8. IT Carveouts in the Era of Agentic AI - Point of View

IT Carveouts key Principle: Like by Like replacement

IT carveouts have traditionally been about identifying and removing IT entanglements from ParentCo, while setting up NewCo’s IT function, systems, infrastructure, and data. The basic principle of IT carveouts has long been a like-by-like replacement approach, mainly to fast-track the carveout timeline. Traditional IT carveouts assumed that the existing SaaS landscape would continue as is, and that any on-premises systems requiring local or hosted infrastructure should ideally be replaced with an equivalent SaaS solution. The IT carveout playbook was clear: set up like-by-like applications, replicate workflows, and stabilise IT operations post-transition. Any system optimisation was typically left for later, after complete IT separation had been achieved.

IT Carveouts in the era of Agentic AI:

Agentic AI is actively reshaping the traditional IT carveout thought process and creating a fundamental mismatch, carving out a business based on a system architecture that may soon become obsolete. Agentic AI is set to challenge one of the core principles of IT carveouts: replacements may no longer be purely like-for-like. Instead, optimization may be needed right at the initial stage of defining the NewCo Target IT Operating Model (IT TOM). IT carveouts may no longer be just about separating systems; they will increasingly be about redefining which systems actually need to exist, and where Agentic AI can replace or augment traditional systems. Key reasons include:

• SaaS tools are no longer passive executors: when combined with AI agents, they become active decision-makers. The enterprise software being carved out today is not the same in function or value as it was even two years ago. Treating it as such during an IT carveout risks mispricing both opportunity and risk
• Entire categories of tools: such as analytics, design, and brainstorming are being compressed or replaced by AI capabilities
• Core enterprise systems: are not disappearing, but their role is shifting from being human-driven to agent-driven
• New layers: including agent orchestration, data platforms, and compute infrastructure are becoming more critical than ever

Agentic AI introduces a new reality across the enterprise SaaS landscape.

• AI collapses tools that were built for human thinking

• AI rebuilds systems that run enterprise workflows

• AI amplifies infrastructure that powers everything underneath

This leads to three clear categories:

1. Impacted (High Risk): 

• Tools focused on analysis, creation, and ideation are being replaced by direct AI outputs

• These are no longer stable dependencies in a carveout, they may become liabilities.

• Most exposed and high-risk SaaS 

• e.g. Tableau · Power BI · Photoshop · Canva · Miro · Notion

2. Evolving (Must Become AI-Native):

• Systems of record, workflows, and collaboration platforms remain, but their interaction model shifts from human-driven to agent-driven execution

• These systems must be re-architected, not just separated

• Rebuilt, not replaced

• e.g. Salesforce, SAP, Workday, ServiceNow, Slack, Microsoft Teams

3. Beneficiaries (Compounding Value): 

• Cloud, data platforms, cybersecurity, and agent infrastructure become more valuable with AI adoption

• These are the new backbone of any carveout strategy

• More AI adoption → more demand - Compounds

• e.g. AWS, Azure, Snowflake, Databricks, CrowdStrike, Okta, LangChain

Having delivered 50+ operational and IT carveouts, CIDEK's recommends a different approach in the Agentic AI era, first sense-checking and analyzing whether NewCo actually needs a traditional system separation, or instead an Agentic AI-driven system redesign. The intention will no longer be to recreate the same technology stack within a new entity, but to build a lean, AI-native operating model where agents orchestrate execution across fewer, stronger systems. As a result, the IT separation process may increasingly need to focus on the following:

• Eliminate before separation: Do not blindly migrate existing tools. During IT DD, identify SaaS layers that are being compressed or replaced by AI, and reflect this within the IT TOM accordingly. Category 1 (Impacted) tools may increasingly be treated as replacement candidates rather than migration candidates.
• Rebuild as AI-native: For systems that remain, such as ERP, CRM, and workflow platforms, the redesign may no longer just be technical, but operational. Organizations may need to rethink how work gets done and how processes fit together. This means shifting from task management to agent-execution pipelines, moving from human-led workflows to autonomous processing, and enabling agent-to-agent interactions across systems. AI readiness may need to be embedded into NewCo systems from the first soft launch, rather than added later as an afterthought.
• Strengthen the core layers: Double down on Category 3 (Beneficiary) infrastructure as a Day 1 strategic priority, rather than treating it as a back-end migration task. This includes data quality and accessibility, cloud and compute scalability, security and identity frameworks, and agent orchestration infrastructure where the NewCo strategy involves deploying AI agents. These are the foundational layers that compound in value as AI adoption scales.

Challenges facing IT Carveouts in era of Agentic AI

• Agentic AI driven IT carveout will still need to consider the overall TSA exit timeline - which means building IT TOM and executing it within the available timeline

• As always, the carveout will continue to have continuous pressure from sell side for a quick and clean separation

• A lack of Agentic AI trustworthy expertise to make decisions favoring the NewCo system architecture, will influence whether business go for a traditional system separation, or AI driven system architecture