エンタープライズAI・2026年4月

エンタープライズAIリーダーが苦労して学んだ5つの実運用における真実

稼働中のAIデプロイメントを管理する200社以上のエンタープライズリーダーが、コスト、制御、ガバナンスについて発見したこと — ベンダーのスライドでは語られなかったこと。
200以上
企業回答者
18以上
参加業界
95%
本番環境でのエージェント運用
81%
今年のAI支出は増加傾向
ご意見をお聞かせください
LinkedIn logo on a black background with white lettering and dot symbol.
Black and white crisscross pattern with intersecting diagonal lines on a white background.
04
05
真実
01
03
05
01
真実
推論コストは問題のごく一部に過ぎない
03
真実
誰も準備ができていないうちに、ツールが爆発的に増えた。
05
真実
ツールの拡大 = 投資とリスク
02
真実
95%がエージェントを実行。その半数は、エージェントがどこへ行くのか追跡できない。
04
真実
見えないものは、統制できない。
200以上の組織 · すべてAIを本番環境で運用中

協力ソートリーダー

Smiling business professional in a blue suit and tie with short brown hair in a circle frame.
linkedin-icon
ギャレット・マロリー
ML シニアマネージャー
Roadie
Two blue location pin icons on a white background, indicating geographic markers or map pointers.
3週間前のカンファレンスで、Googleの発表者が講演を締めくくった言葉が、ずっと私の心に残っています。「モデルは今が一番出来が悪い状態だ」と。変化の度合いを考えると、これは非常に示唆に富む言葉です。たとえここで止まったとしても、すでに破壊的な影響を与えています。私は知る限り最も懐疑的な機械学習の専門家の一人です。機械学習は非常に便利なツールですが、たいていは最適なツールではありません。しかし、新しい世界が広がっていることに気づくような、目を見張る瞬間ってありますよね？私は他のテクノロジーよりも、このテクノロジーでそうした瞬間をより頻繁に経験しています。
Two blue location pin icons on a white background, indicating geographic markers or map pointers.
Smiling man with short dark hair wearing light blue collared shirt in circular profile photo frame.
linkedin-icon
ダニエル・シャー
VP エンジニアリング
Via Transportation
Two blue location pin icons on a white background, indicating geographic markers or map pointers.
I think everyone that has ever used AI models has run into the issue of cost at one point or another. I've heard lots of stories about cost spiking, or about a chain of micro-services in a specific back-end where all of a sudden one change in one part led to a large amount of data running through the system and all of a sudden AI costs have gone through the roof.
Two blue location pin icons on a white background, indicating geographic markers or map pointers.
Woman with long dark hair and glasses smiling wearing blue blouse sitting by window with greenery.
linkedin-icon
Meenal Iyer
SVP Data
SurveyMonkey
Two blue location pin icons on a white background, indicating geographic markers or map pointers.
When the cloud came in the first time, everyone was hit with the same thing — we're not going to have CapEx, there's no more on-prem, everything is going to move up to the cloud. And then they started looking at the cloud costs and the OpEx just increased like crazy. It's going to be quite a learning curve and it's going to be painful, especially in terms of cost — because people think AI coming in is going to reduce costs. No. There are costs to running AI itself which people haven't accounted for.
Two blue location pin icons on a white background, indicating geographic markers or map pointers.
Featured Voices

