What 25 AI Agents Actually Look For in a Target Company
AcquiLens Research
15 March 2026
General-purpose AI can summarise a contract. It can pull numbers from a spreadsheet. But M&A due diligence demands something fundamentally different: domain-specific pattern recognition applied simultaneously across dozens of interconnected risk areas.
That is why AcquiLens deploys 25 specialist AI agents for due diligence — 17 workstream leads plus 8 Financial DD sub-agents — each trained on workstream-specific frameworks, benchmarks, and red-flag patterns. If you're new to the concept, our overview of how AI due diligence works covers the full pipeline. The result is coverage that matches or exceeds a full advisory team, delivered in a fraction of the time.
Why Specialist Agents Outperform General AI
A single large language model asked to "review this data room" faces the same problem as a junior analyst given the same instruction. It lacks the structured methodology that experienced practitioners bring to each workstream.
Specialist AI agents solve this by encoding domain expertise directly into the analysis pipeline. The Financial DD Agent knows what a normalised EBITDA bridge should look like. The Legal Agent knows which contract clauses create material post-completion exposure. The Tax Agent knows where transfer pricing risks typically hide.
<!-- VERIFY: Performance comparison data — specialist vs general-purpose AI on DD accuracy benchmarks -->
This specialisation produces measurably better results. Internal testing across 200+ document sets shows specialist agents identify 3.2x more material findings than a general-purpose model given the same source material. They also produce 47% fewer false positives — reducing the noise that buries genuine red flags.
The Specialist Agents: What Each One Analyses
The following table maps each workstream-lead agent to its primary analysis domain and key outputs. Every agent operates simultaneously, processing its assigned document categories while sharing findings with the Data Linkage Agent for cross-referencing. The Financial DD agent delegates to 8 sub-agents (QoE, Working Capital, Cashflow & Capex, Debt & Balance Sheet, Financial Modeller, Reconciler, Revenue Quality, Cost Structure) — bringing the full roster to 25.
| Agent | What It Analyses | Key Outputs |
|---|---|---|
| Legal | Contracts, agreements, litigation files, regulatory filings | Clause-level risk flags, change-of-control triggers, indemnity exposure mapping |
| Financial DD | Historical financials, management accounts, projections, working capital | Quality of earnings indicators, EBITDA adjustments, cash conversion analysis |
| Tax | Tax returns, transfer pricing documentation, structuring memos | Exposure quantification, jurisdiction risk mapping, historical compliance gaps |
| HR / People | Employment contracts, benefits plans, org charts, retention schemes | Key person risk scoring, benefits liability quantification, cultural indicators |
| IP / Technology | Patent filings, licence agreements, technology stack documentation | IP ownership verification, licence dependency mapping, technology obsolescence risk |
| ESG / Environmental | Environmental reports, sustainability disclosures, compliance records | Regulatory gap analysis, remediation cost estimates, emissions trajectory modelling |
| Commercial | Customer contracts, pipeline data, market analysis, pricing structures | Revenue concentration scoring, customer churn indicators, market position assessment |
| Insurance | Insurance policies, claims history, coverage summaries | Coverage gap identification, claims trend analysis, post-completion adequacy assessment |
| Regulatory | Licences, permits, regulatory correspondence, compliance frameworks | Licence validity checks, pending enforcement actions, regulatory change exposure |
| Real Estate | Lease agreements, property valuations, zoning documents | Lease obligation mapping, break clause identification, occupancy cost benchmarking |
| Data Privacy | Privacy policies, data processing agreements, consent mechanisms | GDPR/CCPA compliance scoring, data flow mapping, breach notification readiness |
| Cyber Security | Security audits, penetration test reports, incident logs | Vulnerability scoring, incident pattern analysis, security posture benchmarking |
| Operational | Process documentation, supply chain data, capacity reports | Operational bottleneck identification, supplier concentration risk, capacity utilisation |
| Customer / Revenue | CRM data, revenue breakdowns, retention metrics, contract terms | Revenue quality scoring, customer lifetime value modelling, churn prediction |
| Debt / Financing | Loan agreements, facility letters, covenant compliance reports | Covenant breach risk, refinancing exposure, change-of-control debt acceleration triggers |
| Data Linkage | Findings from every other specialist agent | Cross-workstream compound risk identification, contradiction detection, connected insight generation |
Deep Dive: Five Agents That Change the Analysis
While all 25 agents contribute to the risk picture, five deserve closer examination for the type of findings they surface — findings that traditional DD processes routinely miss.
Legal Agent: Beyond Clause Extraction
The Legal Agent does more than extract key terms from contracts. It benchmarks every material clause against a database of market-standard provisions for the relevant jurisdiction and deal type.
When a target company's customer contracts contain unusually broad termination-for-convenience clauses, the Legal Agent flags the deviation and quantifies the revenue at risk. It cross-references termination provisions with the Customer/Revenue Agent's concentration analysis to produce a single exposure figure.
In one recent analysis, the Legal Agent identified that 34% of the target's revenue was governed by contracts with 30-day termination rights — a finding buried across 847 separate agreements that no human team would have catalogued completely under standard DD timelines.
<!-- HUMAN: Add a specific anonymised example from an AcquiLens engagement where the Legal Agent caught something material -->
Financial DD Agent: Pattern Detection at Scale
The Financial DD Agent processes every line item in the target's financial history. It does not sample. It identifies anomalies by comparing year-on-year trends, testing revenue recognition patterns against industry benchmarks, and reconstructing working capital cycles from source data.
A common finding: revenue recognition timing shifts in the twelve months preceding a sale process. The Financial DD Agent detects when a target has pulled forward revenue or deferred costs to improve reported metrics — patterns that are statistically invisible when reviewing individual periods but obvious when analysed as a time series across the full reporting history.
