What should buyers examine during AI due diligence before acquiring an AI company in 2027?
Buyers should evaluate AI ownership, intellectual property, revenue quality, unit economics, operational resilience, regulatory exposure, and transferability before determining acquisition value.
AI valuation should reflect durable competitive advantages rather than technological novelty. Revenue quality, customer retention, and fully allocated AI infrastructure costs determine economic sustainability. Intellectual property, data rights, and third-party model dependencies can materially affect acquisition value. Operational resilience, cybersecurity, and regulatory governance are essential components of AI due diligence. Buyers should assess whether AI capabilities and contractual rights remain transferable after acquisition.
A Framework For Testing Ownership, Economics, And Transferability Before Paying For AI
AI is becoming a material component of acquisition underwriting. Yet an AI company's headline growth rate, demonstration quality, or access to a leading foundation model says relatively little about the durability of its economics. Buyers need to establish what the target actually owns, what it licenses, what it can operate reliably, and what would remain valuable if the technology landscape changed. In 2027, the central diligence question is not whether a target uses AI. It is whether AI creates a defensible, transferable source of enterprise value.
Start With The Investment Thesis, Not The Model Demo
Every diligence process should begin by specifying the source of expected value. Is the buyer acquiring differentiated intellectual property, a distribution channel, proprietary data rights, an installed workflow, or a team capable of developing models? These are distinct propositions and should not be valued interchangeably. A company that wraps a third-party model in a polished interface may have an attractive business, but its defensibility may lie in customer relationships and execution rather than in proprietary model technology.
Management should demonstrate how AI capabilities translate into measurable customer outcomes. Useful evidence includes retention by cohort, workflow penetration, time saved, accuracy under real operating conditions, and willingness to pay. Buyers should distinguish controlled demonstrations from production usage, and pilot revenue from repeatable commercial demand. A strong investment thesis identifies both the benefit customers receive and the mechanism through which the company captures a durable share of that benefit.
Revenue Quality And Customer Concentration
AI-related revenue deserves the same scrutiny as any software or services revenue, with additional attention to experimentation. Separate recurring subscriptions from implementation fees, usage-based charges, professional services, and one-time pilot programs. Review renewal behavior, expansion, contraction, churn, and the conversion of trials into paid deployments. Usage growth without improving retention or pricing power can create misleading impressions of traction.
Customer concentration matters particularly when a handful of enterprise design partners account for much of reported revenue. Contracts may contain bespoke commitments, broad termination rights, or pricing concessions that are difficult to replicate. Buyers should review whether customers have placed the product into essential workflows or are still evaluating it under innovation budgets. For consumption-priced products, examine whether revenue growth reflects sustained activity or short-lived inference spikes.
Unit Economics Must Include The Full AI Cost Stack
Traditional software gross-margin assumptions can mislead when inference, retrieval, model orchestration, and human review are substantial variable costs. Diligence should attribute compute, third-party API usage, vector storage, monitoring, evaluation, data labeling, and human-in-the-loop operations to the products and customers that cause them. Where shared infrastructure makes attribution difficult, management should disclose the allocation method and sensitivity of reported margins to that method.
Model prices can decline, but that does not automatically improve profitability. Lower prices may stimulate heavier usage, higher service expectations, or price competition. Buyers should stress-test unit economics against changes in model vendors, token consumption, latency requirements, service-level commitments, and customer pricing. An attractive margin profile should survive realistic workloads rather than depend on a narrow benchmark configuration.
Intellectual Property And The Dependency Map
A rigorous review distinguishes owned code, model weights, training pipelines, fine-tuning assets, data rights, prompts, evaluations, and customer-specific configurations. The mere presence of a proprietary orchestration layer does not establish that a target controls the underlying intelligence. Buyers should obtain a dependency inventory that identifies external model providers, open-source components, licenses, hosting environments, and critical subcontractors.
Contractual rights are as important as technical architecture. Confirm whether training and fine-tuning data may lawfully be used for the intended purposes; whether customer agreements restrict secondary use; and whether vendor terms permit the contemplated commercial deployment. Review assignment, change-of-control, audit, and termination provisions. A product may be technically functional yet commercially impaired if a key data license or model agreement cannot be transferred after closing.