Voices from the Field

Practitioners who are leading the charge of GenAI across the ecosystem
Two blue location pin icons on a white background, indicating geographic markers or map pointers.
The biggest friction is inconsistency: different teams running different models with no shared evaluation standards, which means outputs that can't be compared and integrations that keep breaking on updates. We've had to retroactively build governance around 10+ models instead of designing for it upfront, and that debt shows up as latency in every delivery cycle.
Dmytro Zhorov
VP Technology Lead
EPAM Systems
White lowercase letter e on a blue gradient background with a pixelated effect applied.
Two blue location pin icons on a white background, indicating geographic markers or map pointers.
The biggest friction from model sprawl is the loss of standardization. When multiple teams manage models independently, governance becomes fragmented, observability is inconsistent, and production reliability suffers. This creates hidden costs in the form of duplicated infrastructure, delayed incident resolution, and slower time-to-value for new AI initiatives. It also increases business risk.
Deepak MK
VP Data Science
Examroom AI
Examroom.ai logo featuring green and blue triangle design on white background with text next to it.
Two blue location pin icons on a white background, indicating geographic markers or map pointers.
On what’s going to be different in 2026 - Agentic workflows and agentic solutions to problems like model sprawl and code reviews on generated code. And on biggest security risk in 2026 - Usage level statistics and eval sets.
Holt Calder
Director Data Analytics
Greenhouse Software
Green lowercase letter g logo with dot above, stylized font design element on white background.
Two blue location pin icons on a white background, indicating geographic markers or map pointers.
There are two major issues that we face because of the larger number of models. First, for our own internal models, we face issues in maintaining versions and governance because there are large amounts of training data, parameter information and other artifacts that are associated with a model, and we usually store them manually on s3. The other issue is making sure that the external models that we use through APIs do not expire. This becomes a problem because many teams use external models for their own tasks, and hence, it is difficult to find a coordinated way to upgrade them when a new version is released and the previous version is deprecated. The vendors mail us, and then, we pass on the information through meetings to update and test the new versions.
Apurva Bhargava
ML/AI Engineer
Informed
Informed logo with magnifying glass icon in green and black text on white background.
Respondent Disclosure & Privacy Notice
The individuals and organizations featured in this report participated voluntarily as independent thought leaders and practitioners of enterprise GenAI. All responses were collected through a structured research survey designed specifically to capture practitioner experiences and perspectives — no proprietary, confidential, or commercially sensitive company information was solicited or included. Company logos and affiliations are displayed solely to indicate the respondent's professional context at the time of the survey and do not imply organizational endorsement of any findings, products, or services referenced herein. Respondents who did not provide explicit consent to be identified have been anonymized. Any resemblance of anonymized profiles to specific individuals is coincidental.

Executive Summary

Enterprise AI has crossed the production threshold. Budgets are growing across the board, and model access has never been easier. But beneath the momentum, a more complicated picture is forming.

The tools enterprises chose for speed are now their largest sources of cost opacity, security exposure, and governance debt. These five truths — drawn from 200+ real production deployments — describe what that looks like in practice.
95%
Running AI agents in production — not pilots, not experiments but live.
41%
See inference costs only post-usage — no real-time visibility
76%
Lack fully unified logging across all models and agents
51%
Cannot confirm all tool endpoints are authenticated and secure
The 5 Production Truths
Blue checkmark icon on a white background, symbolizing verification and approval in a digital interface.
Truth 01
Your inference bill is the smallest part
Blue checkmark icon on a white background, symbolizing verification and approval in a digital interface.
Truth 02
95% run agents. Half can't Fully trace them.
Blue checkmark icon on a white background, symbolizing verification and approval in a digital interface.
Truth 03
The tool surface exploded
Blue checkmark icon on a white background, symbolizing verification and approval in a digital interface.
Truth 04
What you can't see, you can't govern
Blue checkmark icon on a white background, symbolizing verification and approval in a digital interface.
Truth 05
The 2026 Paradox: Investment vs Risk prioritization
Headline Stat
41%
see costs post-usage
83%
see token amplification
78%
have 6+ tool endpoints
56%
have no full control layer
#1
Tool expansion = investment & risk
What it Means
Hidden costs — orchestration, tool calls, retries — compound invisibly
Agentic loops multiply costs and risk in ways point monitoring misses
MCP/tool proliferation outpaced security and access control frameworks
Fragmented logging creates compliance gaps even in regulated industries
Organizations are accelerating into their own exposure with full knowledge
Who Responded

200+ Enterprise AI Leaders,
‍All with Agents Running in Production

Every respondent confirmed live AI agent deployments. No POC-only teams. No aspirational respondents. These are the people managing real production risk, right now.
Three stacked orange cubes on pale background, representing blocks or storage containers in a simple graphic.
Industry Distribution
Industry Distribution
34%
Financial Services
22%
Healthcare & Life Sciences
12%
Retail & Consumer
11%
Manufacturing & Industrial
10%
Other
11%
Three interconnected blue cube blocks stacked pyramidally on a white background forming a cohesive structure.
Job Function Distribution
VP/SVP/EVP of Engineering or AI
38%
Director / Sr. Director of AI
29%
CTO / CIO / CDO / CAIO
18%
Head of AI / ML Platform
15%
Three interconnected purple cubes stacked pyramidally on a white background forming a geometric shape structure.
Company Revenue Distribution
$1B+ (Large Enterprise)
44%
$250M – $1B (Mid-Market)
31%
$50M – $250M (Growth)
18%
Under $50M (Scale-up)
7%
Orange icon of two arms with three circles above representing people or support team collaboration.
Named Contributors
Every featured quote is attributed to a real person, verified title, and company. No anonymous surveys dressed up as research.
Orange icon of a person with an orange shield beside them on a white background.
Production-Only Respondents
All participants confirmed active AI agent deployments. Not planning, not piloting — live in production with real cost and risk exposure.
Orange arrow pointing up to a cluster of orange circles on a white background growing upward.
Stats Derived from Raw Data
Every percentage is calculated directly from survey responses. Nothing extrapolated, inferred, or rounded to make a better headline.
Three people icons in orange color with corners framing them on a white background together.
No Vendor Bias
Respondents were not customers. Questions were written to surface problems, not validate solutions. Negative findings were not softened.
Production Truth #01

