Building Cross-Portfolio AI Centers of Excellence in MENA Private Equity

The Structural Problem Behind Scattered AI Experiments

Private equity firms operating across a diversified portfolio face a specific kind of AI failure that rarely gets named correctly. Individual portfolio companies run pilots, engage vendors, and report promising results — yet two years into the effort, no shared intelligence exists, no tooling transfers between assets, and the GP cannot articulate what AI actually contributes to portfolio value. This is the fragmentation trap, and it is more common than most MENA PE firms will admit publicly.

The antidote is not to mandate more pilots. The antidote is to build institutional AI capacity at the fund level — a cross-portfolio AI center of excellence that acts as a production organ, not a research committee.

Why the MENA Context Requires Its Own Playbook

The question of how MENA PE firms build cross-portfolio AI centers of excellence cannot be answered by lifting frameworks from US or European private equity. The MENA operating environment has distinct regulatory jurisdictions across the UAE, Saudi Arabia, Qatar, Bahrain, Kuwait, and Oman, each with evolving data localization and AI governance policies that vary meaningfully. An AI operating model designed for a single jurisdiction will create compliance exposure the moment a cross-border portfolio company is included.

Beyond regulation, MENA portfolios tend to be operationally diverse in ways that challenge a one-size model. A single fund might hold assets in healthcare, logistics, hospitality, and financial services simultaneously. Each vertical has different data structures, different latency tolerances, and different ROI measurement cycles — factors that shape what agents can be shared and what must be built bespoke for each asset.

Arabic-language requirements add a layer of complexity that most Western AI stacks handle poorly. Many enterprise AI tools are English-first by design, which means portfolio companies serving Arabic-speaking customers operate with degraded accuracy in any shared AI layer that was not designed with language parity from the start.

Defining the Center of Excellence: What It Actually Is

A cross-portfolio AI center of excellence is not a consulting team that advises portfolio companies on AI strategy. That model produces decks, not deployments. A properly structured CoE is an operational unit that owns production AI infrastructure, maintains it, evolves it, and redeploys components across portfolio assets on defined timelines.

The CoE has four core functions: capability development, which means building agents and integrations that can serve multiple portfolio companies; deployment governance, which means setting technical and data standards every asset must meet before an agent is activated; ROI measurement, which means maintaining a shared analytics layer that quantifies what each deployment contributes to portfolio value; and talent concentration, which means consolidating AI engineering capacity at the fund level rather than duplicating it imperfectly at every portco.

Each function requires a named owner and a governance mechanism. Funds that treat the CoE as a committee without operational authority will find that portfolio company operating teams override decisions the moment there is friction. Operational authority must be written into management company agreements and investment mandates, not assumed.

Mapping Portfolio Verticals to Shared Agent Architecture

The first methodological step when standing up a cross-portfolio CoE is to map every portfolio company to a capability taxonomy. This is not a technology audit — it is a workflow audit. The question being answered is not "what software do they use?" but "what decisions are being made repeatedly by humans that could be made autonomously by a trained agent?"

Typical MENA portfolio companies surface four to eight high-frequency decision categories in this mapping: procurement approval routing, revenue exception handling, customer escalation triage, compliance document review, demand forecasting, and shift or capacity scheduling. These categories appear across verticals even when the underlying business models look nothing alike. A hospitality asset and a healthcare asset both make daily capacity decisions — the data structures differ but the agent architecture can share a common reasoning backbone.

Once the decision taxonomy is complete, the CoE team clusters portfolio companies by decision similarity rather than by industry. This clustering becomes the foundation of the shared agent roadmap. Agents built for the highest-density clusters generate the greatest cross-portfolio ROI because development cost is amortized across the most assets.

The clustering exercise also reveals where bespoke builds are unavoidable. Certain verticals — particularly regulated financial services or clinical healthcare — will have compliance requirements that make full agent sharing impractical. Identifying these early prevents over-promising to the GP on reuse rates. For a deeper look at how these industry-specific requirements manifest, the methodology underlying AI deployment for transaction diligence in advisory firms is directly applicable to this vertical isolation analysis. (See: AI Deployment for Transaction Diligence in MENA Advisory Firms)

Governance Architecture: Who Decides What

Governance is where most cross-portfolio AI programs fail. Without a clear decision rights framework, portfolio company CEOs resist CoE mandates, IT teams at individual assets ignore integration standards, and the CoE team spends most of its time in escalation conversations rather than building. A functioning governance model distributes authority across three tiers.

The first tier is the GP-level AI investment committee. This committee sets strategy, approves the agent roadmap, allocates CoE budget, and reviews ROI measurement on a quarterly basis. It does not make technical decisions. Its role is to ensure the CoE stays aligned with the fund's value creation thesis and that AI investment is treated as a portfolio-level capital allocation decision.

