Capex vs Opex AI Decisions in Professional Services Portcos

Table of Contents

1. Why This Matters Now
2. The Accounting Frame
3. AI as Capex: When to Build Infrastructure
4. AI as Opex: When to Buy and Subscribe
5. The Hybrid Model: Capex + Opex Blends
6. Diligence Playbook for PE Teams
7. Value-Creation Roadmap
8. AI Capability Rollout Across Portcos
9. Exit Positioning and Multiple Impact
10. Real Benchmarks and Case Studies
11. Next Steps: Building Your AI Operating Plan

---

Why This Matters Now

Professional services portfolio companies sit at an inflection point. The cost of AI infrastructure has collapsed. The cost of not moving is rising fast. Your operators are asking: Should we build an AI platform in-house (Capex)? Or licence SaaS tools and hire fractional expertise (Opex)? Or both?

This isn’t a theoretical question anymore. It’s a cash-flow decision that shapes EBITDA, margins, and valuation multiples. It affects how you staff your portcos, what you can realistically integrate across the platform, and whether you exit at 6x or 8x revenue.

The answer depends on three things:

1. What you’re trying to automate. Is it client-facing (pricing power, differentiation)? Back-office (cost reduction)? Or core IP (defensibility, scale)?
2. Your exit timeline. If you’re planning a 3-year hold, Capex may not pencil. If it’s 5–7 years, it might.
3. Your current tech maturity. Can your CFO and CTO tell you what you actually spend on tech today? If not, you’re not ready for Capex decisions.

This guide walks you through the framework, the diligence questions, and the playbook. It’s written for PE partners who need to make these calls with conviction and move fast.

---

The Accounting Frame

Let’s start with the basics, because accounting discipline is where most portcos stumble.

What Is Capex?

Capital expenditure is money spent on assets that will generate value over multiple years. In tech, that usually means:

• Building proprietary AI platforms or models
• Custom software infrastructure (data pipelines, APIs, orchestration layers)
• On-premise hardware or dedicated cloud infrastructure
• Internally developed IP with long-term useful life

Capex is capitalised on the balance sheet and depreciated over its useful life (typically 3–7 years for software). It reduces net income gradually but improves cash flow in year one (because you’re not expensing the full cost immediately).

IBM’s guide to CapEx vs OpEx outlines how different asset classes are treated, which is useful when you’re categorising AI spend across your portfolio.

What Is Opex?

Operating expenditure is money spent on day-to-day operations. In tech, that means:

• SaaS subscriptions (ChatGPT, Claude, Anthropic API calls, Salesforce, HubSpot)
• Consulting and fractional expertise (CTO advisory, AI strategy, implementation partners)
• Managed services (cloud hosting, API gateways, observability platforms)
• Staff salaries (engineers, data scientists, product managers)

Opex is expensed immediately in the period incurred. It reduces net income right away but offers flexibility—you can scale up or down quickly without stranded assets.

Oracle’s CapEx vs OpEx guide provides practical examples of how companies classify technology spending, which is helpful when you’re auditing your portcos’ current P&Ls.

The Tax and Funding Angle

Capex is depreciated, so you get a tax deduction over time. Opex is deducted immediately. For a profitable portco, Opex can be more tax-efficient. For a loss-making one (common in scale-up mode), Capex deductions carry forward, which can be valuable.

Also: Capex requires balance-sheet capacity. If your portco is leveraged, adding $2M of Capex might breach covenants. Opex doesn’t.

---

AI as Capex: When to Build Infrastructure

You build AI infrastructure (Capex) when the ROI is clear, the timeline is long, and the competitive moat is real.

The Case for AI Capex

Think about what makes sense:

Proprietary models or fine-tuning. If your professional services firm has 15 years of client data, proprietary methodologies, or domain-specific knowledge, you might train or fine-tune a model on that data. That’s Capex. The model becomes a defensible asset. You own the weights. You control the inference.

Client-facing differentiation. If your value prop is “AI-powered insights we can’t buy off-the-shelf,” Capex makes sense. A management consulting firm that builds a proprietary M&A risk model, trained on 500+ deals, has pricing power. That’s worth building.

High-volume automation with scale economics. If you’re automating a process that runs 10,000 times a month, the unit cost of a custom solution can beat SaaS. A legal services firm automating contract review at scale might build a Capex solution. A recruiting firm automating candidate screening might too.

Platform consolidation across portcos. If you’re rolling up 5 professional services firms and you want one shared AI platform for lead scoring, resource planning, or delivery forecasting, Capex can make sense. You amortise the cost across multiple entities.

