Share
AI’s value creation potential is significant, but technology alone will not deliver measurable returns. For private equity sponsors, success starts with the leadership, governance, and operating change to convert efficiency into EBITDA.
This chapter is an excerpt from our Private Equity Value Creation Playbook. Get your copy.
AI has quickly become one of private equity’s most talked-about value creation levers, but it is also one of the most under-realized. Many firms see AI as a way to automate manual work, improve decision-making, expand margins, and ultimately improve EBITDA. The technology has matured, proven use cases continue to expand, and vendors are introducing new capabilities to market at a rapid pace. Yet many portfolio companies still struggle to turn AI investments into measurable financial results. Even when efficiency improves, those gains often never reach the P&L.
Here we examine why AI initiatives stall, what separates tool adoption from true operating capability, and how leading firms build AI into the business in ways that create lasting value.
Why AI initiatives fail
Many organizations approach AI as if deploying the technology alone will create value. This is a fundamental error. AI creates value only when it’s supported by reliable data, integrated into everyday operations, and governed in ways that manage risk while enabling responsible adoption.
Most failures fall into five familiar patterns:
1. AI deployed onto broken data foundations
The most common mistake is deploying AI before the underlying data is ready to support it. AI models are only as good as the data they consume. Fragmented data, inconsistent definitions, poor data quality, and unclear lineage lead to outputs that are inconsistent, or simply wrong.
Organizations that skip this foundational work quickly lose confidence in the results. The model outputs cannot be reconciled with operational reality. The business loses trust in AI as a decision-support capability. The investment is written off, and the organization defaults back to the manual processes AI was supposed to replace.
2. Objectives are vague and unmeasured
The second failure is unclear business objectives. “Deploy AI” is not an objective. “Improve operations with AI” is not an objective. Objectives must specify what business outcome is being pursued, how it will be measured, and how AI is uniquely positioned to deliver it.
Without clear objectives, AI initiatives lose direction. Pilots multiply without clear success criteria. Vendors are evaluated on technology features rather than business outcomes. The organization accumulates AI tools without accumulating AI value.
3. Efficiency gains are not converted to ROI
The third failure is efficiency without ROI. AI can reclaim time, such as in accounts payable processing, customer service triage, inventory forecasting, and document review. But reclaimed time may not be a financial return. It could be if the employees are hourly, for example, or if throughput is in a manufacturing (or similar) environment, but less so for salaried employees or administrative departments.
Sponsors who measure AI success by general efficiency metrics – hours saved, throughput improved, error rates reduced – may discover that those gains never reach the P&L. Headcount does not automatically decrease. Capacity is not automatically redeployed to revenue-generating work. Throughput improvements do not always translate to revenue growth. To see real ROI, further tailored strategy is needed.
4. AI adoption is treated as tool adoption, not operating change
The fourth failure is treating AI as a software deployment rather than an operating model change. Buying software, training employees, and launching the platform is not the same as adopting AI.
If the workflow remains unchanged, AI simply sits on top of the existing process and delivers only incremental improvement. When AI is integrated into workflow and operating model design, it becomes embedded in how the business operates, driving greater efficiency, better decisions, and more sustainable value creation.
5. Shadow AI and governance gaps create risk without value
The fifth failure is shadow AI – the AI use you don’t know about. Employees deploy consumer AI tools (ChatGPT, Claude, Copilot) without governance, data access controls, or an understanding of the regulatory and competitive implications.
Organizations that delay AI governance often discover that employees have already adopted AI in ways that create regulatory, competitive, and reputational risk. The cleanup is expensive; the prevention would have been cheap.
AI creates value only when it’s supported by reliable data, integrated into everyday operations, and governed in ways that manage risk while enabling responsible adoption.
Are your portfolio companies experiencing shadow AI or failed pilots?
In the complete Private Equity Value Creation Playbook, this AI chapter (13) also includes our 8-point AI Value Creation Failure Mode Checklist and a set of Reflection Questions for board reviews. Download the Full Playbook & Diagnostic Tools: https://www.cohnreznick.com/insights/private-equity-value-creation-playbook
Tool adoption vs. operating capability
The most successful sponsors treat AI as an operating capability rather than a software purchase. The distinction matters. Tool adoption focuses on implementing technology. Operating capability changes how work gets done.
The reframe rests on four elements:
1. Use case specificity vs. generic deployment
Generic AI deployment fails. Specific use case deployment succeeds. The organization must define which business process AI will improve, what data it will use, and how the outputs will inform decisions.
Specificity prevents the “wandering pilot” problem: It forces the organization to identify a concrete business outcome before evaluating technology, and helps ensure that vendor selection is driven by use case fit rather than feature breadth.
