For many SaaS founders and portfolio operators, the AI conversation has already shifted.
The early phase was about experimentation: trying ChatGPT, drafting content faster, summarizing meetings, and automating small tasks. But that phase is ending. The harder question now is not whether companies will use AI. It is how leaders govern it without slowing down the business.
That was the core tension explored in a recent conversation with Philip Belcher, a former Cisco leader turned CEO advisor. His perspective is useful because it avoids both extremes: the boosterism that treats AI as magic and the panic that treats it as uncontrollable. Instead, he frames AI as something more familiar to executives: another major technology wave that creates competitive upside only when paired with leadership, strategy, execution, and governance.
For executives, investors, operating partners, and CFOs/CROs, that framing matters. The winners in AI will not simply be the firms with the most tools. They will be the organizations that integrate AI into operating models, keep humans accountable, and preserve customer trust while improving speed and efficiency.
Key Takeaways
- AI is not a strategy. It is an enabling capability that should support a clear business purpose, strategic priorities, and execution plan.
- Governance must start at the leadership level. AI risk is not just technical; it is operational, reputational, legal, and cultural.
- Generative AI and agentic AI are fundamentally different. One helps create outputs; the other increasingly acts inside workflows and systems.
- The bigger opportunity is in embedded AI. Durable value often comes from AI operating inside finance, customer service, and operational processes.
- Ignoring AI is now a strategic risk. Leaders do not need deep technical mastery, but they do need enough fluency to ask the right questions.
- Verification remains essential. AI can accelerate analysis and preparation, but outputs still require human review.
- Customer experience is the test. If AI improves responsiveness, accuracy, and service, it creates value. If it adds friction, it destroys it.
- People and processes remain the leading indicators. AI adoption works best when tied to culture, training, workflow design, and accountability.
- Over-automation can backfire. Removing every human touch may look efficient internally while weakening trust externally.
- Action item: Build an AI governance framework that defines approved use cases, escalation paths, review controls, and ownership by function.
sbb-itb-01010c0
AI Is the Next Wave, Not the Final Destination
Belcher’s most grounded point is also the most important: AI should be understood as part of a longer technology arc.
Executives have seen this pattern before. The internet, CRM platforms, ERP systems, and automation tools all arrived with outsized claims. Some did reshape industries. Some disappointed. Most created value only after businesses learned how to implement them intelligently.
That is the right lens for AI.
Treating AI as a standalone miracle tends to produce shallow adoption: scattered pilots, duplicated tools, no process redesign, and no accountability. Treating it as "just another tool" can be equally dangerous if that leads to underestimating the speed of change.
A more useful view is this:
- AI is foundational
- AI is not self-justifying
- AI creates leverage only when attached to business outcomes
This distinction matters for boards and executive teams. If the strategic conversation is centered on "What AI tools should we buy?" the company is already asking the wrong question. The better question is: Where can intelligence, automation, and decision support materially improve our economics, resilience, or customer experience?
Generative AI Gets Attention. Agentic AI Changes Operations.
One of the sharper distinctions in the discussion was between generative AI and agentic AI.
Generative AI is now familiar. It drafts, summarizes, researches, and produces first-pass outputs. Most companies are already experimenting with it in marketing, sales, support, and internal productivity.
Agentic AI is more consequential.
In practical terms, this refers to AI embedded in systems and workflows that can make decisions, trigger actions, and execute tasks with a degree of autonomy. In a SaaS environment, that can mean:
- triaging inbound support issues
- handling billing follow-up
- routing customer requests
- updating CRM records
- flagging churn signals
- managing workflow exceptions
- supporting collections and revenue operations
For a CFO or CRO, this is where AI stops being a novelty and starts becoming an operating model issue.
The opportunity is clear: if AI can absorb repetitive, rules-based, context-rich work, teams can spend more time on judgment, customer engagement, and strategic priorities.
But that same shift creates the core governance challenge. Once AI moves from content generation into decision-making and process execution, leaders have to answer more difficult questions:
- What decisions can AI make on its own?