Your inference bill is the smallest part  of the problem

Blue checkmark icon on a white background, symbolizing verification and approval in a digital interface.
Inference cost is a minority share
The inference bill, despite being the most visible metric, accounts for only 15–20% of total AI production cost.
Blue checkmark icon on a white background, symbolizing verification and approval in a digital interface.
Majority of costs are hidden in system complexity
The remaining ~80% of costs arise from less visible layers such as orchestration overhead, tool call chains, embedding generation, retry loops, and engineering effort for debugging non-deterministic behavior.
Blue checkmark icon on a white background, symbolizing verification and approval in a digital interface.
Cost visibility is delayed and reactive
A significant 41% of organizations lack real-time cost monitoring, becoming aware of expenses only after usage has already occurred, without proactive alerts or budget controls.
Blue checkmark icon on a white background, symbolizing verification and approval in a digital interface.
Misconfigurations amplify unseen costs dramatically
Errors like an agent executing a 400-call chain instead of 4 can go unnoticed until billing cycles, leading to unexpected and substantial cost overruns.
Blue checkmark icon on a white background, symbolizing verification and approval in a digital interface.
Advanced visibility practices define cost control maturity
Enterprises with better cost management have transitioned from post-hoc billing analysis to token-level tracing across the entire AI stack, enabling granular and proactive oversight.
Three interconnected blue cube blocks stacked pyramidally on a white background forming a cohesive structure.
Spend Trajectory in 2026
Growing aggressively (>30%)
42%
Growing moderately (10–30%)
39%
Stable or consolidating
19%
Three interconnected blue cube blocks stacked pyramidally on a white background forming a cohesive structure.
AI Cost Visibility
Donut chart divided into three distinct purple sections on a white background for data visualization purposes.
After usage (post-hoc billing)
41%
Partial real-time visibility
35%
Full real-time token-level tracking
24%
80%
Hidden Costs
The remaining 80% hides in orchestration overhead, tool call chains, embedding generation, retries, and debugging non-deterministic behavior.
Organizations report budget increases in 2026
81%
AI Spend Growing
61%
Spend Controls in Place
No spend caps or rate limits of any kind across AI deployments
Inference Bill
Misleads
01
Truth
Misconfiguration Risks
41%
Lack Real-Time
Visibility
Of organizations only see AI costs after usage, with no real-time alerts or budget controls in place.
FROM THE FIELD
Two blue quotation mark icons on white background for citation and dialogue representation purposes.
What surprised us most about GenAI costs in production: inference can explode 10x beyond training costs, and context bloat multiplies rapidly — spiking latency in the process. Model sprawl has created significant orchestration overhead — coordinating multiple models drives up compute costs and error rates, while unmanaged sprawl introduces friction in data flow that compounds over time.
Bijit Ghosh
Engineering
Wells Fargo
WF logo in white serif font on dark background.
Two blue quotation mark icons on white background for citation and dialogue representation purposes.
In pilots, cost looks linear and predictable—API usage, a few engineers, modest infrastructure. In production, cost becomes non‑linear and systemic. Inference spend grows faster than expected as usage scales, but that’s only part of it. The bigger surprise is how much cost sits around the model: orchestration, monitoring, security controls, compliance reviews, integration with legacy systems, and ongoing human oversight.
Mukta Maheshwari
AVP Engineering
State Street
State Street logo with three horizontal lines and company name in black text on white background.
Production Truth #02

95% Run Agents. Half Can't Fully Trace  Where They Go.

