AI Voice for Private Equity: Cross-Portfolio Value Creation

Private equity value creation in 2026 has a new operational lever: voice AI deployed across portfolio companies at fund level, not portco level. The economics are straightforward — a single shared voice AI platform serving five portfolio companies costs a fraction of five independent vendor contracts, and the EBITDA uplift attributed at portco level improves the fund's multiple in ways that cost rationalisation and headcount reduction alone cannot replicate cleanly.

What makes this a 2026 thesis rather than a 2023 experiment is the maturity of enterprise voice AI platforms and the availability of a credible attribution framework. McKinsey's "State of AI 2025" found that only 6% of enterprises capture material EBIT impact from AI. PE portfolio companies, precisely because fund-level operational mandates can compress the governance and procurement timeline, are positioned to be in that 6% — if the deployment architecture is built for multi-portco economics from the start.

This post covers the two deployment models (centralised platform vs per-portco), the fund-level governance structure, the EBITDA attribution framework, common use cases across portfolio company types, and the diligence lens PE deal teams should apply to target companies with high call-centre exposure.

This analysis is published by the team behind Dilr Voice — an enterprise voice AI platform deployed across multi-entity organisations in 40+ countries — and DATS, our AI consulting and placement practice that works with PE firms and portfolio operators on deployment sequencing and EBITDA attribution.

Why PE Firms Are Making Cross-Portfolio AI Voice a Value Creation Mandate

The thesis is grounded in a structural feature of PE portfolios: many portfolio companies in the same fund share a customer-operations problem. BPO businesses, financial services operators, healthcare administration firms, logistics companies, and SaaS businesses with high-touch customer success functions all share an oversized cost-to-serve in their contact centre or outbound sales function. Voice AI directly addresses that cost line.

What is new in 2026 is that voice AI platforms have matured past the point where only well-resourced enterprise IT teams can deploy them. A single fund-level deployment — one platform vendor, one commercial relationship, one governance wrapper — can be extended to five or ten portfolio companies with portco-specific configurations. The shared compliance framework, shared telephony infrastructure, and shared analytics layer reduce the per-portco deployment cost and timeline significantly compared to five independent implementations.

The procurement model also changes under a fund mandate. Where a portco-led purchase might take 4–9 months through procurement, legal, IT, and compliance, a fund-level mandate with pre-negotiated terms reduces the portco's implementation timeline to the configuration and integration layer — typically 8–12 weeks per portco. That compression matters enormously when a fund has a 4–5 year hold period and wants to show EBITDA improvement by Year 2.

There is a second structural argument. PE portfolio companies often have a knowledge disadvantage relative to large enterprise buyers: they may not have the internal AI expertise to evaluate voice platforms, negotiate commercial terms, or govern a deployment once it is live. A fund-level approach — where operational partners or a specialist advisor sets the platform selection criteria, the governance standards, and the EBITDA attribution methodology — transfers that expertise across the portfolio without each portco having to build it independently.

Cross-referencing the BCG "Widening AI Value Gap" (2025): the firms capturing 2.5× EBIT from AI are Future-Built organisations that have defined governance, repeatable deployment patterns, and portfolio-level investment in AI capability. That description maps precisely to what a well-run PE fund can create for its portfolio companies in 12–18 months.

The Two Deployment Models: Centralised Platform vs Per-PortCo

Every PE firm faces the same initial architecture decision: does the fund operate a single shared voice AI platform that all portfolio companies access, or does each portco make its own platform decision within a fund-defined framework?

Model 1: Centralised Fund Platform

The fund negotiates a single commercial agreement with a voice AI vendor. Portfolio companies access the platform as sub-accounts or tenant instances within a multi-tenant architecture. The fund's operational partners configure governance standards, integration templates, and compliance guardrails at the platform level. Each portco configures its own scripts, personas, and integrations within that shared environment.

Advantages: lowest total cost of ownership (single commercial negotiation, volume discount, shared infrastructure), fastest per-portco deployment (governance framework already exists), easiest cross-portfolio analytics (same data schema, same reporting layer), and strongest negotiating leverage with the vendor (total contract value is the fund's aggregate, not any single portco's).

