For a decade, artificial intelligence sat in the “upside” column of the investment memo: a source of optionality that might improve returns but was never counted on to deliver them. That era is over. AI has moved into the base case: leading investors now underwrite AI-driven revenue and margin improvements directly into deal models, investment committee materials, and value creation plans. When AI is an underwriting assumption rather than a hope, the question that determines returns is no longer whether AI creates value. It is whether the organization can execute.
Execution, however, is precisely where most enterprises fail. MIT research found that roughly 95% of enterprise generative AI pilots deliver no measurable P&L impact. The reason is not because the technology falls short, but because organizations cannot integrate it into workflows, govern it, and scale it. For a private equity firm, that failure rate is not an abstraction. It is the gap between the value creation plan the deal was priced on and the EBITDA the portfolio actually delivers.
This paper argues that private equity is structurally better positioned than any other ownership model to close that gap, but only if firms stop treating AI as a collection of company-by-company experiments and start operating it as a portfolio-wide capability. We will describe a two-level operating model:
We will also outline a governance framework, built on defined autonomy levels, that makes AI spend auditable and ties it to outcomes. Together, these elements convert AI execution in private equity from a scattered set of initiatives into an ecosystem-wide flywheel in which every deployment makes the next one faster, cheaper, and more certain.
The economics of private equity have tightened. With entry multiples elevated and leverage more expensive, Bain & Company estimates that today’s deals require roughly 12% annual EBITDA growth to generate acceptable returns. The historical norm is closer to 5%. That growth must come from somewhere, and increasingly it is being underwritten to come from AI.
McKinsey’s 2026 global private markets research observes that leading firms now reflect AI upside and downside directly in diligence and operational value creation plans, across pricing, sales effectiveness, customer support, software development, and back-office automation. The report also found that firms are ceasing to underwrite AI as a long-dated option in favor of near-term, executable use cases that can move performance within the holding period.
The market data supports the shift. In FTI Consulting’s 2026 survey of 200 private equity fund and operating leaders, the overwhelming majority reported positive financial impact from AI initiatives across their portfolios. The top priority cited was revenue acceleration, not cost reduction.
Boston Consulting Group found that PE-backed companies that systematically build AI capabilities across functions achieve nearly twice the return on invested capital of those that do not. The dispersion between AI leaders and laggards is becoming a dispersion in fund returns.
When AI sits in the base case, an AI plan that fails to execute is no longer a missed opportunity. It is a broken deal model.
Yet adoption remains uneven. Roughly six in ten PE-backed companies have adopted AI in some form: ahead of the broader economy but behind venture-backed peers. However, adoption says little about depth. Most portfolio companies are running point tools and pilots, not operating models. The firms that will win the next cycle are those that treat AI as a core business capability of the franchise itself. It must be embedded in how they conduct due diligence, how they build value creation plans, how they govern their companies, and how they tell the story at exit.
Ambition is not the constraint; execution is. The MIT study referenced earlier found that the causes of AI’s failure to deliver P&L impact are consistent, and have little to do with model quality. Pilots built outside real workflows. No owner accountable for a business outcome. Data foundations that cannot support production use. No governance to decide what scales, what stops, and what it all costs.
Private equity leadership has drawn the obvious conclusion. Among the largest and most sophisticated sponsors, a visible maturity spectrum has emerged. The best-in-class have industrialized AI on both sides of the house:
Others have built dedicated data and AI groups, structured value-creation playbooks, and portfolio AI teams with disciplined ROI tracking. At the other end of the spectrum, AI adoption exists but is fragmented (partner-led, use-case driven, lacking a common platform), which means every portfolio company starts from zero and nothing compounds.
The lesson from the leaders is not that they picked better technology. It is that they built execution capability and operating discipline, and made it institutional rather than personal. That is the capability this paper describes how to build.
A Fortune 500 company that cracks AI in one division still has to fight its own org chart to replicate the win. A private equity firm that cracks AI in one portfolio company owns the governance rights, the board seats, and the operating cadence to replicate it across ten, twenty, or fifty more companies. And, they can underwrite the next acquisition knowing exactly what the playbook delivers. Repeatability is the asset class’s native advantage. Control ownership, concentrated governance, defined holding periods, and a professional operating function are exactly the conditions under which AI execution succeeds.
Capturing that advantage requires operating at two levels simultaneously.