Blue checkmark symbol on a white background, within a blue hexagonal shape outline.
Agentic AI is now the default
Agentic AI has moved from a concept to a default pattern: almost every enterprise in this survey runs autonomous agents in production that coordinate tasks, call tools, and make decisions without human confirmation at each step.
Blue checkmark symbol on a white background, within a blue hexagonal shape outline.
Agentic workflows create long call chains
When an agent runs, it doesn’t make one model call; it creates a chain of actions — prompt, tool call, retrieval, re‑prompt, decision — and each link can amplify token usage and cost.
Blue checkmark symbol on a white background, within a blue hexagonal shape outline.
Lack of real‑time chain inspection
Most organizations cannot inspect this chain in real time. Only about half have the tracing infrastructure to see agent decisions step‑by‑step as they happen.
Blue checkmark symbol on a white background, within a blue hexagonal shape outline.
Adoption driven by real productivity gains
The adoption curve has been steep and is mostly fueled by genuine productivity gains, not just experimentation.
Blue checkmark symbol on a white background, within a blue hexagonal shape outline.
Not just a cost issue, also compliance
This lack of visibility is not just a cost problem; for regulated industries, it becomes a compliance and auditability problem with serious consequences.
95%
Agents in
Production
Confirmed running AI agents in live production environments
83%
Token Amplification in Agentic Workflows
Observe significant token amplification, agent chains consuming 3–15×expected tokens
Three blue stacked cubes icon on white background, representing blocks or construction.
AI Cost Visibility
Blue sections in a donut chart on a white background divided into four parts.
Full step-by-step agent tracing
46%
Partial tracing (output-level only)
49%
No structured tracing in place
5%
83%
 See Token
Amplification
83% of respondents observe token amplification, where a task that should cost ~1,000 tokens ends up consuming 8,000, 15,000, or more through the chain.
An agent doesn't make one model call; it creates a chain of actions- prompt, tool call, retrieval, re prompt, decision amplifying token usage and cost.
Agents Create Long Call Chains
61%
No Spend Caps
61% report no spend caps on agentic workflows, meaning a runaway agent loop has no automated stop condition.
No Real-Time
Chain Inspection
02
Truth
Agentic AI Is
Now the Default
Compliance & Audit Risk
This lack of visibility is not just a cost problem; for regulated industries, it's a compliance and auditability problem with serious consequences.
FROM THE FIELD
Two blue location pin icons on a white background, indicating geographic markers or map pointers.
Early model sprawl created duplication and limited observability. We mitigated this by building a centralized LLM library to abstract provider access and standardize telemetry. While this significantly improved governance and cost control, it introduced some friction in terms of slower experimentation and dependency on the platform team for new model integrations.
Karan Kakwani
Machine Learning Engineering
Apollo.io
Black starburst shape with eight pointed rays on light gray background, forming a circular pattern.
Two blue location pin icons on a white background, indicating geographic markers or map pointers.
The real  shock was that inference costs are only about 15–20% of total spend; the other 80% is hiding in data pipelines, evaluation loops, drift remediation, and integration work that no pilot ever surfaces. We essentially funded a visible line item and inherited four invisible ones.
Dmytro Zhorov
VP Technology Lead
EPAM Systems
Large white lowercase letter e on a gradient dark gray background with subtle light effect.
Two blue location pin icons on a white background, indicating geographic markers or map pointers.
As we move from single-turn interactions to multi-step autonomous pipelines, the blast radius of a bad decision grows fast. We don't yet have the equivalent of a circuit breaker — something that catches when an agent is about to do something it shouldn't, mid-flight. Traditional security tooling wasn't built for this and we're somewhat improvising.
Anonymous
ML Engineering
Daugherty Enterprises
Daugherty Enterprises Foundation logo featuring heart shaped design with fork and spoon incorporated elements.
Production Truth #03