The second tier is the CoE operating team. This team makes all technical decisions: agent architecture, integration standards, data governance policies, deployment timelines, and vendor selection. It has authority to require portfolio companies to meet data quality and integration standards as a condition of receiving CoE-built agents. This authority must be explicit, not aspirational.

The third tier is the portfolio company AI liaison. Each portco designates one senior operator — typically from operations or technology leadership — who is accountable to both the portco CEO and the CoE operating team. This person is responsible for internal adoption, surfacing workflow constraints that affect agent performance, and managing the local data environment to CoE standards.

Funding the CoE: Allocation Models That Work

How the CoE is funded determines its independence and its staying power. Three allocation models are in common use across institutional PE: management company funding, portco assessment, and carry-linked investment.

Management company funding is the cleanest model for early-stage CoEs. The management company treats the CoE as an operational overhead investment, similar to how it funds finance, legal, and deal team infrastructure. This model preserves portco P&L from CoE charges during the build phase, which reduces resistance from portfolio company operating teams. The limitation is that it caps CoE scale at what the management company is willing to expense directly.

Portco assessment models charge each portfolio company a periodic fee for CoE services, typically scaled by revenue, headcount, or agent count deployed. This model creates a self-funding mechanism and aligns portco operating teams with CoE outputs — they are paying for what they receive. The risk is that portcos with tight EBITDA margins resist assessment charges before they have experienced tangible agent value.

Carry-linked investment treats the CoE as a fund-level investment that earns a return on exit. Portfolio companies that benefit from CoE-built AI systems carry a notional AI value contribution on their balance sheet, which the GP monetizes at exit as part of the overall value creation story. This model is more complex to structure but aligns incentives most directly with the private equity return model. Whichever model is chosen, the GP should set a minimum three-year funding horizon for the CoE. AI infrastructure that compounds intelligence needs time to demonstrate the value that justifies its cost.

Building the Deployment Timeline: From Assessment to Production

The deployment timeline for a cross-portfolio AI CoE should be structured in three phases, each with defined gates that prevent advancing before foundations are in place.

Phase one runs for the first several months and covers portfolio readiness assessment, data environment auditing across all assets, and the build of the CoE's technical infrastructure. By the end of phase one, the CoE should have a functioning data pipeline from at least two portfolio companies, a deployed agent in production at one asset, and a documented integration standard that every other portco must meet before receiving a deployment.

Phase two covers the rollout of shared agents to clustered portfolio companies. The CoE builds the highest-priority shared agents first, activates them at each portco in the cluster, and runs a parallel ROI measurement process that captures the value being generated. This phase typically runs for six to twelve months depending on portfolio size and the complexity of integration work at individual assets. Deployment timelines are driven primarily by data readiness at the portco level — clean, structured, accessible data reduces deployment time significantly.

Phase three is the compounding phase. Agents trained on production data from multiple portfolio companies begin to generate insights that no single portco could produce on its own. Cross-portfolio pattern intelligence emerges: pricing signals observed across hospitality assets inform retail pricing agents, and procurement exception patterns from one logistics company surface risk signals useful to another. This is the phase where the CoE becomes structurally valuable — not just operationally useful.

ROI Measurement: Building the Analytics Framework

ROI measurement for a cross-portfolio AI program must be designed before deployment begins, not retrofitted after agents are running. The most common failure mode is measuring the wrong things: agent uptime, query volume, or user adoption scores that capture activity but not value creation.

A PE-grade analytics framework measures AI contribution along three dimensions. The first is operational efficiency — reductions in labor cost or cycle time attributable to agents replacing or accelerating human decisions. The second is revenue contribution — additional revenue generated by agents making better decisions than humans on pricing, upsell timing, or customer routing. The third is risk avoidance — compliance incidents, fraud events, or operational errors that agent monitoring prevented.

Each dimension requires a baseline measurement taken before agent activation. Without baselines, the CoE cannot demonstrate attribution, and the GP investment committee will struggle to justify continued CoE funding in annual reviews. The analytics layer should be maintained centrally by the CoE and reported to the GP investment committee quarterly, with portco-level breakdowns available to each asset's management team.

It is also worth building a portfolio-level AI value attribution model that contributes to exit preparation. Buyers in MENA private equity transactions are increasingly applying scrutiny to operational AI programs during due diligence. A well-documented AI value story — with traceable analytics, owned infrastructure, and demonstrated ROI measurement history — can influence exit multiples in sectors where operational efficiency is a primary value driver. For further context on how AI analytics create enterprise value in adjacent contexts, see AI Deployment for Engagement Delivery in MENA Management Consulting Firms.