2. Maturity prerequisites vs. aspiration
Successful AI capabilities require digital maturity.
| AI Type | Examples | Digital Maturity Required |
|---|---|---|
| Embedded | Recommendation engines, document classification, pattern recognition | Low |
| Generative | Content creation, conversational interfaces, analytical synthesis | Moderate |
| Agentic | Autonomous decision-making, multi-step workflows | Advanced |
3. ROI mechanics: Efficiency vs. EBITDA conversion
Again, remember that efficiency is not necessarily ROI, and organizations that measure AI success purely by baseline efficiency metrics – hours saved, throughput improved – often discover that those gains never reach EBITDA.
The conversion plan must be defined before deployment. If leadership cannot explain how reclaimed time will be converted into financial return, AI will produce efficiency without ROI every time.
4. Process redesign vs. workflow automation
Automation makes existing processes faster. Process redesign makes existing processes obsolete. AI capability development requires process redesign.
If the process is automated without being redesigned, the AI becomes an overlay producing marginal value. If the process is redesigned around AI capability, the AI becomes the foundation producing transformational value. The difference is whether the sponsor treats AI as a tool to support existing work or as a capability to redefine how work is done.
The role of leadership, incentives, and governance
Technology alone won’t make AI successful. Leadership is required. The tone, urgency, and accountability set by the executive team determine whether AI becomes a priority or remains a technology pilot.
Tone at the top: AI owned by leadership, not delegated to IT
If the CEO, CFO, and operating partner are not visibly committed to AI – reinforcing it in operating reviews, tying it to performance metrics, and holding teams accountable – it will not happen. When the CEO talks about AI in operating reviews, when AI use cases are on the board agenda, when AI outcomes are tracked alongside EBITDA, the organization understands that AI is the priority. When these signals are absent, AI competes for attention with operational priorities – and loses.
Incentive alignment: Compensate labor redeployment, not just time savings
Incentives should reward converting AI efficiency into financial return. This includes:
-
- Tying functional leader compensation to labor redeployment outcomes, not just AI deployment milestones
- Rewarding capacity reallocation from automated tasks to revenue-generating work
- Penalizing efficiency claims that do not produce measurable EBITDA impact
When leadership is compensated to convert AI efficiency to ROI, behavior changes. When AI deployment is treated as the success metric, organizations deploy AI without capturing value.
Governance and approval processes: Shadow AI controls and data access discipline
Effective AI governance creates just enough friction to make sure that AI is deployed safely and value is captured.
AI governance isn’t there to slow down AI deployment. It’s there to make sure the deployment actually produces value rather than risk. Approval gates, data access controls, and use case review aren’t bureaucracy – they’re the discipline that prevents shadow AI from creating exposure faster than authorized AI creates value.
Good governance gives people clear boundaries, protects sensitive data, and allows AI to be adopted with confidence.
Sequencing AI by digital maturity
Successful AI adoption begins by matching the right type of AI to the organization’s use case and level of digital maturity. Choosing the wrong type of AI for the organization’s maturity almost always leads to disappointing results.
Embedded vs. Generative vs. Agentic AI
As mentioned earlier: Not all AI is the same. Embedded, generative, and agentic AI each require different levels of digital maturity and create value in different ways.
-
- Embedded AI is integrated into existing software platforms – CRM recommendation engines, ERP forecasting, document classification. It requires modest digital maturity, integrates with existing workflows, and produces incremental value. Most middle-market portfolio companies should start here.
- Generative AI creates new content – written outputs, analytical synthesis, conversational interfaces. It requires a stronger digital foundation and thoughtful workflow redesign to deliver meaningful results.
- Agentic AI executes multi-step workflows autonomously – taking actions, making decisions, completing transactions. It demands a high level of digital maturity and significant process redesign, but it also has the potential to create the greatest value.
Sequencing by maturity level: The judgment call
Organizations that pursue agentic AI before their data foundation is ready often end up with expensive failures. At the other end of the spectrum, organizations with strong digital maturity may limit their potential if they stop at embedded AI.
The judgment call requires honest maturity assessment:
-
- Data foundation quality and lineage
- Workflow standardization and documentation
- Talent capacity for AI-enabled work
- Governance infrastructure for AI risk
- Capital availability for AI investment
An honest assessment leads to better sequencing decisions than an overly optimistic one.
Organizational readiness: People, process, technology – in that order
Successful AI adoption starts with people, then process, and finally technology. Technology can accelerate execution, but it cannot replace skilled people, well-defined use cases, or strong governance.