The tool surface exploded before ‍anyone was ready

Orange checkmark symbol on a white background within an abstract shape.
Tool surface exploded first
The Model Context Protocol and tool‑calling APIs gave AI agents full reach into databases, APIs, internal services, and external systems, but they did not ship with any governance framework for when every team starts wiring their own endpoints.
Orange checkmark symbol on a white background within an abstract shape.
Explosion of active tool endpoints
78% of enterprises now have six or more tool endpoints in active use, turning each endpoint into an independent attack surface, billing surface, and data access path.
Orange checkmark symbol on a white background within an abstract shape.
Endpoints as independent attack/billing surfaces
These tool endpoints are not just “neat integrations”; when called by an agent they can access customer data, trigger downstream processes, or incur third‑party API costs, and this multiplies across dozens of teams, hundreds of agents, and thousands of daily calls.
Orange checkmark symbol on a white background within an abstract shape.
Big gap in tool inventory vs. security review
There is a significant, largely unmeasured gap between a company’s actual tool inventory and the security review of that inventory.
Orange checkmark symbol on a white background within an abstract shape.
MCP governance gap
The MCP governance gap means the tooling for agent‑to‑system connections scaled far faster than the organizational capacity to review, approve, and audit them.
78%
Active Tool Endpoints per Organizations
Have 6 or more active tool/MCP endpoints across AI systems
51%
Authentication Confidence across tool Endpoints
Cannot confirm all tool endpoints are properly authenticated
Three interconnected orange cubes stacked in a pyramid formation on a pale yellow background.
Active Tool Endpoints per Organizations
Donut chart divided into four sections of varying yellow and orange shades on white background.
6–15 endpoints
44%
16+ endpoints
34%
2–5 endpoints
22%
78%
6+ Endpoints
78% of enterprises now have six or more tool endpoints in active use, each acting as an independent attack surface, billing surface, and data access path.
There is a significant, largely unmeasured gap between a company's tool inventory and its security review of
that inventory.
Gap:
Inventory vs.
Security Review
51%
Unconfirmed
51% cannot confirm all tool endpoints are properly authenticated and access-controlled; for the other 49%, "confirmed" often means "we think so," not "verified systematically."
03
Truth
Endpoints as
Attack & Billing
Surfaces
MCP
Governance Gap
The MCP governance gap means the tooling for agent-to-system connections scaled far faster than the organization's ability to review, approve, and audit them.
FROM THE FIELD
Two orange quotation marks on a white background for citation and dialogue indication purposes.
There are quite a few operational frictions caused by model sprawl in the organization — different models and tools producing different or even contradictory results, regulatory and compliance risks, production issues when moving from a prototype to a mass production model, as well as an efficiency hit.
Anonymous
Principal Engineer
Intel
Intel logo on white background with black lettering and registered trademark symbol.
Two orange quotation marks on a white background for citation and dialogue indication purposes.
It also complicated cost management and observability. API usage and GPU workloads grew in parallel, and without unified tracking, it became harder to attribute spend or measure ROI at a use case level. On the governance side, varying licensing terms and data handling policies across models required additional review cycles.
Omkar Basarikatti
ML Engineering
Acceldata
Acceldata logo with dark blue text on white background displays company brand name clearly.
Production Truth #04

What you can't see,  you can't govern

Purple verified checkmark icon on white background, symbolizing approval and confirmation in digital interfaces and design.
76% lack unified logging
76% of organizations do not have fully unified logging across all their AI models and agent workflows. Different models log to different systems, agent decisions are not correlated with inference logs, and there is no single view of what the full AI stack did on a given day — a nightmare for CISOs and compliance officers behind the shiny demos.
Purple verified checkmark icon on white background, symbolizing approval and confirmation in digital interfaces and design.
No single view of the AI stack
Because logs are siloed, the full chain of AI behavior (from prompt to agent decision to model call) cannot be reconstructed in one place, making incident investigation, cost attribution, and compliance reporting extremely fragile.
Purple verified checkmark icon on white background, symbolizing approval and confirmation in digital interfaces and design.
No universal enforcement points
Without that central layer, there is no single place to enforce prompting policies, apply content filters universally, route traffic between models by cost or capability, or systematically record decisions for audit. Every model is accessed directly, often with the same API key and minimal logging.
Purple verified checkmark icon on white background, symbolizing approval and confirmation in digital interfaces and design.
Direct model access is a risk
Direct, unmediated access to models means policies are implemented ad hoc, security boundaries are inconsistent, and the same vulnerable pattern repeats across teams and tools.
Purple verified checkmark icon on white background, symbolizing approval and confirmation in digital interfaces and design.
Audits already flagging this as a finding
Several enterprise leaders in this survey said compliance audits are already surfacing the lack of unified AI logging and control as a material finding, forcing attention and resource allocation.
76%
Lack fully unified logging across all AI models and workflows
56%
Have no centralized control layer for AI access and policy enforcement
Three interconnected purple cubes stacked pyramidally on a white background forming a geometric shape structure.
Logging Coverage Breakdown
Donut chart with three purple sections on a white background showing data distribution.
Fully unified (all models + agents)
24%
Partial (some models logged)
61%
Minimal or no structured logging
15%
Three interconnected purple cubes stacked pyramidally on a white background forming a geometric shape structure.
Top Governance Concerns Cited
Data leakage / PII exposure in prompts
1st
Lack of audit trail for model decisions
2nd
No policy enforcement between users & models
3rd
76%
Lack Unified
Logging
76% of organizations do not have fully unified logging across AI models and agent workflows, so each system logs separately and behavior cannot be reconstructed as a whole.
Without unified logs, there is no single view of what the full AI stack did on a given day, making incident analysis & compliance reporting highly fragile.
No Single View
of the AI Stack
56%
Lack Central
Control Layer
56% of enterprises have no centralized control layer between users or agents and the models they call, so there is no common policy enforcement point.
No Universal
Enforcement
Points
04
Truth
Direct Model
Access Is Risky
Compliance
Consequences
In regulated industries like finance, healthcare, and insurance, this gap has direct compliance consequences and weakens the foundation beneath shiny AI demos.
FROM THE FIELD
Two purple quotation marks on a white background, symbolizing quote or citation indicators.
Model sprawl increases evaluation complexity, since performance comparisons across providers and model versions are not standardized. Governance and access control reviews take longer as the surface area expands.
Ameer Azam
Data Scientist
Pixis.ai
Pixis logo with stylized lettering and design on a white background.
Two purple quotation marks on a white background, symbolizing quote or citation indicators.
Model sprawl  triggers "FinOps" and security chaos, making it nearly impossible to track per-feature ROI or enforce consistent safety guardrails across the organization.
Lucas Hendrich
CTO
Forte  Group
Teal colored logo with stylized letter G and arrow on white background.
Two purple quotation marks on a white background, symbolizing quote or citation indicators.
Managing so many models has made things more complicated day-to-day. It’s harder to keep track of updates, make sure everything works together, and stay consistent across teams. It definitely slows things down and adds extra work.
Velia Carboni
EVP and CDTO
VF Corporation
Blue circular logo with white serif letters V and F.
Production Truth #05

