McKinsey & Company reveals how gen AI is accelerating outside-in diligence, enabling organisations to make faster, more informed investment decisions by leveraging proprietary data, customised prompts, and specialised AI agents, though highlighting the importance of human oversight and disciplined integration.
Before committing to major investments, leaders face the challenge of conducting rigorous outside-in diligence under tight timelines and often incomplete data, all while navigating rising expectations for value creation. Traditionally, this process has demanded weeks of painstaking manual effort involving the collection and synthesis of public data, mining seller data rooms, expert consultations, and benchmarking. However, the advent of generative AI (gen AI) is transforming this landscape by enabling faster insight generation, broader analytic scope, and sharper strategic clarity.
According to a detailed analysis by McKinsey & Company, gen AI tools can now synthesize vast quantities of both public and proprietary data, spot trends and outliers, and even propose hypotheses that human analysts might overlook. Some advanced tools extend their capabilities by providing preliminary insights into the potential value within an asset, significantly shortening the time needed for initial assessments. Despite this promise, many organisations have only recently begun integrating gen AI into their diligence processes, and its implementation remains uneven with significant operational hurdles to overcome.
A core insight from McKinsey’s experience supporting gen-AI-enabled diligence across various domains—ranging from public company transformations to private equity assessments—is that deploying gen AI effectively requires more than plugging in a tool. Success depends on properly training AI models, crafting precise prompts, integrating proprietary data, and maintaining experienced human judgment throughout. The firm outlines five strategies to elevate diligence efforts using gen AI:
-
Customising Models with Proprietary Data
The real power of gen AI emerges when companies leverage their unique proprietary information. Off-the-shelf models may suffice for generic questions, but models trained on organisations’ own data—such as customer transaction histories, prior merger synergy realisations, and operational metrics—can unlock substantial competitive advantages. For instance, a SaaS company employed a gen AI model trained on its proprietary customer data to identify underutilised features, predicting higher upsell potential overlooked by competitors. In another example, anonymised datasets from over 1,600 enterprise transformations provided rapid insight for an oil and gas company, compressing lengthy analyses into days. However, activating proprietary data demands overcoming challenges of fragmentation and data hygiene, requiring systematic cleaning, tagging, and security efforts prior to integration. -
Optimising Peer Set and Benchmark Selection
Peer comparison remains fundamental in outside-in diligence, but its execution often teeters on art more than science. Gen AI assists by scanning industry databases, earnings transcripts, patent filings, and even foreign-language sources to build dynamic peer sets grounded not just on surface similarities but deeper operational aspects—customer segments, cost structures, and supply chains. Such dynamic benchmarking can reveal overlooked competitors and market opportunities, as demonstrated by a medtech company that expanded its competitive analysis to Asian market players previously missed. AI also enables rapid iteration across different peer sets to test benchmarking robustness, thus enhancing confidence in conclusions. -
Constructing Effective Prompts Like a Product Manager
Gen AI’s effectiveness hinges on the quality of input prompts. Many users err by treating gen AI like a search engine, issuing short, vague queries that yield poor results. Instead, crafting structured, context-rich prompts specifying roles, constraints, and data sources leads to sharply focused and actionable outputs. For example, a diligence team improved their analysis of customer acquisition cost by framing their query with precise parameters including peer benchmarks, business model, and recent financial data—transforming a generic question into an industry-specific insight package. Developing a shared, reusable prompt library tailored to diligence tasks can institutionalise this approach. -
Building Specialized AI Agents for Focused Tasks
Leading teams are creating purpose-built gen AI agents dedicated to specific diligence functions such as peer selection, market sizing, or synergy assessment. These agents, designed with clear responsibilities, inputs, and constraints, collaborate within integrated workflows to accelerate and refine analyses while mitigating hallucination risks. An example includes a peer selector agent that sifted through extensive documents, feeding outputs to downstream agents to rapidly compose comprehensive investment theses. Identifying high-value repeatable processes and applying targeted agents can streamline complex multi-step diligence activities. -
Treating Gen AI as an Amplifier, Not a Decision-Maker
Despite its eloquence, gen AI can produce confidently flawed outputs if data inputs are poor or misaligned. McKinsey cautions against overreliance without appropriate human oversight. Some diligence teams embed governance layers including rigorous prompt design, audit logging, and system segregation to safeguard data security and ethical compliance. In practice, structured checks have caught issues such as overstated synergies, underscoring the critical need for human review. Building organisational trust in gen AI insights requires transparency about its limitations and institution-wide mandates for risk-based oversight.
For organisations looking to harness gen AI in diligence, McKinsey recommends a phased approach: inventory and prepare proprietary data, codify and test leading prompts, pilot targeted AI agents in high-impact areas, appoint AI champions to steward adoption, and establish disciplined feedback loops for continuous improvement. Embracing these steps signals a broader operating model shift. The new model revolves around the diligence team as orchestrators, data engineers maintaining curated data pipelines, analysts iterating prompt design, knowledge managers capturing best practices, and risk teams enforcing safeguards.
Beyond diligence, McKinsey has demonstrated how gen AI is reshaping broader M&A activities, facilitating quicker target sourcing, more efficient negotiation, and faster integration, all critical to capturing synergies before value erosion occurs. Internally, the firm’s own generative AI platform—Lilli—has streamlined consultants’ workflows, helping them save time while improving insight quality, though its development encountered significant technical challenges and learning curves alike.
Other applications of gen AI span product portfolio optimisation, where companies leverage the technology to automate complex analysis, improving margins and revenue growth through smarter portfolio decisions. Yet across domains, the consistent theme remains: gen AI requires disciplined integration, ongoing validation, and a partnership with human expertise to unlock its transformative potential safely and effectively.
In summary, generative AI is poised to revolutionise outside-in diligence by reducing manual effort, accelerating timelines, and enhancing the depth and breadth of insight. Companies that swiftly adapt—training models on proprietary information, perfecting prompts, embedding governance, and reimagining analysts’ roles—will gain superior agility and confidence in their investment decisions. This emerging landscape demands both technological acumen and organisational readiness to realise gen AI’s full promise in creating strategic advantage.
Source: Noah Wire Services
Subscribe to Industry Updates