Reliability, Security, And Operational Resilience
AI diligence should assess performance under conditions resembling the target's actual customers and workflows. Request evaluation protocols, representative test sets, documented failure modes, incident records, rollback procedures, and evidence of changes across model versions. Accuracy alone is insufficient: buyers should examine consistency, traceability, latency, availability, and how the system responds when confidence is low or external dependencies fail.
Security review should address access controls, tenant isolation, secrets management, prompt injection, data leakage, logging, and third-party exposure. For AI agents that can take actions, inspect the scope of permissions and the boundaries around financial transactions, data modification, and external communications. A system that produces useful answers may still be unsuitable for enterprise deployment if it cannot constrain what an agent is authorized to do.
Governance And Regulatory Exposure
Regulatory exposure depends on sector, geography, use case, and the role the product plays in consequential decisions. Buyers should map applicable requirements rather than assume that one generic AI policy resolves all obligations. Examine documentation of intended uses, risk assessments, human oversight, data retention, incident escalation, and customer-facing claims. In regulated sectors, determine whether the product's evidence trail would withstand a customer's procurement review or a supervisory inquiry.
The diligence team should also test how management substantiates claims about fairness, explainability, reliability, and compliance. Marketing language should not be treated as proof of control effectiveness. Where an AI system influences lending, employment, healthcare, insurance, or other high-consequence outcomes, the burden of demonstrating governance can be materially greater than for a low-risk productivity tool.
Team, Architecture, And Integration Risk
The technical team's expertise should be evaluated against the target's actual differentiation. A company whose competitive position depends on proprietary research requires a different talent profile from one whose advantage is enterprise integration and distribution. Buyers should identify key-person dependencies, undocumented systems, and the practical difficulty of transferring operational knowledge after closing.
Integration planning should begin during diligence. Can the product fit the acquirer's identity systems, security controls, procurement processes, data architecture, and support model? Are there contractual or technical barriers to consolidating infrastructure? An acquisition can destroy value when the buyer underestimates the cost of maintaining multiple model stacks or overestimates the ease of combining customer data.
Valuation: Separate Durable Advantages From Temporary Scarcity
AI valuations often embed expectations about future growth, margins, and strategic scarcity. Buyers should build scenarios that distinguish the base business from optionality. A useful model tests customer adoption, retention, pricing, infrastructure costs, competitive substitution, and required investment. It should identify which assumptions create most of the valuation and whether those assumptions are supported by observed operating evidence.
Strategic value can justify a premium, but it should be tied to specific synergies: distribution access, product acceleration, defensible data, cost savings, or an otherwise difficult-to-build capability. Those synergies should be measured net of integration cost, execution risk, and time to realization. A premium justified solely by the presence of AI is not a substitute for underwriting.
A Practical Diligence Decision Framework
An effective process produces a concise decision record covering five questions. What is being acquired? What evidence proves customers value it? Which economics persist at scale? Which rights and dependencies survive a transaction? What risks could prevent the buyer from realizing the investment thesis? The answer should include unresolved matters, their potential financial consequences, and any protections needed in transaction documentation.
For boards, private equity sponsors, and strategic acquirers, the objective is not to eliminate uncertainty. It is to distinguish risks that can be priced, controlled, or contractually allocated from risks that undermine the acquisition rationale. In AI transactions, disciplined diligence can help buyers move decisively without confusing technological promise with bankable enterprise value.
AI due diligence in M&A requires buyers to assess whether a target’s AI capabilities create sustainable and transferable enterprise value. This includes examining intellectual property ownership, revenue quality, infrastructure costs, customer retention, and dependence on third-party models. Buyers must also evaluate operational resilience, cybersecurity, regulatory exposure, and contractual transferability. A disciplined diligence framework connects these findings to valuation, transaction structuring, and post-acquisition integration risk.
When to speak with Chatsworth
You may benefit from an advisory conversation if your board is evaluating timing, valuation expectations, buyer universe quality, or diligence readiness. Chatsworth provides senior-led perspective on process design and execution risk independently of whether a mandate results.
Speak with the team →This article is published by Chatsworth Securities LLC (CRD #40804) for informational purposes only and does not constitute legal, tax, or securities advice. See our Terms of Use.