The 2026 Paradox: ‍‍
Expanding Into the Gap

Teal checkmark symbol on a white background, within a rounded, six-sided shape icon.
Tool ecosystem expansion as top priority
Tool‑ecosystem expansion — adding more MCP servers, connecting more internal systems to AI agents, and wiring more enterprise APIs — is the #1 investment priority, cited by 27% of respondents.
Teal checkmark symbol on a white background, within a rounded, six-sided shape icon.
Competitive pressure to expand
Organizations that connect more tools, automate more workflows, and give agents more reach will create compounding advantages over those that wait, driven by strong competitive pressure.
Teal checkmark symbol on a white background, within a rounded, six-sided shape icon.
Every new tool is a governance backlog item
Every new tool connection is also a new entry into the governance backlog — a new endpoint that must be authenticated, logged, audited, and secured. That backlog is already behind.
Teal checkmark symbol on a white background, within a rounded, six-sided shape icon.
Building tool surface first creates structural liability
They are putting a centralized AI gateway in place — a layer that sees all model calls, enforces all policies, and logs all decisions — and then expanding the tool surface on top of it, not the reverse.
Teal checkmark symbol on a white background, within a rounded, six-sided shape icon.
Liability grows with each integration
This liability becomes harder to remediate with every new integration, as the surface grows larger and more fragmented.
27%
#1 INVESTMENT PRIORITY IN 2026
Tool Ecosystem Expansion
More MCP servers. More internal APIs connected to agents. More tool endpoints giving AI greater reach across enterprise systems. This is where the budget is going.
31%
#1 RISK FACTOR IN 2026
Tool Surface Security
Uncontrolled tool proliferation, unauthenticated endpoints, and no central enforcement layer. The very thing enterprises are investing in is also what keeps their CISOs up at night.
2026's
Defining Paradox
The same capability enterprises plan to invest in most aggressively is also their single largest stated risk, defining the 2026 AI landscape.
They put a centralized AI gateway in place, a layer that sees all model calls, enforces all policies, and logs all decisions, then expand the tool surface on top of it.
Building Control Layer First
Winning
Enterprises Act
Differently
Enterprises that navigate this paradox successfully build control infrastructure before they expand, not after.
Liability Grows
with Each
Integration
05
Truth
Tool Ecosystem
Expansion Priority
Governance
Pressure Arrives
Too Late
By the time external governance pressure arrives, the tool surface is too large and fragmented to be addressed quickly or coherently.
FROM THE FIELD
Two blue quotation mark icons on a white background, symbolizing quotes or testimonials in typography design.
Doing differently? Implementing unified GenAI governance with real-time cost tracking, policy-based access controls, and centralized observability across models and tools.
Kumar Gautam
Principal Architect
Pure Storage
Pure Storage company logo with grey and black design on white background.
Two blue quotation mark icons on a white background, symbolizing quotes or testimonials in typography design.
We are establishing shared infrastructure, governance frameworks, and clear decision-making around model usage, access control, and cost accountability. We are also prioritizing repeatable accelerators instead of one-off solutions, so every successful implementation becomes a reusable template for faster scale across the organization.
Deepak MK
VP Data Science
Examroom AI
Grey Examroom.ai logo on white background with triangle symbol next to text.
2026 Investment Priorities