Talent Architecture: Who Runs the CoE

The CoE cannot be staffed with generalists who have an interest in AI. It requires practitioners who have built production AI systems — not people who have managed vendor relationships or run AI strategy projects. The distinction matters because most of the hard problems in a cross-portfolio AI program are engineering problems, not strategy problems.

The minimum CoE team for a mid-sized MENA private equity portfolio — typically five to fifteen portfolio companies — is a technical lead with agentic AI deployment experience, two to three AI engineers capable of building and maintaining production agents, a data engineer who owns the cross-portfolio data infrastructure, and an AI operations analyst who runs the ROI measurement layer.

These roles are difficult to hire in MENA at the moment. The regional supply of practitioners with production agentic AI experience is limited, and competition from sovereign wealth funds, large banks, and technology companies makes it expensive. Most MENA PE firms find that a hybrid model works better in the early years: a small internal team that owns governance, standards, and measurement, supplemented by an external sovereign AI infrastructure partner for the build and deployment work. The internal team develops capability over time as agents are transferred to portco ownership under clearly documented IP terms.

Selecting External Infrastructure Partners

The choice of AI infrastructure partner for a cross-portfolio program has more long-term consequence than almost any other decision the CoE makes. The wrong choice creates vendor lock-in, produces agents the portco cannot modify or own, and builds dependencies that persist through exit — reducing the transferability of AI assets to incoming buyers.

The selection criteria that matter most for PE cross-portfolio deployments are: client ownership of all source code, agents, data, and IP; proven production-grade exception handling rather than demo-environment performance; genuine vertical depth across the industries present in the portfolio; and a pricing model that scales transparently by agent count and integration complexity rather than opaque per-seat or usage charges.

Labarna AI operates as sovereign production intelligence — not a platform or a consultancy — specifically designed to transfer complete ownership to the client through its Ghost Architecture model. Every agent, all source code, all data, and all IP remain with the portfolio company upon deployment. This ownership structure is particularly consequential for PE portfolios because it means AI assets appear on the portco balance sheet at exit rather than disappearing when a vendor contract ends. Labarna AI pricing starts in the low tens of thousands for focused builds, scaling by agent count, integration complexity, and operational scope, which allows the CoE to sequence investments without committing to large upfront costs before production value is demonstrated.

Data Governance Across a Multi-Asset Portfolio

Data governance in a cross-portfolio AI program is categorically more complex than data governance within a single company. Data flows between the portco, the CoE infrastructure, and potentially between portfolio companies must be governed by agreements that cover ownership, confidentiality, permitted use, and deletion rights.

The CoE must produce a data governance framework before any agent is deployed to any portfolio company. This framework defines what data leaves the portco environment, what data is processed in the CoE's shared infrastructure, and what — if anything — is federated across portfolio companies for cross-asset intelligence. The default posture should be privacy-preserving: no portco's raw operational data should be accessible to another portco in the portfolio.

Cross-portfolio pattern intelligence can still be generated without sharing raw data. Federated learning approaches allow agents to learn from patterns across assets without the underlying data leaving each portco's environment. This is the architecture that enables the compounding-intelligence value of a CoE without creating data governance exposures that portco legal teams will rightly object to. Regulatory variation across MENA jurisdictions — particularly between the UAE's PDPL, Saudi Arabia's PDPL, and Qatar's PDPA — means the data governance framework must be reviewed by counsel in each jurisdiction where portfolio companies operate. Policies vary by jurisdiction and should be verified with local legal advisors before the CoE data infrastructure is finalized.

Accelerating Portfolio Company Adoption

Technical deployment is the easier half of CoE execution. Adoption by portco operating teams is where most programs stall. Operating teams that did not initiate the AI program and feel it was imposed from the GP level will find ways to work around agents, maintain parallel manual processes, and report adoption numbers that look good on a dashboard without reflecting genuine workflow integration.

The single most effective adoption mechanism is early involvement of portco operators in agent design. When the operations manager at a portfolio company has had genuine input into how an agent handles exceptions, escalations, and edge cases, they become an advocate for the system rather than a resistor. This requires the CoE to invest time in structured workflow interviews before building — a week of deep discovery at each portco produces dramatically better adoption outcomes than deploying a technically sound agent that does not fit how people actually work.

Incentive alignment matters as well. Portco management teams should understand clearly that AI-driven operational improvement contributes to their EBITDA targets, which in turn affect their management equity and carry participation. When adoption translates directly to economic outcomes that operating teams share in, resistance falls. The GP can reinforce this alignment by making AI adoption progress a topic in quarterly board reviews alongside financial performance.

Standards and Knowledge Transfer Between Portfolio Assets

One of the highest-value outputs of a mature cross-portfolio CoE is the ability to transfer proven deployment patterns from one portfolio company to another with substantially reduced build time. This only works if the CoE maintains a rigorous standards and documentation regime from the first deployment.