At the portfolio company level, each business needs an execution path matched to its situation: its data readiness, technology estate, talent, and its place in the investment lifecycle. At the firm level, the sponsor needs institutional constructs that carry learning across companies and investment cycles, so the capability belongs to the franchise rather than to individual deal teams or vendors.
The two levels reinforce each other: Portfolio deployments generate the evidence and assets the firm-level constructs codify, and the firm-level constructs make every subsequent portfolio deployment faster and more certain. Most firms today have neither level working systematically. The remainder of this paper describes how Coforge approaches both.
The base-case era moves the starting line into due diligence. If AI value is priced into the deal, the deal team needs to know, before close, whether the target can actually deliver it. Coforge runs this as a compressed pre-deal sprint (we call it Track 0) that de-risks the AI thesis and converts it into a first-100-days execution plan. The sprint establishes an AI and technology baseline covering data readiness, architecture, security and compliance posture, and delivery maturity. It builds a use-case value thesis that shortlists the highest-ROI AI plays with sizing and feasibility, and defines the delivery and operating plan in terms of talent, governance, vendor landscape, and dependencies.
The outputs are deliberately decision-grade rather than encyclopedic:
The sprint costs a fraction of a percent of deal value and determines whether the AI line in the model is an asset or a liability.
No two portfolio companies need the same engagement. A founder-led industrial business two years from a platform modernization needs something different from a software company whose product roadmap is being rewritten by generative AI. Coforge structures post-close execution as three tracks, applied individually or in combination, each designed to meet a company where it is.
| Track 1: AI Strategy | Track 2: Targeted AI Solutions | Track 3: Enterprise Technology Foundation | |
| Focus | A clear, business-aligned AI strategy and readiness baseline | Production-ready AI solutions deployed rapidly into existing workflows | Modernizing the technology stack so AI can scale reliably across core systems |
| What happens |
|
|
|
| Output | Sequenced AI roadmap with prioritized use cases and expected EBITDA uplift | Enterprise-wide AI deployment — faster and more cost-effective than custom builds | AI-ready platforms that remove execution friction and accelerate EBITDA impact at scale |
The tracks are deliberately modular. A company might enter through a Track 1 strategy engagement and graduate into Track 2 deployment; another might need Track 3 modernization before any AI ambition is realistic, because no model performs well on top of a fragile data estate. What matters is honest sequencing: the discipline to fix foundations where they are weak, and the speed to deploy where they are strong.
In our experience, the results compound quickly. Engineering organizations adopting governed, AI-assisted delivery typically see development cycles compress by roughly a quarter, with legacy modernization efforts cutting manual reverse-engineering effort by half or more.
Execution models only work when incentives do. Engagement structures for portfolio companies should reflect PE economics rather than traditional IT services economics:
When the execution partner’s commercial model depends on the same EBITDA bridge the sponsor underwrote, strategy and delivery stop being separate conversations.
Everything described so far can be done one company at a time, and that is precisely the trap. Company-by-company execution, no matter how successful, leaves the firm with anecdotes instead of assets. The learning walks out the door when the deal team moves on or the company is sold. Two institutional constructs convert individual wins into franchise capability: the AI Center of Excellence (CoE), and the Portfolio AI Command Center.
A firm-level AI CoE is a jointly-operated construct built by the sponsor and execution partner that serves as the connective tissue between the firm and its portfolio on everything AI. The AI CoE owns the reusable assets: the reference architectures, the vetted use-case playbooks by industry and function, the diligence toolkits, the governance standards, and the vendor and model evaluations. It persists across portfolio companies and across investment cycles, which is what distinguishes it from a succession of consulting engagements. Modeled on the practices of the most advanced sponsors but tailored to each firm’s operating model and scale, the AI CoE is what allows the tenth portfolio deployment to cost a fraction of the first and deliver in a fraction of the time.
The second construct is a Portfolio AI Command Center: a central operating and reporting layer that aggregates AI activity, outcomes, and readiness across every portfolio company into a single view. For the operating group, it answers the questions that today require a quarter’s worth of board meetings to assemble: which companies are ahead of or behind their AI value creation plans, which use cases are repeating across the portfolio and should be productized, where spend is concentrating and what it is returning. For the investment committee, it sharpens capital allocation between competing initiatives. And for investor relations, it converts a scattered set of AI anecdotes into a quantified, portfolio-wide narrative. This is increasingly a requirement in fundraising, as LPs move from asking whether a manager “uses AI” to asking for evidence of what it has produced.