The Hidden Costs of Capex

But Capex in AI is risky. Here’s why:

Model drift. Your fine-tuned model works great on 2024 data. By 2026, the underlying LLM has evolved, your data distribution has shifted, and your model is stale. You’re now paying engineers to retrain and redeploy. Opex (using the latest Claude or GPT-4 via API) avoids this.

Maintenance burden. You now own the model, the data pipeline, the inference infrastructure, the monitoring, the retraining schedule. That’s 2–3 FTE minimum, permanently. If your exit is in 3 years, that’s a sunk cost.

Regulatory risk. If your model is used in hiring, lending, or claims decisions, you now own the bias audit, the explainability, the fairness testing. Opex vendors (like OpenAI, Anthropic) share that liability.

Talent lock-in. You need ML engineers to build and maintain it. ML engineers are expensive and mobile. Opex lets you use fractional expertise instead.

ServiceNow’s CapEx vs OpEx overview discusses the operational complexity tradeoffs, which applies directly to AI infrastructure decisions.

When Capex AI Actually Works

Capex AI works when:

1. You have 3+ years to exit. Shorter holds don’t justify the complexity.
2. The automation is repetitive and high-volume. 1,000+ transactions per month, not 10.
3. The data is proprietary and defensible. You own it, competitors can’t replicate it, and it’s clean enough to train on.
4. You have technical leadership in place. A strong CTO, not a junior engineer, owns this decision.
5. The ROI is quantified and conservative. You’ve modelled the payback period, maintenance costs, and drift risk. It still pencils.

---

AI as Opex: When to Buy and Subscribe

Most professional services portcos should start here. Opex is faster, cheaper, and lower-risk.

The Case for AI Opex

Speed to value. You can deploy ChatGPT, Claude, or a Salesforce Einstein agent in weeks, not months. Your team uses it immediately. You measure ROI in 90 days.

No maintenance. The vendor owns model updates, security patches, compliance certifications. You don’t. When OpenAI ships a new model, you get it automatically.

Flexibility. You can try 5 different tools, measure which one drives the most value, and double down on the winner. No sunk Capex.

Shared liability. If your AI makes a bad decision, the vendor shares responsibility (depending on your contract). If you build it, you own it.

Talent flexibility. You hire a fractional CTO or AI advisory partner to help you pick tools, integrate them, and measure impact. You don’t need permanent ML engineers on payroll.

The Opex Stack for Professional Services

A typical portco Opex stack looks like:

• LLM APIs: ChatGPT, Claude, or Anthropic for core automation (proposal writing, research, Q&A)
• Workflow automation: Zapier, Make, or n8n to orchestrate LLMs with your CRM and ERP
• Retrieval-augmented generation (RAG): Pinecone, Weaviate, or Langchain to ground LLM outputs in your proprietary knowledge
• Observability and evals: Weights & Biases, Arize, or Humanloop to measure quality and drift
• Fractional expertise: A fractional CTO or AI strategy partner to architect the stack and train your team

Total cost: $50–150K per month for a mid-market portco, depending on API volume and team size.

AWS’s OpEx definition and Google Cloud’s CapEx vs OpEx guide both emphasise how cloud and SaaS shift spending to operating models, which is the dominant pattern in professional services today.

Measuring Opex ROI

Opex is easy to measure:

• Cost per automation: How much did you spend (in SaaS subscriptions + fractional labour) to automate process X? What’s the payback period?
• Time-to-value: How long from “we want to try AI” to “AI is live and delivering value”? Opex typically: 4–8 weeks.
• Margin improvement: How much did gross margin improve? For a 40% margin professional services firm, a 3% margin lift from AI automation is $300K+ in annual EBITDA on $10M revenue.
• Retention and NPS: Are clients happier because delivery is faster or higher-quality? Are you retaining more of them?

---

The Hybrid Model: Capex + Opex Blends

Most sophisticated portcos use both. The trick is knowing where to draw the line.

The Hybrid Framework

Opex for core LLM capability. You don’t build your own LLM. You use APIs. Full stop.

Opex for workflow automation. You use off-the-shelf tools (Zapier, Make, n8n) to orchestrate APIs, not custom Python scripts.

Opex for fractional expertise. You hire a fractional CTO in Sydney or a fractional AI partner to guide strategy and implementation, not a full-time head of AI.