Disadvantages: portcos with complex regulatory environments (e.g., FCA-regulated financial services entities) may have data-residency or isolation requirements that conflict with a shared tenancy model. Some portcos may have legacy telephony infrastructure that is difficult to integrate with a shared platform. Exit complexity: when a portco is sold, its voice AI dependency on a fund-level platform must be resolved — either the portco buys its own license, the platform supports a clean handover, or there is a transition period. The MSA clause set governing portco separation must be negotiated upfront.

Model 2: Per-PortCo Deployment with Fund-Level Standards

Each portfolio company makes its own vendor selection, but the fund defines the evaluation criteria, the compliance standards, the SLA requirements, and the EBITDA attribution methodology that all portcos must meet. The fund's operational partners may recommend a shortlist of preferred vendors but do not mandate a single platform.

Advantages: portcos with specific regulatory, integration, or operational requirements can choose the best fit for their context. Exit is clean — each portco owns its own vendor relationship. Portcos build internal AI capability rather than depending on a fund-level resource.

Disadvantages: higher aggregate cost (no volume discount, each portco negotiates independently), slower per-portco deployment (each portco runs its own procurement cycle), and less visibility for the fund across portfolio company performance (different data schemas, different reporting cadences).

The Hybrid Model (Most Common in Practice)

Most PE firms with serious voice AI mandates land on a hybrid: a preferred-vendor framework negotiated by the fund (with volume pricing and standard legal terms), portcos choosing within that framework (or justifying exceptions to an operational partner review), and fund-level governance standards that all portcos must meet regardless of vendor. This gives portcos enough autonomy to accommodate their specific context while the fund captures most of the commercial leverage and governance efficiency of a centralised model.

The build vs vendor decision at portco level is almost always resolved in favour of vendor within a PE context — custom-build timelines are incompatible with a fund's hold period and value creation timeline.

Centralised Platform Economics: How the Cost Model Actually Works

The economics of a centralised fund platform depend on three variables: the number of portfolio companies, the aggregate call volume across those companies, and the per-minute or per-resolution pricing model the fund negotiates.

A fund with five portfolio companies each handling 50,000 voice interactions per month has an aggregate volume of 250,000 interactions. At typical per-minute pricing of $0.07–$0.12/minute (industry range for enterprise voice AI, per the voice AI TCO analysis), the aggregate spend at list price is $17,500–$30,000 per month. A fund negotiating a single commercial agreement at that aggregate volume should expect 20–35% volume discount relative to five independent contracts at portco-level volumes. That discount alone represents £36,000–£84,000 per year in aggregate savings across the portfolio — before any implementation cost savings are counted.

The implementation cost saving is larger. A single governance framework, a single data processing agreement, a single security audit, and a single legal review — shared across five portcos — eliminates 80% of the legal and compliance cost that five independent portco purchases would incur. At typical enterprise voice AI procurement legal costs of £15,000–£40,000 per contract, the fund-level approach saves £60,000–£160,000 in procurement overhead alone.

The total cost of ownership picture is further improved by shared analytics infrastructure. Five portcos with independent platforms produce five separate reporting systems with incompatible schemas. A fund-level platform produces a single analytics layer that operational partners can use to identify cross-portco performance patterns — which portcos are running which scripts, which containment rates are achievable, and which portco implementations are underperforming relative to portfolio peers. That visibility is operationally valuable and has no equivalent in a per-portco model.

One cost line the fund-level model introduces that per-portco does not: the operational partner resource required to govern the shared platform. Someone at fund or advisory level must own the platform relationship, manage configuration access controls, govern the compliance framework, and ensure portco-level configurations do not create fund-level regulatory exposure. For most mid-market PE funds this is a 0.25–0.5 FTE operational partner function, which is easily justified against the procurement and analytics savings above.

The AI voice ROI framework provides a starting point for modelling per-portco savings, but PE operational partners need a portfolio-level ROI model that aggregates those calculations across hold-period assumptions and weights them by portco revenue contribution.