1. People First: Data stewards and AI literacy
Before deploying AI, organizations need the people and skills to support it. This includes:
-
- Data stewards who own data quality, lineage, and access for the use cases AI will support
- Functional leaders with sufficient AI literacy to evaluate use cases, vendor capabilities, and deployment outcomes
- Technical talent capable of integrating AI capabilities with existing systems and workflows
2. Process Second: Use case definition, governance, redesigned workflows
With the right people in place, attention can shift to process. This includes:
-
- Defining specific use cases with measurable business outcomes
- Establishing AI governance – approval gates, data access controls, model review
- Redesigning workflows around AI instead of layering it onto existing processes
3. Technology Last: Embedded → generative → agentic
Technology deployment follows people and process readiness. As discussed, the sequence within technology matters: embedded AI first, generative AI second, agentic AI last. Organizations that move to generative or agentic AI before mastering embedded AI often discover they aren’t ready to support the added complexity. The maturity gap becomes the obstacle.
AI as a multi-year system: Readiness cadence, operating model rationalization, continuous improvement
AI is not a one-time deployment. Organizations that create lasting value build it into the way they operate over the entire hold period. The most successful sponsors treat AI as a long-term operating capability, rather than a series of disconnected pilots.
1. Establish AI readiness discipline early
The First 100 Days should establish the AI readiness baseline:
-
- Digital maturity assessment across data, process, talent, and governance dimensions
- Use case inventory prioritized by business impact and feasibility
- Governance framework covering data access, model approval, and risk management
- Talent gap assessment and remediation plan
Organizations that establish this discipline early build a stronger foundation for AI. Those that skip it often discover they’re deploying technology before the business is ready to support it.
2. Use AI to rationalize the operating model
AI creates an opportunity to rethink how the business operates. When AI can eliminate manual work, organizations should ask whether the process still needs to exist at all. The same applies to workflows: Redesign them where AI creates a better way of working instead of layering technology onto the existing process.
Organizations that use AI only to support existing ways of working often miss its greatest potential. Those that use it to rethink how work gets done create far more value.
3. Monitor, measure, and adjust continuously
AI performance should be monitored as rigorously as any other value creation initiative. Metrics include:
-
- AI use cases deployed vs. planned
- Efficiency gains captured vs. modeled
- ROI conversion rate (efficiency to EBITDA)
- Shadow AI incidents and governance compliance
- Talent capacity expansion
AI performance should be a standing agenda item in operating reviews, with clear accountability for results.
Looking further ahead: The frontier of technology and AI
Predictive and agentic AI have the potential to reshape how portfolio companies create value, but expectations need to match today’s realities.
As these technologies mature, they will increasingly be able to:
-
- Execute autonomous knowledge work including financial analysis, customer service triage, and document processing
- Orchestrate cross-enterprise workflows spanning supply chain, finance, and commercial functions
- Score predictive demand and customer health in real time, enabling proactive intervention
- Model real-time scenarios for capital allocation, M&A evaluation, and strategic decisions
As always, these technologies won’t replace strategy, leadership, or accountability. They deliver the greatest value when the fundamentals – people, process, and governance – are already in place.
For most middle-market companies, the priority should be building strong data foundations, governance, and embedded AI before investing heavily in predictive or agentic AI.
In conclusion: From AI investment to business value
AI has the potential to become one of private equity’s most powerful value creation levers, but only when it drives meaningful operational change. The gap between investing in AI and creating value from it isn’t a technology problem. It’s an execution problem driven by digital maturity, governance, process redesign, and the ability to convert efficiency into measurable results.
The sponsors who realize AI value over the hold period:
-
- Treat AI as an operating capability, not simply another software tool.
- Match the type of AI to the organization’s digital maturity.
- Convert efficiency into measurable financial results.
- Establish AI governance early so innovation can move forward without creating unnecessary risk.
- Build in the right order: People first, process second, technology last.
- Approach AI as a long-term capability that evolves throughout the hold period.
Organizations that invest in leadership, governance, digital maturity, and disciplined execution are far more likely to realize AI’s potential. Technology is only one part of the equation; lasting value comes from integrating AI into the way the business operates.
Related services
Our solutions are tailored to each client’s strategic business drivers, technologies, corporate structure, and culture.
Receive CohnReznick insights and event invitations on topics relevant to your business and role.
Any advice contained in this communication, including attachments and enclosures, is not intended as a thorough, in-depth analysis of specific issues. Nor is it sufficient to avoid tax-related penalties. This has been prepared for information purposes and general guidance only and does not constitute legal or professional advice. You should not act upon the information contained in this publication without obtaining specific professional advice specific to, among other things, your individual facts, circumstances and jurisdiction. No representation or warranty (express or implied) is made as to the accuracy or completeness of the information contained in this publication, and CohnReznick, its partners, employees and agents accept no liability, and disclaim all responsibility, for the consequences of you or anyone else acting, or refraining to act, in reliance on the information contained in this publication or for any decision based on it.