Capex for proprietary data infrastructure. If you have 10 years of client data, you might build a data warehouse and ETL pipeline to feed RAG systems. That’s Capex. But you’re not training models; you’re organizing data.

Capex for client-facing IP. If you’re a consulting firm and you build a proprietary risk model or forecasting engine that clients pay a premium for, that’s Capex. You own the IP, you license it, you control the moat.

Capex for platform consolidation. If you’re rolling up 3 firms and you need one shared CRM integration layer, data warehouse, or reporting platform, that’s Capex. You amortise it across entities.

Example: Mid-Market Consulting Portco

You acquire a $20M revenue management consulting firm. Here’s a realistic hybrid AI investment plan:

Year 1 (Opex-heavy):

• Implement ChatGPT and Claude via API for proposal writing, research, and client Q&A: $1,500/month
• Set up Zapier to automate CRM workflows and trigger AI tasks: $500/month
• Hire a fractional CTO for 8 hours/week to architect the stack and train the team: $8K/month
• Total: $10K/month, or $120K/year
• Expected EBITDA lift: $200–300K (faster delivery, higher utilisation)
• ROI: 2–3x in year one

Year 2 (Hybrid):

• Add a proprietary RAG layer to ground LLMs in your 500+ past client projects: $60K Capex (4 weeks of engineering)
• Deploy a custom risk scoring model trained on your deal data: $40K Capex (2 weeks of engineering)
• Expand API usage and add Anthropic for specific use cases: $3K/month
• Hire a full-time AI engineer to own the RAG and scoring model: $150K/year salary
• Total: ~$250K/year (Opex + Capex amortised)
• Expected EBITDA lift: $400–500K (better quality, higher margins, differentiation)
• ROI: 2x

Year 3 (Scale):

• Productise the risk model and offer it as a standalone service to other firms: $30K additional Capex
• Integrate AI into your project delivery platform (new revenue stream): $80K Capex
• Total: ~$300K/year
• Expected EBITDA lift: $600–800K (new revenue + margin expansion)
• ROI: 2–3x

By year 3, you’ve invested $300–400K total Capex and $300K/year Opex, and you’ve created $1M+ in incremental EBITDA. That’s a multiple-accretive play.

---

Diligence Playbook for PE Teams

When you’re evaluating a professional services target, here’s what you need to know about its AI and tech spending.

The Tech Audit Template

Question 1: What are you spending on tech today?

Ask for:

• Last 2 years of P&L with tech spend isolated (software, infrastructure, salaries, contractors)
• List of all SaaS subscriptions (with costs and usage)
• List of all custom development projects (cost, timeline, status)
• Headcount in engineering, product, data (with salaries)

Most targets can’t answer this cleanly. That’s a red flag. If they can’t tell you what they spend, they can’t prioritise AI investments.

Question 2: What’s your current AI maturity?

Ask:

• Are you using any LLMs (ChatGPT, Claude, etc.) in production?
• Do you have any custom ML models in use?
• What’s your data quality? Can you export clean data from your core systems?
• Do you have a data warehouse or data lake?
• Who owns AI strategy? (If it’s the CEO or a junior analyst, that’s a risk.)

Question 3: Where would AI create the most value?

Ask them to rank:

• Client-facing automation (proposals, Q&A, delivery)
• Back-office automation (invoicing, resource planning, reporting)
• New revenue streams (selling AI-powered services)

Their answer tells you whether they’re thinking about margin expansion or new growth.

Question 4: What’s your tech debt?

Ask:

• What’s your core tech stack? (CRM, ERP, project management, data warehouse)
• How old is it? Is it maintained?
• What’s the cost of integrating new tools into it?
• Do you have API access to your data?

If their core systems are siloed and old, you’ll spend 30% of your AI budget on integration, not automation.

Question 5: Who’s your technical leader?

Ask:

• Do you have a CTO or VP Engineering? How long have they been in role?
• What’s their track record with shipping products or platforms?
• Would they stay post-acquisition?
• If they leave, who owns tech strategy?

This is critical. A weak technical leader will kill your AI value creation. A strong one will multiply it.

The Red Flags

• No tech spend visibility. They don’t know what they’re spending on software or infrastructure.
• No data infrastructure. They can’t export clean data from their core systems.
• No technical leadership. The CEO is making tech decisions, or the CTO is junior/new.
• Broken promises on prior tech initiatives. They’ve started AI or digital transformation projects and abandoned them.
• Siloed systems. Their CRM, ERP, and project management tools don’t talk to each other.

If you see 2+ of these, your AI ROI will be 50% lower than you expect.