Fund-Level vs PortCo-Level Governance

The governance question in PE voice AI is not just about who owns the vendor relationship. It is about who is accountable for regulatory compliance, who resolves conflicts between portco operational needs and fund-level standards, and how the governance model handles portco exit.

Fund-Level Governance Responsibilities

At fund level, governance should cover: platform vendor selection and commercial terms; the compliance framework that all portcos must meet (GDPR data residency, FCA Consumer Duty where applicable, sector-specific obligations such as FCA for financial services portcos or CQC for healthcare portcos); minimum SLA standards; the EBITDA attribution methodology; and the cross-portfolio analytics governance (what data can be shared across portcos vs what must remain portco-isolated).

The fund should maintain a cross-portfolio AI register — a cross-regulator AI tool inventory is now expected by ICO, FCA, and EU AI Act for any enterprise deploying AI in customer-facing roles. For a PE fund operating regulated portfolio companies, this register must exist at portco level and be consolidated at fund level for LP reporting and regulatory diligence.

PortCo-Level Governance Responsibilities

Each portfolio company retains accountability for: the operational configuration of the voice AI within their specific business context; their sector-specific regulatory compliance (a debt recovery portco owns its FCA Consumer Duty compliance, an NHS subcontractor portco owns its DCB0129 clinical safety obligations); their integration with their own systems (CRM, telephony, ticketing); and their customer-facing disclosure and consent architecture.

Portco governance should include a steering group that meets at least monthly, with attendance from Operations, IT/Systems, Compliance, and Finance (for EBITDA tracking). The fund's operational partner should hold observer or advisor status on that steering group, especially in the first 12 months. The cross-portfolio AI voice governance framework provides the structural starting point — PE funds should adapt it to include fund-level escalation paths and cross-portco peer review.

Exit Planning

Governance of portco exit from the shared platform must be defined upfront. The key questions: does the portco purchase its own vendor license on exit (and at what price), or is the portco's voice AI programme migrated to an alternative? Who owns the call data and transcripts produced on the shared platform — the portco or the fund? What is the transition assistance obligation if the platform is central to the portco's customer operations?

These questions are much easier to resolve at the outset than during a sale process, when both the fund and the incoming buyer are under time pressure. The MSA with the voice AI vendor should include explicit portco-exit provisions, and the fund's governance framework should define the internal policy. The vendor exit and offboarding analysis covers the technical portability questions in detail.

EBITDA Attribution: The Framework That Gets It Onto the Contribution Statement

The most important governance document in a PE voice AI deployment is the EBITDA attribution framework — the agreed methodology for measuring and reporting the EBITDA impact of the voice AI programme at portco and fund level.

Without a pre-agreed attribution framework, voice AI savings get lost in operational variance, absorbed into headcount reductions that are attributed to restructuring, or disputed between the portco CFO and the fund's operational partners at board level. With a pre-agreed framework, the contribution statement is clean, the IRR attribution is defensible, and the sale process narrative includes a verified EBITDA add-back.

What the Attribution Framework Must Cover

1. Baseline definition. Before the voice AI programme goes live, the portco must lock a baseline cost-to-serve: total cost of the call centre function (salaries, benefits, seat costs, telephony, QA, training) divided by total handled call volume. This baseline is the measurement datum. Any reduction in cost-to-serve after go-live is attributable to the programme, net of platform costs.

2. Attribution scope. The framework must define which cost lines are in scope for attribution (direct call centre costs, yes; management overhead, usually no; real estate, only if headcount reductions trigger lease events). It must also define which revenue impacts are in scope — for portcos using AI voice for outbound sales or collections, revenue uplift from improved contact rates is a legitimate attribution item, but requires a separate methodology (contact-rate uplift × historical conversion rate × average order value).

3. Containment rate tracking. The primary operational metric is containment rate — the percentage of calls fully handled by the AI agent without escalation to a human. Every percentage point of containment rate improvement represents a reduction in handled-call volume for human agents. The framework should define how containment rate is measured (at call level, not call-minute level), what threshold constitutes a "contained" call (call concluded without escalation and customer did not call back within 24 hours), and how the metric is reported to the fund.