What Enterprise Leaders Are Doing About It

Despite the challenges, the data shows clear momentum around specific investments
Here's where enterprise AI budget is flowing in 2026
27%
Tool ecosystem expansion
25%
More Agents
14%
Centralized Control Layer
12%
More Models
10%
Security Enforcement
10%
Cost Visibility
ADDITIONAL VOICES
Two blue quotation mark icons on white background for citation and dialogue representation purposes.
We've moved from evaluating models to evaluating the infrastructure around models. The model is almost a commodity — what differentiates production at scale is routing, logging, cost control, and the ability to swap without rewriting your applications.
Daniel Shir
VP Engineering
Via Transportation
Dark blue Via logo featuring a stylized map pin icon with a circular shape and lettering.
Two blue quotation mark icons on white background for citation and dialogue representation purposes.
We run a hybrid model stack — some proprietary, some open source, some fine-tuned. The only way to manage that sanely is with a unified gateway layer. Without it you'd need a different integration, a different logging approach, a different cost model for each. It doesn't scale.
Lucas Hendrich
CTO
Forte Group
Forte Group company logo displayed on a white background with black text.
Featured Voices

The Leaders Who Shaped This Report

Some practitioners whose depth of insight drove the five production truths — from global banks
navigating compliance to AI-native startups building governance from scratch.
Alan Chan
Vice President
IT
Evercore logo on a white background with blue serif text displaying the company name.
Biggest AI Risk in 2026?
Architectural complexity  
It’s not simply requests × price. The costs skyrocket because of longer-than-expected prompts context stuffing, tool traces, multi-step agent loops, retries and fallbacks.
Priority
Tool Expansion
Linda Hermer
CEO
Vantage Precision Health logo featuring gradient V symbol with DNA sequence and company name text.
Biggest AI Risk in 2026?
Regulatory non-compliance
Compared to gen AI, it's been challenging (to say the least) to manage security with agentic AI systems. One problem is over-permissioned agents, which are often granted excessive access to internal databases, SaaS applications, and APIs.
Priority
Tool Expansion
Satish Tatini
バイスプレジデント
VFコーポレーション
Blue circular logo with white serif letters V and F.
2026年の最大のAIリスクは？
アーキテクチャの複雑性
ユーザー入力には悪意のあるコンテンツや機密性の高いコンテンツ（プロンプトインジェクションやデータ抜き取りの試みなど）が容易に含まれる可能性があり、モデルは本質的に何が安全で何が安全でないかを認識しないことに気づきました。これは、特にモデルが内部システムに接続されている場合に弱点となります。
優先事項
モデルの増加
バラージ
CTO
Vertafore company logo with orange geometric symbol and dark blue text on white background.
2026年の最大のAIリスクは？
セキュリティ上の脆弱性
コスト、セキュリティ、トレーニングは、モデルの急増により摩擦が生じている主要な側面のひとつです。
優先事項
エージェントの増加
クマール・ゴータム
クマール・ゴータム
PureStorage logo with orange geometric symbol and company name in black and gray text.
2026年の最大のAIリスクは？
コストの予測不能性
2026年の生成AIのリスクは、モデル、ツール、ワークフローの各層にわたる統一された可視性と制御の欠如である可能性があります。これにより、ポリシーの適用に一貫性がなくなり、監査上のギャップが生じるでしょう。
優先事項
集中制御レイヤー
ダニエル・シャー
ダニエル・シャー
Blue location pin icon next to Via Transportation text indicating transportation service location marker.
2026年、最大のAIリスクとは？
セキュリティリスク
モデルの変更やドリフトが最終的に比較的決定論的な結果をもたらすことを確実にするため、テストインフラストラクチャを構築する必要がありました。
優先度
その他のエージェント
Gen AI調査レポート全文を読むにはサインインしてください
AIインフラストラクチャに対する考え方を変える5つの知見
Wavy lines in light purple on a white background creating an abstract pattern design element.
200以上の組織 · すべてAIを本番環境で運用中

