving pilots into production, and enterprise AI security covers the security controls in more depth.

Ten representative production deployments

These examples are illustrative rather than a ranking. The final core contains 150 bounded deployment units.

Organization	Workflow	Autonomy	Published outcome
C.H. Robinson source	Transactional freight quoting	A3	Quote coverage ~60–65%→100%; response time 17–20 min→~32 sec
Allianz source	Project Nemo claims workflow	A2	Claim processing/settlement time -80%; final payout remains human-approved
Bank of Singapore / OCBC source	Source of Wealth Assistant	A2	Preparation up to 10 days→~1 hour
Deutsche Bank source	Third-party-risk evidence review	A2	Structured review ~3 hours→~30 minutes
Delivery Hero source	Herogen software-delivery agent	A3	9% of company code-change requests at 18% rollout
Glide source	Slipstream engineering pipeline	A4	Median merge-to-production 23 hours→14 minutes
Hospital for Special Surgery source	Patient scheduling and triage	A3	Direct appointment booking and confirmations; 24/7 production
Travelers source	AI Claim Assistant	A3	~85–90% claim-filing completion after scaling
ZKH Group source	AI Smart Workbench	A3	520,000+ autonomous system operations in 2025
Avalara source	Software provisioning/deprovisioning	A3	78% automated; ~1,500 hours saved monthly

Methodology and limitations

The Brief Script researched 392 public records and selected 150 qualifying deployment-level cases. The evidence window was frozen at September 2, 2026. A case entered the core only when public evidence supported a named organization, a bounded workflow, real deployment or production use, and behavior above simple generation or retrieval.

Autonomy was coded separately from architecture. A multi-agent system did not automatically receive a higher autonomy level; the coding depended on what the system could actually do. Evidence was scored on a 100-point protocol covering quantified outcomes, baseline disclosure, measurement period, deployment scale, action detail, human oversight, production maturity, business materiality, adopter evidence, and disclosed limitations.

The final core contains 150 deployments from 119 organizations. Multiple deployments from the same organization were allowed only when the trigger, action chain, and outcome were meaningfully distinct. The evidence score measures disclosure and verifiability—not whether a deployment was objectively better than another deployment.

This is not a random sample and should not be interpreted as market prevalence. Public-source research overrepresents organizations willing to disclose deployments and outcomes. Some metrics originate in company or vendor claims and cannot be independently audited from public material. Categories such as time, cost, quality, and revenue can overlap.

Research parameter	Frozen rule
Research universe	392 public records
Final core	150 qualifying deployment-level cases
Organizations	119
Evidence cutoff	September 2, 2026
Non-vendor/adopter/independent core	64 cases (42.7%)
Evidence scoring	100-point disclosure/verifiability protocol
Core eligibility	A1–A5 behavioral threshold; A0 excluded
Interpretation	Sample findings, not market-prevalence estimates

The takeaway

Enterprise AI agents are already doing real work in production. The important shift is not from humans to machines; it is from human execution to machine execution inside explicit boundaries.

The strongest evidence in 2026 sits at A3: agents that can act automatically on the routine path, escalate exceptions, and leave a clear control surface for people. That is where autonomy and evidence quality align most consistently in this dataset.

The next test is economic. Enterprises have become much better at proving that agents save minutes, increase throughput, and expand automation coverage. The market will become more mature when those operational gains are routinely connected to cost per task, revenue, loss avoidance, and return on invested capital.

Until then, the right question is not whether a company has AI agents. It is whether those agents can act inside a measurable workflow, under auditable permissions, with evidence that the work is actually better.

FAQ

What are enterprise AI agents?

Enterprise AI agents are AI systems that use tools, data and workflow steps to complete business tasks with some degree of independence, rather than only generating text. In this study, a system counted only if it showed evidence of independent tool use or action inside a real workflow. The most common production pattern was bounded autonomy: agents execute routine steps automatically and escalate exceptions to people.

What is the difference between AI agents and chatbots?

A chatbot responds to prompts. An AI agent can plan steps, choose tools, read and write to business systems and act toward a goal. That is why agents need stronger permissions, approval rules and logging than chatbots, and why this study excluded assistant-only systems (A0) from its core sample.

Which AI agent platforms do enterprises use?

The main enterprise options in 2026 include Salesforce Agentforce, Microsoft Copilot Studio with Agent 365, ServiceNow AI Agents, Google Gemini Enterprise and its Agent Platform, AWS Bedrock AgentCore, OpenAI Frontier and UiPath. The right choice usually follows where the workflow and its data already live, such as CRM, Microsoft 365, the service desk or a cloud provider. Our AI agent platforms comparison covers pricing, models and governance for each.

How much do enterprise AI agent platforms cost?

Pricing varies by model. As of October 2026, Salesforce charges $500 per 100,000 Flex Credits or $2 per conversation, Microsoft sells 25,000 Copilot Studio credits for $200 a month, Google’s Gemini Enterprise Business edition starts at $21 per seat per month, and AWS AgentCore bills purely on consumption. ServiceNow and OpenAI Frontier do not publish list prices. License fees are usually a minority of total cost once integration, data work and governance are included.

How should you choose AI agents for enterprise workflows?

Start from the workflow, not the vendor. The strongest deployments in this dataset had high volumes of repetitive decisions, a measurable cycle time, clear exception states and reversible actions. Ask what the agent reads, which tools it can choose, what it can change, which actions need approval, and what happens when a case falls outside policy.

Which enterprise workflows are best suited to AI agents?

In this dataset, software engineering and customer service were the most mature: all 13 software cases and 90.5% of the 21 customer-service cases reported quantified outcomes. Payments and treasury were technically live but thinly measured, with only 15.8% of 19 cases reporting a quantified outcome.

What is agent sprawl and why does governance matter?

Agent sprawl is the growth of agents across teams without central visibility, ownership or controls. Gartner expects an average global Fortune 500 company to run over 150,000 agents by 2028, and predicts 40% of enterprises will demote or decommission autonomous agents by 2027 after governance failures. An inventory, scoped permissions, action logs and a shutdown process are the minimum controls.

Are AI agents ready for full enterprise autonomy?

Mostly not, and the production data does not suggest companies want it. Only 17 of 150 deployments were end-to-end autonomous (A4), and no multi-agent orchestration case cleared the evidence bar. The evidence is strongest for bounded A3 systems that act on the routine path and hand exceptions to people.