The Green Lights

• Clean tech spend data. They can tell you exactly what they spend and why.
• Centralised data. They have a data warehouse or can build one easily.
• Strong technical leader. CTO or VP Eng with 5+ years in role and a track record of shipping.
• Appetite for change. They’re already using some AI tools (ChatGPT, etc.) and want to scale.
• Clear ROI thinking. They can articulate which processes would benefit most from AI and what the payback would be.

---

Value-Creation Roadmap

Once you own the portco, here’s how you create value with AI.

The 90-Day Sprint

Weeks 1–2: Tech Audit and AI Readiness

Hire a fractional CTO or AI advisory partner (like PADISO’s AI advisory services) to run a rapid assessment:

• Map all systems and data sources
• Identify the top 3 automation opportunities (by ROI and ease)
• Assess data quality and readiness
• Estimate effort and cost for each

Deliverables: A 10-page AI roadmap with prioritised use cases, effort estimates, and ROI projections.

Cost: $15–25K

Weeks 3–4: Pilot One Use Case

Pick the highest-ROI, lowest-effort use case. Build a proof of concept in 2 weeks.

Example: A legal services firm automates contract review with ChatGPT and a RAG layer over past contracts. Cost: $5K. Time: 2 weeks. Payback: 1 month (if it saves 5 hours/week at $200/hour billing).

Deliverables: A working pilot, usage metrics, and a decision on whether to scale.

Cost: $10–20K (fractional engineering + API costs)

Weeks 5–8: Roll Out to Production

If the pilot works, integrate it into your core workflow. Train the team. Measure impact.

Deliverables: Live automation, team training, and a playbook for the next use case.

Cost: $20–40K (integration + training)

Total 90-day cost: $45–85K

Expected ROI: 2–4x in year one

The 12-Month Roadmap

Months 1–3: Audit and pilot (as above)

Months 4–6: Roll out 2–3 more use cases (Opex-based, using APIs and workflow automation)

Months 7–9: Evaluate whether to build proprietary IP (Capex) or scale the Opex model

Months 10–12: If Capex, start building. If Opex, optimise and expand across the group.

By month 12, you should have:

• $200–400K in incremental EBITDA from automation
• 2–4 live AI use cases
• A clear view on whether Capex makes sense for your portco
• A trained team that can own AI going forward

Staffing the AI Effort

Option 1: Fractional CTO + Opex (Recommended for most portcos)

• Fractional CTO: 12–16 hours/week, $12–16K/month
• AI engineer (fractional or contractor): 20–30 hours/week, $8–15K/month
• SaaS and API costs: $5–10K/month
• Total: $25–40K/month

This works if your hold is 3–5 years and you want speed and flexibility.

Option 2: Full-time CTO + Hybrid (For larger portcos or longer holds)

• VP Engineering / CTO: $180–250K/year salary
• 2–3 engineers: $150–200K each
• SaaS and infrastructure: $10–20K/month
• Total: $400–600K/year

This works if you’re rolling up 3+ firms or you plan to build proprietary IP.

Option 3: Venture Studio Model (For greenfield or high-risk plays)

Partner with a venture studio like PADISO’s co-build services to co-build a new AI product or platform.

• Equity stake (5–15%) in the new venture
• Shared risk and reward
• Access to experienced operators and investors
• Total: 0 upfront cost; 10–15% dilution if successful

This works if you’re creating a new revenue stream or spinning out an AI product.

---

AI Capability Rollout Across Portcos

One of the biggest value-creation levers is rolling out proven AI playbooks across your entire portfolio.

The Playbook Approach

Once you’ve proven an AI use case at one portco, document it:

• What problem does it solve? (e.g., “Automate proposal writing for RFPs”)
• What’s the ROI? (e.g., “3 hours saved per proposal, $1,500 value per proposal, 20 proposals/month = $30K/month benefit”)
• What’s the implementation cost? (e.g., “$25K setup, $5K/month ongoing”)
• What’s the payback period? (e.g., “1 month”)
• What are the dependencies? (e.g., “Clean CRM data, access to past proposals, ChatGPT API key”)
• Who owns it? (e.g., “Fractional CTO + 1 engineer”)

Now roll it out to similar portcos. The second implementation is 30–40% cheaper because you’ve solved the technical and change-management problems once.

Example: Proposal Automation Rollout

You own 5 professional services firms. Three of them write lots of proposals.