- What confidence threshold is required?
- When must a human review or override?
- Who is accountable if the outcome is wrong?
- How is the system monitored over time?
This is where many companies are currently exposed. They are adopting AI faster than they are defining authority, controls, or ownership.
Governance Is Not a Brake on Innovation. It Is What Makes Scaling Possible.
Belcher’s central argument is that all technology requires governance, and AI simply raises the stakes because it can learn, adapt, and increasingly act.
That is an executive issue, not merely an engineering issue.
In PE-backed and growth-stage SaaS businesses, AI governance should be treated as a cross-functional operating requirement with implications for:
- compliance
- financial controls
- data privacy
- customer commitments
- workforce design
- brand reputation
- model risk
- vendor management
This matters because governance failures rarely show up first as technical failures. They show up as business failures:
- incorrect customer communication
- flawed financial categorization
- biased recommendations
- unauthorized actions
- inconsistent service outcomes
- employee confusion about what is allowed
- reputational damage when something "automated" goes public
One example raised in the conversation involved conflicting guidance in higher education: some classes banning AI, others requiring it, while institutions simultaneously buy enterprise licenses. The point is broader than academia. Many organizations are rolling out AI before aligning policy, incentives, and operating norms.
That creates ambiguity, and ambiguity is expensive.
What good AI governance looks like in practice
For executive teams, governance does not need to begin as a heavy bureaucracy. It should begin as a clear management system.
At minimum, that system should define:
1. Approved use cases
What AI is permitted to do in sales, finance, support, HR, and product - and what is off-limits.
2. Decision rights
Which actions require human approval, and which can be automated within guardrails.
3. Data rules
What internal, customer, or regulated data can be used with which tools and vendors.
4. Validation standards
How outputs are checked for accuracy, bias, completeness, or compliance risk.
5. Monitoring and escalation
How errors are reported, corrected, and used to improve future controls.
6. Functional accountability
Who owns AI outcomes in each department: not just IT, but the business leader closest to the process.
The key insight: governance should increase confidence, not slow every experiment to a halt.
The Real Competitive Divide: Businesses With a Plan vs. Businesses Chasing Noise
One of the strongest parts of the discussion had little to do with AI directly. It had to do with planning.
Belcher emphasized a point many executives know but too few enforce: businesses need a clear purpose, vision, mission, and strategy. Without that, every new tool looks important, every vendor pitch sounds urgent, and every competitor announcement triggers panic.
That is especially true in AI, where the barrier to entry feels low and the number of ideas is exploding.
This is why many leadership teams are currently overwhelmed. They are not drowning in technology; they are drowning in unprioritized possibilities.
For founders and operators, AI should be filtered through a short list of strategic tests:
- Does this support our business model?
- Does it solve a real workflow bottleneck?
- Does it improve unit economics or customer outcomes?
- Does it strengthen differentiation or merely copy the market?
- Can we govern it responsibly at our current maturity level?
Without that filter, organizations end up with "AI theater": pilots with no owner, dashboards with no decisions, and subscriptions with no ROI.
This is especially relevant for private equity environments, where value creation depends on focus. AI should not become another portfolio-wide mandate that produces presentations instead of measurable impact. It should be deployed where it can affect throughput, margin, retention, pricing, service quality, or decision speed.
Why Customer Experience Still Decides the Outcome
A useful counterbalance in the conversation was Belcher’s insistence that technology does not erase the fundamentals of business. Customers still evaluate companies based on whether their needs are understood and resolved.
That sounds obvious, but AI adoption often breaks this rule.
Teams can become so focused on reducing labor or increasing volume that they forget the customer’s side of the interaction. A process may become faster internally while becoming colder, more confusing, or less trustworthy externally.
This is where many automation initiatives fail.
Belcher noted that organizations that understand sales well usually understand something deeper: the business exists to serve customer needs profitably. AI can help with that. It can reduce administrative drag, speed preparation, improve responsiveness, and surface insights before a meeting or customer conversation.