4. Headcount bridge. If the programme results in headcount reduction (or avoids headcount additions in a growing portco), the framework must define the cost per FTE avoided — fully loaded, including recruitment cost, training cost, and benefits — and the methodology for attributing FTE avoidance to the voice AI programme rather than other operational changes. Most PE funds accept a 70/30 attribution split on FTE reductions in mixed-cause periods (70% to voice AI, 30% to operational factors), adjusted by evidence from the containment rate data.

5. Audit trail. The framework must define the evidence pack required to support the attribution at sale diligence: call volume data pre and post go-live, containment rate trends, cost-to-serve comparison, and the platform cost incurred. The voice AI auditability framework provides the logging and audit architecture that produces this evidence pack as a natural output of the platform.

6. Hold-period projection. At the point of fund-level investment decision, the framework should include a 3–5 year EBITDA projection: Year 1 (go-live and ramp, partial-year savings), Year 2 (full-run-rate containment, defined savings), Year 3+ (expansion to additional use cases, compounding improvement as conversation design matures). The programme expansion playbook provides the operational model for Year 2 expansion.

Common Portfolio Company Use Cases by Sector

Not all portfolio company types have the same voice AI opportunity. The use case and ROI profile vary significantly by sector, and the fund-level value creation thesis should be calibrated accordingly.

BPO and Business Services

BPO portfolio companies are the highest-ROI voice AI opportunity in a PE portfolio. Their entire revenue model is based on providing call-handling services to third-party clients, and voice AI converts their cost-of-delivery while their client contracts are typically structured around outcomes (handled calls, resolution rates) rather than inputs (agent hours). An AI-augmented BPO can absorb volume growth without headcount growth and improve margins on existing contracts simultaneously. The risk is client contract terms — many BPO MSAs have provisions requiring client consent for automation of handling functions.

Financial Services (Lenders, Servicers, Insurers)

Financial services portcos — consumer lenders, mortgage servicers, insurance claims operations, debt purchasers — have high call volumes, structured call types (collections, claim intake, balance enquiry, account servicing), and a mature FCA regulatory framework to navigate. The compliance architecture is more demanding (Consumer Duty, FCA complaint handling rules, vulnerable customer detection), but the ROI is strong because the cost-per-call baseline is high and the call types are well-suited to automation. The fintech collections and KYC analysis provides the compliance framework for financial services voice AI deployment.

Healthcare and Social Care

Healthcare portcos — GP management organisations, community pharmacy chains, occupational health providers, dental chains — have high appointment-scheduling, recall, and triage call volumes. The ROI is clear (appointment fill rates, no-show reduction), the regulation is CQC and ICO/NHS IG rather than FCA, and the patient-sensitivity overlay requires more careful conversation design than a commercial context. Fund-level governance in healthcare must include a clinical safety framework (DCB0129 where applicable) and a patient consent architecture.

SaaS and Technology

SaaS portfolio companies with high-touch customer success or inside sales functions use voice AI for outbound renewal calls, expansion conversation initiation, and churn-prevention check-ins. The ROI model is different from cost reduction — it is revenue retention and expansion (NRR improvement). Attribution requires connecting voice AI contact activity to NRR outcomes, which requires a 3–6 month data series before the attribution is statistically valid.

Logistics and Field Services

Logistics and field service portcos use voice AI for dispatch notifications, delivery confirmation, slot booking, and exception management. High call volume, highly structured interactions, and a clear cost-per-handled-interaction baseline make these strong ROI candidates. The fund-level governance consideration is that logistics portcos often operate across multiple time zones and language markets — the cross-portfolio platform must support multilingual deployment.

Due Diligence Through a Voice AI Lens

PE deal teams conducting diligence on acquisition targets with significant call-centre operations should assess voice AI automation potential as a value creation line item alongside the standard operational improvement thesis. The assessment has three components.