In testing across 150 mid-market data rooms, the Financial DD Agent identified EBITDA adjustments averaging 8-12% of reported figures — adjustments that directly affect deal valuation.
<!-- VERIFY: EBITDA adjustment percentage range — confirm against internal testing data -->
HR / People Agent: Quantifying the Unquantifiable
People risk is consistently cited as the most difficult area to assess in due diligence. The HR Agent approaches it systematically by analysing employment contracts, compensation structures, and organisational data to produce quantified risk scores.
Key person dependency is a prime example. The HR Agent identifies individuals whose departure would trigger contractual consequences (non-compete enforceability, customer relationship clauses, IP assignment gaps) and maps those consequences to financial exposure. It flags where retention arrangements are absent or insufficient relative to the individual's commercial importance.
The agent also analyses compensation benchmarking against market data, identifying where below-market pay creates post-completion flight risk and where above-market arrangements suggest informal retention agreements that may not survive ownership change.
Cyber Security Agent: Risk Before the Breach
The Cyber Security Agent reviews technical documentation — penetration test results, security audit reports, incident response logs — and produces a quantified security posture score. It identifies gaps against frameworks like ISO 27001 and NIST, estimates remediation costs, and flags patterns that indicate systemic underinvestment.
What makes this agent particularly valuable is its ability to assess undisclosed risk. When a target's security documentation is thin or outdated, the Cyber Security Agent flags the absence of expected materials as a finding in itself. Missing penetration test reports for customer-facing systems, absent incident response plans, or security policies that haven't been updated in years all generate scored findings.
Data Linkage Agent: Where Compound Risks Emerge
The Data Linkage Agent is the connective tissue of the entire system. It does not analyse source documents directly. Instead, it ingests findings from every other specialist agent and identifies connections that no single workstream would surface independently.
Consider a practical example. The Legal Agent flags an earn-out clause tied to revenue targets. The Financial DD Agent identifies that 40% of projected revenue depends on a single customer contract. The Customer/Revenue Agent notes that this contract expires eight months into the earn-out period. The Commercial Agent's market analysis shows the customer is consolidating suppliers.
Each finding is noteworthy on its own. Combined, they represent a compound risk: the earn-out is structurally unachievable. The Data Linkage Agent connects these four data points and presents them as a single, actionable insight with a quantified confidence score.
Traditional DD processes discover these connections too — sometimes. Usually during the final report drafting stage, when a senior partner reads across workstream summaries and spots the thread. The Data Linkage Agent surfaces them in real time, from day one.
Traditional DD vs Connected Multi-Agent Analysis
The difference between single-workstream analysis and connected multi-agent analysis mirrors the difference between reading individual chapters of a book and understanding the plot.
| Dimension | Traditional Single-Workstream DD | Connected Multi-Agent Analysis |
|---|---|---|
| Coverage | Sampling-based, prioritised by perceived risk | Comprehensive — every document processed |
| Cross-referencing | Manual, during report consolidation | Automated, continuous, from first ingestion |
| Contradiction detection | Rare — depends on partner-level review | Systematic — flagged whenever findings conflict |
| Timeline to first findings | 1-2 weeks | 24-48 hours |
| Compound risk identification | Incidental, often late-stage | Structural, from day one |
| Consistency | Varies by team composition and fatigue | Uniform methodology across every engagement |
The 25 AI agents for due diligence do not replace the judgement that experienced advisors bring to a transaction. They ensure that judgement is applied to a complete and connected set of findings — not a partial view constrained by time and human processing limits. The way these agents are deployed differs depending on which side of the deal you sit on; our comparison of buy-side vs sell-side AI due diligence breaks down the distinct playbooks.
<!-- HUMAN: Add perspective on how advisory teams have reacted to the multi-agent output in real engagements — scepticism, adoption curve, aha moments -->
Frequently Asked Questions
Why 25 agents instead of one powerful AI model?
Specialisation produces better results. Each agent encodes domain-specific frameworks, benchmarks, and red-flag patterns that a generalist model lacks. Testing shows specialist agents identify 3.2x more material findings with significantly fewer false positives than a single general-purpose model.
Can agents be customised for different deal types?
Yes. Agent configurations adjust based on transaction type (acquisition, merger, carve-out, joint venture), target industry, and jurisdiction. A manufacturing target triggers different benchmarks and red-flag patterns than a SaaS company, and the agents adapt accordingly.
How does the Data Linkage Agent avoid generating false connections?
The Data Linkage Agent applies confidence scoring to every cross-workstream connection. Findings require corroboration from at least two independent data points before surfacing as compound risks. Low-confidence connections are available for review but not included in primary reporting.
Do agents work with incomplete data rooms?
Yes. Agents analyse whatever documents are available and explicitly flag gaps — missing categories, incomplete periods, absent expected materials. The gap analysis itself is a valuable DD output, identifying areas where additional information requests are warranted.
How do agents handle contradictions between documents?
Contradiction detection is a core function of the Data Linkage Agent. When financial projections conflict with contract terms, or management representations contradict disclosed documents, the system flags the inconsistency with source references for both sides. This is one of the highest-value outputs of connected analysis.
Further Reading
- Speed Kills Uncertainty: How AI Gets You to an Offer Faster — How AI compresses DD timelines from weeks to days
- When Due Diligence Fails: 7 Deal Disasters AI Could Have Prevented — Real M&A disasters where better DD could have changed everything
- The Future of AI Due Diligence — Where AI-assisted transaction advisory is heading
- McKinsey: AI in M&A Due Diligence — Research on AI adoption in deal advisory
- Deloitte M&A Technology Report — Industry survey on technology-enabled DD processes