Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstalliTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Complexity can be a major barrier to enterprise AI because agents must work across an organization’s existing applications, data, integrations, and rules. In a September 2026 CIO opinion article, Shannon Bell, OpenText’s executive vice president, chief digital officer, and chief information officer, argues that accumulated IT complexity can make AI harder to deploy and operate. Her account is a useful executive perspective, not proof that complexity is the single cause of enterprise AI failures. The practical response is to remove unnecessary dependencies, prepare data and governance, and grant agents only the authority their demonstrated performance and the task’s risks justify.
Why complexity makes enterprise AI harder
An AI agent does not operate in isolation. To complete a business task, it may need to retrieve information from several systems, interpret different data formats, follow access rules, and pass work between applications. Every integration adds dependencies that must be maintained, secured, and monitored.
When systems are customized, fragmented, or poorly documented, the agent also encounters more exceptions. The resulting engineering and operational effort can outweigh the value of automating the task. AI may expose existing problems in an environment rather than resolve them automatically.
Free tools Windows power users keep installed
One-click scans. No signup required.
Bell reports that OpenText started with more than 1,500 applications and later reduced its application landscape by more than 300 applications and consolidated over 40 data centers. Those are figures from her description of her own organization, not industry benchmarks or targets that every company should copy.
#1 Best Overall
Bell’s article also cites CIO News reporting that 68% of CIOs say technical debt from past integrations is blocking their ability to scale AI. The article does not provide the underlying survey methodology, so the figure is best treated as a statistic reported by the opinion piece, not a standalone measure of all CIOs.
Which complexity should you remove?
Simplification does not mean forcing every workload into one architecture. Bell describes OpenText as operating across data centers, public and private clouds, and sovereign environments to meet differing workload requirements. Security, regulation, data sovereignty, and business needs can make some complexity necessary.
The useful distinction is between complexity that serves a purpose and complexity that accumulated without one. In an AI initiative, review each dependency and ask whether it supports a current requirement or simply persists because of earlier growth or integration decisions.
Recommended Free Tools
- Keep complexity with a clear requirement: controls needed for security, regulatory obligations, sovereignty, or a genuine business need.
- Investigate complexity without a current purpose: duplicate applications, unnecessary handoffs, redundant data paths, and integrations that add support work without helping the target workflow.
- Avoid simplification for its own sake: a uniform stack can be a poor fit when workloads have different operational or location requirements.
How to scale an AI agent safely
Start with a task whose inputs, expected outputs, and consequences are understood. Map the systems and information involved before granting access, and decide what success and failure look like. For consequential decisions, have the agent analyze and recommend while a person makes the final call.
- Map the workflow. List the systems the agent will connect to, the information it needs, its proposed permissions, and the expected result. Include the handoffs and exceptions that people currently manage.
- Choose a bounded task. Prefer a workflow with known inputs and outputs over an open-ended mandate. Define acceptable results and the errors that require escalation.
- Begin with limited authority. Let the agent analyze or recommend before allowing it to take consequential actions. Match human review to the impact of a mistake.
- Evaluate outcomes. Record whether recommendations were correct and, when they were not, what differed. Use those observations to improve the workflow and judge whether performance is reliable enough for more responsibility.
- Expand cautiously. Increase the agent’s permissions or scope only when evaluations support the change. Keep activity visible, monitor outcomes after deployment, and retain a way to stop the agent.
Bell describes OpenText’s network and security operations team using a resolution agent to analyze incidents and recommend a resolution, with a human making the final decision at the time described. This is an example from her account, not an independently tested case study.
Prepare data, permissions, and oversight
An agent needs relevant, trustworthy context to produce useful work. Before connecting one to enterprise information, identify what data exists, where it resides, how it is governed, what may be exposed to AI, and where sensitive information will be processed. Bell says data quality affects whether business users can rely on an agent’s output.
Access should reflect the agent’s role and the task—not simply mirror broad access available to a user or system account. Make the agent’s activity and the data it uses visible to appropriate people. Continue evaluating behavior in production and preserve a stop mechanism so the organization can respond if the agent acts unexpectedly.
Complexity is not only a technical problem
Integration and data issues matter, but organizational agreement can also slow modernization. A CapTech research announcement reports results from a Harris Poll survey conducted May 8–22, 2026, among 302 director-level-and-above IT decision-makers at US organizations with at least 250 employees that were already using AI beyond the pilot phase. CapTech is a consultancy that sells related services, so these findings should be read in that context.
Best Value
| Finding | Survey result and scope |
|---|---|
| Stakeholder misalignment and decision-making versus technical barriers | 56% said stakeholder misalignment and decision-making hinder modernization more than technical barriers; 44% pointed to technical barriers. CapTech/Harris Poll, May 8–22, 2026; 302 qualifying US IT decision-makers. |
| Implementation versus internal decision-making | 86% agreed their organizations could rapidly implement AI but internal decision-making slowed progress. Same survey population and dates. |
| Shared understanding of tradeoffs | 90% believed modernization decisions would improve if stakeholders established a shared understanding of tradeoffs. Same survey population and dates. |
CapTech CTO Brian Bischoff says that clear ownership, governance, and measures of success help organizations turn AI initiatives into business outcomes. For a project team, that means agreeing early on who owns the workflow, who approves changes in agent authority, which tradeoffs are acceptable, and how success will be measured.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What other surveys say about agent barriers
A vendor-produced Tray.ai infographic reports that 38% of surveyed enterprise leaders identified integration complexity as the biggest barrier to scaling AI agents, 57% cited security concerns as the top barrier to agent success, and 79% expected data challenges to affect agent rollouts. Tray.ai says the findings are based on a survey of more than 1,000 enterprise leaders across industries focused on agent development and deployment strategies. The infographic does not state the survey dates or fuller sampling details, and Tray.ai sells products in this area; these figures are useful context, not neutral product-performance evidence.
Bell’s CIO article also cites McKinsey figures saying nearly two-thirds of enterprises worldwide have experimented with agents, while fewer than 10% have scaled them to tangible value. Because the article does not identify the exact McKinsey report or fieldwork, treat that comparison as a secondary reference rather than a standalone primary statistic.
A practical readiness checklist
Before expanding an agent beyond an initial task, confirm that the organization can answer these questions:
- Which business outcome should the agent produce, and how will the team judge whether it did so correctly?
- Which applications, integrations, and data sources are essential to that task, and which dependencies could be removed?
- Where does the required data reside, how reliable is it, and what rules govern its use and processing?
- What is the least access the agent needs, and which actions require human approval?
- Who owns the workflow and its governance, and have stakeholders agreed on the tradeoffs and success measures?
- How will the team review agent activity and performance, handle errors, and disable the agent if needed?
Use these criteria to compare implementation approaches rather than assume one vendor or architecture is best: integration burden, data quality and governance, security controls, fit for hybrid or sovereign requirements, clarity of responsibilities, human review and rollback, visibility and evaluation, and the organization’s ability to agree on ownership and outcomes.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