この研究を支えるコミュニティ全体

Acceldata logo with blue text on white background, data company branding.
オムカー・バサリカッティ
MLエンジニア
Vertafore company logo with orange geometric symbol and dark blue text on white background.
バラジ
最高技術責任者
Via Transportation logo featuring blue and black text with a location pin icon next to it.
ダニエル・シャー
VP エンジニアリング
Greenhouse Software logo with stylized letter G and text next to it in green and black.
ホルト・カルダー
アナリティクスディレクター
Forte Group logo featuring a stylized heart symbol with an arrow on a white background.
ルーカス・ヘンドリック
最高技術責任者
Interactions logo featuring orange and white symbol next to company name in dark text.
スリニバス・バンガロール
SVP AI
Intel logo with black and blue text on white background, official brand emblem symbol.
匿名
プリンシパルエンジニア
Wells Fargo logo on white background with black square and text.
ビジット・ゴーシュ
 エンジニアリング
State Street logo with three blue lines next to the company name in dark blue text.
ムクタ・マヘシュワリ
ソフトウェアエンジニア AVP
Apollo.io logo with yellow square and black asterisk symbol next to text.
カラン・カクワニ
MLエンジニア
PureStorage logo with orange geometric symbol and company name in black and gray text.
クマール・ゴータム
プリンシパルアーキテクト
EPAM Systems company logo with epam text in black on white background.
ドミトロ・ジョロフ
テクノロジー担当ヴァイスプレジデント
VF Corporation logo featuring blue circle with white letters and text on white background.
サティシュ・タティニ
データおよびデジタル担当ヴァイスプレジデント
Pixis.ai logo with dark green abstract symbol next to bold text on white background.
アミール・アザム
データサイエンティスト
Kiwi.com logo with green square and white letter K and circle next to website text.
イェレナ・アレクシッチ
データサイエンス責任者
Evercore logo in blue text on a white background, representing financial services and investment banking firm.
アラン・チャン
IT担当ヴァイスプレジデント
Caesars Entertainment logo with gold and black color scheme and circular emblem icon design element.
アミット・アガルワル
AIおよびデータ担当ヴァイスプレジデント
VF Corporation logo featuring blue circle with white letters and text on white background.
ヴェリア・カルボーニ
執行副社長 兼 最高デジタル・技術責任者
回答者層
さまざまな業界の200名以上の回答者がAIイニシアチブを主導
自動車
ヘルスケア
保険
ロジスティクス
メディア・出版
その他多数

残りのすべてを扱いやすくするレイヤー

5つの実運用における真実すべてにおいて、これらの課題を管理している企業と、課題を蓄積している企業とを隔てる単一の構造パターンが存在します。チーム、エージェント、そしてあらゆるモデルおよびツールエンドポイントの間に位置する単一のレイヤーである集中型AIゲートウェイは、コストの帰属を可能にし、ガバナンスを強制可能にし、セキュリティを監査可能にし、リスクを伴う拡張という矛盾を実際に乗り越えられるようにします。それがなければ、追加する新しいモデルや接続する新しいツールが増えるたびに、システム全体の管理はより困難になります。それがあれば、その逆が真実となります。
これこそがTrueFoundryのAIゲートウェイが構築された目的です。抽象化レイヤーとしてではなく、大規模なエンタープライズAIが実際に必要とする運用制御プレーンとして。
200以上
本調査のために実運用経験を共有したエンタープライズリーダー
18以上
金融サービスからヘルスケア、テクノロジーまで、代表される業界
5
実運用における真実 — データセット全体で十分に一貫しており、普遍的と呼べるパターン
調査方法

本調査について

200以上
受領した回答総数。本レポートの分析は、確認済みのエンタープライズ本番環境導入からの検証済み回答に基づいています。
32
モデルアクセス、エージェントワークフロー、コスト可視性、ツールエンドポイント、セキュリティ体制、ガバナンス、2026年の優先事項を網羅する調査項目。
2026年3月～4月
調査期間。すべての回答者は、AIを積極的に導入している組織のエンタープライズ実務家（VPレベル以上、または直接的な生産責任を持つシニア個人貢献者）です。
回答者の開示とプライバシーに関する通知
本レポートに掲載されている個人および組織は、エンタープライズGenAIの独立したソートリーダーおよび実務家として自発的に参加しました。すべての回答は、実務家の経験と視点を具体的に把握するために設計された構造化された調査を通じて収集されており、企業独自の機密情報や商業的に機密性の高い情報は一切求められておらず、含まれていません。企業ロゴと所属は、調査時点での回答者の専門的な背景を示す目的でのみ表示されており、本レポートで言及されている調査結果、製品、またはサービスに対する組織的な承認を意味するものではありません。特定されることへの明示的な同意を提供しなかった回答者は匿名化されています。匿名化されたプロファイルが特定の個人と類似している場合でも、それは偶然の一致です。