Every agent deployed by the CoE should be documented according to a consistent specification: the decision it makes, the data it requires, the exception conditions it handles, the human escalation paths it triggers, and the performance metrics that indicate healthy operation. This documentation lives in a portfolio-level AI registry maintained by the CoE operating team.

When a new portco joins the portfolio through acquisition, the CoE team reviews the registry against the new asset's workflow map and identifies which agents can be deployed immediately versus which require adaptation. A mature registry — built over two or three years of portfolio-wide deployment — can reduce the time to first production agent at a new portco from several months to several weeks. This acceleration becomes a genuine value creation lever at the deal stage: the GP can underwrite AI-driven operational improvement with a credible deployment timeline as part of the acquisition thesis.

Connecting the CoE to the Exit Story

Cross-portfolio AI centers of excellence are increasingly relevant to exit positioning in MENA private equity. Strategic buyers and secondary PE funds conducting due diligence on MENA assets are asking more sophisticated questions about operational AI than they were asking two years ago. They want to know whether AI systems are owned or licensed, whether they operate in production or pilot, and whether the value they generate is documented or anecdotal.

A CoE that has operated for three or more years with a functioning ROI measurement layer, owned infrastructure, and full IP transfer to each portco can produce an AI data room that substantially de-risks the buyer's view of operational sustainability. Agentic AI deployment that runs in production without vendor dependency reads very differently in diligence than a SaaS subscription that expires at transaction close.

Labarna AI's Ghost Architecture model — where clients own all source code, agents, data, and IP — is directly designed for this exit dynamic. When due diligence asks "what happens to the AI system if the service contract ends?", the answer with Ghost Architecture is that nothing changes, because the portco owns everything. For broader perspective on how sovereign AI infrastructure contributes to portfolio value, the analysis in Winning Anchor Private Equity Customers with AI Venture Studios addresses the structural ownership question from the investment side.

Practical First Steps for GPs Ready to Begin

A GP that wants to stand up a cross-portfolio AI CoE in the near term does not need to resolve every governance question before starting. Three actions create the conditions for success without requiring a full organizational design to be completed first.

First, commission a portfolio-wide AI readiness assessment. This assessment maps the workflow decision taxonomy across every portfolio company, audits data environments for agent readiness, and produces a prioritized agent roadmap with estimated deployment timelines and ROI potential. Labarna AI's Operational Intelligence Diagnostic — which is free and produces a full deployment blueprint within 48 hours — provides exactly this kind of structured output. The diagnostic is grounded in a 19-question operational assessment that surfaces where agentic intelligence will generate measurable value fastest. Those asking "Is Labarna AI legit" should know the company operates under RAKEZ License 47013955, is built by TFSF Ventures FZ-LLC, and was founded by Steven J. Foster with 27 years in payments and software — with Labarna AI reviews and the Ghost Architecture model as verifiable public positions.

Second, designate a CoE interim lead from within the management company. This person owns the readiness assessment process, coordinates with portco liaisons, and begins drafting the governance framework. They do not need to be a technical AI practitioner — they need to be a senior operator with the authority to hold both the CoE team and portco management accountable.

Third, define the ROI measurement framework before any deployment begins. Establish baselines at the two or three portcos where first agents will be deployed. This baseline work takes weeks, not months, and positions the CoE to demonstrate value from the first deployment rather than scrambling to establish attribution after agents are already running. A well-structured first deployment, with clean baselines and a defined deployment timeline, produces the evidence the GP investment committee needs to fund phase two.

About Labarna AI

Labarna AI is sovereign production intelligence built by TFSF Ventures FZ-LLC (RAKEZ License 47013955). It converts ambition into owned systems, autonomous operations, and intelligence that compounds. Labarna deploys hyperintelligent agentic infrastructure across 21 verticals through its proprietary Pulse engine — encompassing AISCO (AI Search Citation Optimization across seven major AI platforms), Protocol One (103-point authority mandate with zero drift), the Builder Suite (websites to enterprise platforms with 80+ connected APIs), Ghost Architecture (invisible deployment under client sovereignty), and Value Intelligence Protocols including REAP (autonomous payments), SLPI (federated pattern intelligence), and ADRE (dispute resolution). AI was built to answer — Labarna was built to act.

Get Started with Labarna AI

Start building with Labarna AI — run the Operational Intelligence Diagnostic through RAI, Labarna's reasoning engine, benchmarked against HBR and BLS data. Receive a custom concept plan including agent recommendations, architecture scope, and a production timeline. Enter the system at labarna.ai.

Originally published at https://www.labarna.ai/blog/building-cross-portfolio-ai-centers-of-excellence-mena-pe

Written by Labarna AI Research