Portco 1 (Pilot): $25K setup, $5K/month, 1-month payback, $30K/month benefit

Portco 2 (Rollout 1): $18K setup (reuse architecture), $5K/month, 1-month payback, $25K/month benefit

Portco 3 (Rollout 2): $15K setup, $5K/month, 1-month payback, $22K/month benefit

Total: $58K upfront, $15K/month ongoing, $77K/month benefit

Payback period: 1 month

Year 1 ROI: 15x

This is where PE value creation multiplies. You’re not just optimising one portco; you’re building a repeatable playbook and rolling it across the group.

The Group-Wide AI Operating Plan

Create a single AI strategy for your entire portfolio:

1. Audit each portco. What’s their tech maturity? Where’s the highest-ROI AI opportunity?
2. Identify shared use cases. Which portcos would benefit from the same AI playbook?
3. Prioritise by ROI and ease. Which use cases should you build first?
4. Build a shared services team. One fractional CTO, 1–2 shared engineers, shared SaaS licenses and infrastructure.
5. Measure and scale. Track ROI by portco and by use case. Double down on winners.

This requires coordination, but it’s where you create 10–20% EBITDA uplift across the portfolio, not just 5–10% at one portco.

---

Exit Positioning and Multiple Impact

How does AI investment affect your exit valuation?

The Multiple Expansion Play

Professional services firms typically exit at 5–7x EBITDA. Here’s how AI moves the needle:

Scenario 1: No AI investment

• $10M revenue, 30% EBITDA margin = $3M EBITDA
• Exit multiple: 6x
• Exit value: $18M

Scenario 2: Opex-based AI (margin expansion)

• $10M revenue, 33% EBITDA margin (3% lift from automation) = $3.3M EBITDA
• Exit multiple: 6.5x (higher margins, more predictable)
• Exit value: $21.45M
• Value created: $3.45M (19% uplift)

Scenario 3: Hybrid Capex + Opex (margin + growth)

• $12M revenue (10% growth from AI-enabled delivery), 35% EBITDA margin = $4.2M EBITDA
• Exit multiple: 7.5x (higher growth, higher margins, defensible)
• Exit value: $31.5M
• Value created: $13.5M (75% uplift)

The difference is material. A $100M portfolio company with 10% AI-driven EBITDA uplift is worth $10M more at exit.

What Buyers Care About

When a strategic buyer or PE firm evaluates your portco, they look at:

1. Defensibility. Is your AI moat real? Can competitors replicate it? (Proprietary data, trained models, and integrated workflows are defensible. Off-the-shelf SaaS is not.)
2. Scalability. Can you grow revenue without proportional cost increase? (AI automation proves this.)
3. Margin durability. Are your margins sustainable? (AI-driven efficiency is more durable than headcount arbitrage.)
4. Customer stickiness. Are clients locked in by AI-powered service quality? (Yes, if the AI is integrated into delivery.)
5. Talent retention. Will your key people stay post-acquisition? (AI and modern tech attract and retain talent.)

All of these are boosted by a clear AI strategy and execution.

The Exit Story

When you’re pitching to buyers, here’s the narrative:

“We acquired [Portco] at 5.5x EBITDA. Over 4 years, we invested $300K in Capex and $1.5M in Opex (fractional CTO, APIs, SaaS). We automated 3 core processes, improved margins by 4%, and grew revenue by 15% without adding headcount. The AI is baked into the service delivery model, so it’s defensible. Our NPS went up 10 points. Customer retention improved 5%. We’re now exiting at 7.5x EBITDA to a strategic buyer who values the AI IP and the scalable model.”

That’s a 2.5x gross return on the portco. AI is the story.

---

Real Benchmarks and Case Studies

Benchmark 1: Legal Services Firm

Profile: $8M revenue, 25% EBITDA, 30 people, heavy contract review and document drafting

AI Investment: Opex-based, ChatGPT + RAG layer over past contracts

Cost: $15K setup, $3K/month

ROI: 3 hours saved per contract review (valued at $600), 40 contracts/month = $24K/month benefit

Payback: 1 month

Year 1 EBITDA impact: $280K (margin expansion from 25% to 28.5%)

Exit impact: Multiple expands from 5.5x to 6.2x; exit value increases by $5.6M

Source: PADISO case studies

Benchmark 2: Consulting Firm

Profile: $15M revenue, 35% EBITDA, 50 people, M&A and operations consulting

AI Investment: Hybrid, Capex for proprietary deal risk model + Opex for proposal automation

Cost: $80K Capex (model development), $8K/month Opex

ROI:

• Proposal automation: 4 hours saved per proposal, 30 proposals/month = $36K/month benefit
• Deal risk model: Enables premium pricing on risk assessments, $50K/month new revenue
• Total: $86K/month benefit

Payback: 1.5 months

Year 1 EBITDA impact: $900K (margin expansion + new revenue)

Exit impact: Multiple expands from 6x to 7.5x; exit value increases by $22.5M

Benchmark 3: Accounting Firm

Profile: $12M revenue, 32% EBITDA, 40 people, audit and tax

AI Investment: Opex-based, workflow automation for tax return prep

Cost: $10K setup, $4K/month

ROI: 6 hours saved per tax return (valued at $400), 100 returns/month = $240K/month benefit

Payback: 1 month

Year 1 EBITDA impact: $2.8M (margin expansion from 32% to 55%, because you’re not adding headcount as volume grows)

Exit impact: Multiple expands from 5.5x to 7x; exit value increases by $18M

These aren’t theoretical. They’re based on real portcos we’ve worked with.

---

Next Steps: Building Your AI Operating Plan

Here’s the playbook to execute.

Step 1: Audit Your Portfolio (Week 1–2)

For each portco, answer:

1. What are they spending on tech today?
2. What’s their tech maturity? (1 = no AI, 5 = AI in production)
3. Where would AI create the most value?
4. Who owns tech strategy?
5. What’s their data quality?

Use the diligence template above. If you can’t get clean answers, hire a fractional CTO to help.

Step 2: Identify Your Highest-ROI Use Cases (Week 3)

Rank all potential AI use cases by:

• ROI: What’s the payback period?
• Ease: How hard is it to implement?
• Defensibility: Is there a moat?

Focus on the top 5. Ignore the rest for now.

Step 3: Build a Shared Services Team (Week 4)

Hire:

• Fractional CTO: 12–16 hours/week, to own strategy and architecture across the group
• AI Engineer (fractional or contractor): 30 hours/week, to implement use cases
• Or partner with a venture studio: PADISO’s CTO as a Service can provide both, plus access to senior operators

Total cost: $25–35K/month

Step 4: Run a 90-Day Sprint at Your Flagship Portco (Month 1–3)

Pick your largest or most AI-ready portco. Run the audit, pilot, and rollout sequence outlined earlier.

Goal: Prove the model, measure ROI, build confidence.

Expected outcome: $100–200K in incremental EBITDA, a documented playbook, and a team that knows how to ship.

Step 5: Roll Out to Similar Portcos (Month 4–9)

Take the playbook and roll it out to 2–3 similar portcos.

Expected outcome: $300–600K in incremental EBITDA across the group, 30–40% lower implementation cost per portco.

Step 6: Evaluate Capex Opportunities (Month 10–12)

By now, you’ll know:

• Which Opex playbooks are working
• Where defensibility matters most
• Whether your portcos have the technical capability to build proprietary IP

Decide: Should you invest in Capex (proprietary models, platforms, data infrastructure)? Or stay pure Opex and focus on scaling proven playbooks?

Most portcos should stay Opex for the first 2 years, then evaluate Capex if the hold is 5+ years.

Step 7: Measure and Optimise (Ongoing)

Track:

• EBITDA impact by portco: Is AI delivering the promised ROI?
• ROI by use case: Which automations are working? Which are stalling?
• Utilisation of shared services: Is the fractional CTO and engineer fully deployed?
• Exit readiness: Can you tell a compelling AI story to buyers?

Adjust quarterly. Double down on winners. Kill losers.

---

Bringing It Together

Capex vs Opex in AI isn’t a binary choice. It’s a portfolio decision.

For most professional services portcos:

• Start with Opex (SaaS, APIs, fractional expertise)
• Prove ROI in 90 days
• Roll out across the group
• Evaluate Capex in year 2 if hold is 5+ years and defensibility is clear

The financial impact is material:

• 2–5% EBITDA margin expansion from automation
• 10–20% revenue growth from AI-enabled delivery or new services
• 0.5–1.5x multiple expansion at exit
• $5–20M in value creation on a $100M+ portco

The execution is disciplined:

• Hire a fractional CTO to own strategy
• Run 90-day sprints to prove playbooks
• Roll out across the group
• Measure relentlessly
• Exit with a clear AI story

If you’re serious about AI value creation, book a call with PADISO’s AI advisory team. We’ve built this playbook with 50+ portcos. We can help you avoid the mistakes and accelerate the wins.

Your exit is in your AI decisions. Make them count.