The Command Center turns “How is AI going?” from a board-meeting anecdote into a portfolio-wide, quantified answer for the operating group, the investment committee, and LPs alike.
The most common failure mode in enterprise AI is not technical; it is fiscal. Organizations invest in AI and early agent deployments without a consistent framework to govern autonomy, measure outcomes, and make value visible and repeatable at scale. The result is the pattern every CFO recognizes: AI investment with no visible return. As agentic AI (systems that act, not just answer) enters portfolio operations, the governance gap becomes a risk gap as well.
Coforge addresses this with a framework we call Autonomy Level Pricing (ALP). The premise is simple: AI systems should be classified, governed, and paid for according to how autonomously they operate, on a defined scale from AL0 (fully human-operated) to AL4 (fully autonomous within policy guardrails).
Each level carries codified accountability like policy guardrails, a registry of deployed agents, and escalation paths, enabling operational and reputational risk to be managed deliberately as autonomy increases. Progress is tracked through two master metrics: the Human Intervention Rate (how often people must step in) and the Acceptance Rate (how often the system’s output is accepted as-is). Together they make autonomy provable rather than assumed, on a live evidence base rather than retrospective claims.
For a private equity owner, the framework does three things:
Operated together, these elements form a flywheel that runs the length of the fund lifecycle. In diligence, the firm underwrites AI value with evidence from its own portfolio rather than market analogies. It can price more aggressively where it knows the playbook works and more cautiously where it knows the hidden costs. During the hold, each deployment adds assets to the Center of Excellence and data to the Command Center, so execution gets faster and cheaper with every company: the second rollout of an AI-enabled customer-operations stack reuses the architecture, the governance template, and the lessons of the first. At exit, the story changes register: From “this company uses AI” to “this company runs a governed, measured AI operating model with a documented earnings contribution,” a claim that buyers can verify and pay for.
The compounding extends beyond any single fund. A firm that institutionalizes AI execution builds a reputation asset: management teams increasingly choose sponsors partly on the operational capability they bring, and LPs increasingly allocate on it. The fragmented alternative, where each portfolio company negotiates its own vendors, builds its own governance, and repeats its own mistakes, leaks value at every seam. In an environment where deals are priced to require double-digit EBITDA growth, few firms can afford the leak.
Private equity firms have no shortage of AI strategies; what they are missing is the capacity to execute them at portfolio scale. Coforge is that execution engine: A partner that translates your firm’s AI ambition into repeatable, scalable operations. Our capabilities are built on three key strengths:
The approach is proven in production, not in pilots. We build the governed enterprise AI foundations that scale generative AI adoption across development organizations, and deliver AI-powered legacy modernization that decodes decades-old systems and regenerates them for the AI era. The common pattern across these engagements also happens to be the central thesis of this paper: Value comes from embedding AI into daily operations with the same rigor as any core capability. It must be governed, measured, and owned.
None of this requires a multi-year program to begin. Two paths can start immediately and run in parallel. The first is portfolio-direct: Jointly prioritize three to five portfolio companies based on live triggers (an upcoming refinancing, a margin gap against the value creation plan, a technology modernization already budgeted, etc.). Run AI strategy and maturity assessments for each, and deploy delivery pods against the highest-value use cases, with results reported to boards within a quarter.
The second is firm-level: Review the AI and technology dimensions of value creation plans across the portfolio, map opportunities to the right execution track, and design the AI Center of Excellence and Portfolio AI Command Center so all learnings from the portfolio path are captured from day one.
The sequencing logic matters more than the pace. Portfolio engagements generate live results and evidence, while firm-level constructs turn that evidence into permanent capability. You can choose the pace, but you can no longer choose whether to build the capability. AI is in the base case now. The only question is whether your execution will follow.
Global Head, Strategic Solutions and Business Transformation Consulting
Pedro Silva works with client executives and Coforge’s technology and advisory partners to co-author transformation agendas – translating enterprise priorities into board-level value cases, and into the operating and commercial models that make them executable. With experience across strategy and corporate finance, management consulting, enterprise transformation, and principal investing, Pedro brings deep expertise in large-deal origination, commercial model design, and AI-enabled value creation. He works with executives who have moved past AI experimentation and need it to show up in the numbers.