Call Volume and Mix Analysis

The first diligence question is: what is the target's handled call volume, and what percentage of that volume is automatable? Automation potential correlates with call type structure. Highly structured calls (appointment booking, balance enquiry, status update, collections outreach) typically have 60–80% automation potential. Less structured calls (complex complaints, advisory conversations, high-emotion interactions) have 0–20% automation potential. A target with 80% of its call volume in structured call types is a strong automation candidate; a target with 80% advisory or complaints volume is a weaker one.

The KPI framework for voice AI programmes provides the analytical structure for this assessment: ask the target to produce call type distribution data, average handle time by call type, and first-call resolution rates by call type. Those three data points are sufficient for an initial automation potential assessment.

Technology and Integration Readiness

The second diligence question is: what is the target's telephony and CRM environment, and how difficult is it to integrate? Targets on Salesforce, Dynamics 365, or standard cloud telephony platforms (Twilio, RingCentral, Genesys Cloud) are much faster and cheaper to integrate than targets with on-premise telephony, proprietary CRM systems, or telephony that is deeply embedded in their operational workflows.

Integration complexity is the primary driver of timeline variance in voice AI deployments. A target with modern cloud infrastructure can be live in 8–12 weeks; a target with on-premise telephony and a bespoke CRM may take 6–9 months just for the integration layer. The budget and integration ownership analysis is useful for identifying who inside the target organisation controls the technology stack decisions that will determine deployment speed.

Regulatory Exposure

The third diligence question is: what is the target's regulatory environment, and does it create any voice AI deployment constraints that are not obvious from the automation potential analysis? FCA-regulated targets require Consumer Duty compliance architecture. Healthcare targets require clinical safety frameworks. EU-facing targets require Article 50 AI Act disclosure architecture by August 2026 (for new deployments) or December 2026 (for synthetic audio marking). The fund-level compliance framework should define minimum standards across all of these — but diligence should flag targets where the regulatory environment significantly increases the compliance build cost or timeline.

Implementation Sequencing: Which PortCo First

The sequence in which portfolio companies go live on the fund's voice AI platform materially affects the total value created and the governance effort required.

The first portco to go live should be selected for proof-of-model, not for maximum EBITDA potential. Choose the portco with the simplest integration environment (cloud telephony, standard CRM), the most clearly structured call types, and a management team that is genuinely bought in to the programme. A successful first deployment — clean go-live, visible containment rate improvement, attributable cost reduction — provides the evidence that convinces the more sceptical portco management teams and creates a replicable implementation template.

The second and third portcos should be sequenced by integration readiness, not by EBITDA potential. A portco with 10× the savings potential but 12 months of integration complexity should sit behind a portco with 3× the savings potential but 8 weeks of integration. The fund's hold-period arithmetic typically favours speed over scale in the first 18 months.

The highest-complexity portcos — those with on-premise telephony, bespoke CRM, or regulated data environments requiring isolated tenancy — should go last. By the time they are in deployment, the fund-level governance framework is established, the operational partner team has deployment experience, and the voice AI vendor relationship is strong enough to allocate dedicated professional services resource to a complex implementation.

Portfolio-level sequencing should be reviewed at the quarterly operational partner meeting, with updates to the projected EBITDA contribution timeline at each review. The pilot purgatory analysis identifies the structural failure modes that cause PE voice AI initiatives to stall between portco deployments — most are governance failures (no owner, no exit criteria, no cross-portco steering) rather than technical failures.

A PE fund running this well should expect the following timeline: Quarter 1 (platform selection, fund-level governance framework, first portco scoping); Quarter 2 (first portco go-live, second portco scoping, attribution baseline locked); Quarter 3 (first portco at run-rate, second portco go-live, third portco scoping); Year 2 (full portfolio on platform, EBITDA attribution visible in portco financials, narrative ready for sale process or LP reporting).

Want to see the economics in action? Try Dilr Voice on a live portco use case (free, $20 credits), get an AI placement diagnostic for the highest-readiness portco, or speak to DATS about a fund-level deployment framework.

Written by the Dilr.ai engineering team — practitioners who ship enterprise AI in production. Follow us on LinkedIn for shipping notes, or subscribe via the RSS feed.