But if AI creates more friction - bad invoices, robotic support loops, impersonal outreach, opaque decisioning - it damages the very thing it was supposed to improve.
For revenue leaders, this suggests a simple principle:
Use AI to remove work from your team, not care from your customer experience.
In practice, that means:
- automate preparation, not empathy
- automate classification, not accountability
- automate routing, not relationship ownership
- automate repetitive follow-up, not complex judgment
The companies that get this right will likely outperform because they will pair machine efficiency with human clarity.
The ERP Warning Applies to AI
One of the most useful analogies in the discussion came from older enterprise technology rollouts.
Belcher referenced the long history of ERP implementations that failed to produce expected value. The lesson was not that ERP was useless. The lesson was that technology underdelivers when it is installed for the wrong reasons or without process alignment.
The same warning applies to AI.
If AI is introduced merely to cut costs, eliminate headcount, or signal innovation to the market, it will often trigger internal resistance and customer friction. If it is connected to well-understood workflows, trained teams, and measurable outcomes, it has a far better chance of creating value.
This is especially important in finance and operations functions, where leaders are rightly cautious. AI can help with categorization, forecasting support, reconciliation prep, workflow routing, and exception handling. But those benefits only hold if the underlying process is already reasonably coherent.
Automating a broken workflow does not fix it. It usually scales the confusion.
Human Review Is Still Non-Negotiable
Another practical point from the conversation: AI can be impressively useful and surprisingly wrong at the same time.
Belcher gave a simple example of using AI to analyze a spreadsheet, only to find that most of the categorization was incorrect and required manual correction. That is an important reminder for executives hearing overly broad claims about autonomy.
AI can accelerate:
- research
- preparation
- summarization
- first-pass analysis
- pattern detection
But acceleration is not the same as reliability.
For CFOs, CROs, and operating partners, this means AI outputs should be classified the same way any decision-support system would be classified:
- low-risk outputs can be lightly reviewed
- medium-risk outputs need structured validation
- high-risk outputs require explicit human approval
The problem is not that AI makes mistakes. Every system does. The problem is when organizations stop designing for mistake detection.
That is where governance and process discipline matter most.
Leaders Can No Longer Opt Out
A recurring theme in the discussion was that leaders can no longer hide behind technical discomfort.
They do not need to become machine-learning specialists. But they do need enough familiarity to govern AI sensibly, ask good questions, and make trade-offs intelligently.
This is a major cultural shift.
In earlier eras, some executives could treat technology as someone else’s domain. That is no longer viable. AI now touches decision quality, labor design, customer interaction, and competitive positioning too directly.
For founders and executives, the new minimum standard is not expertise. It is informed stewardship.
That includes being able to ask:
- Where are we using AI today?
- Which processes rely on it?
- What data is involved?
- What controls exist?
- What outcomes are we measuring?
- Where could it create strategic upside?
- Where could it create hidden downside?
Leaders who cannot answer those questions are not neutral. They are exposed.
A Better Operating Model: People, Processes, Customers, Financials
Toward the end of the discussion, Belcher referenced the balanced scorecard framework and highlighted a sequence that deserves more attention in AI conversations.
He emphasized that sustainable performance tends to flow in this order:
- People
- Processes
- Customer outcomes
- Financial outcomes
That ordering is highly relevant for AI deployment.
Too many AI strategies begin at step four. They start with cost savings or revenue hopes. But better adoption usually starts at the top:
People
Do teams understand the tools, risks, and expected behaviors?
Processes
Are workflows clear enough for AI to assist without creating errors or confusion?
Customer outcomes
Will the change improve speed, service, consistency, or insight in ways the customer can feel?
Financial outcomes
Will those improvements translate into retention, growth, margin, or efficiency?
This sequence gives executives a more disciplined roadmap. It also explains why many AI initiatives underperform: they skip the organizational groundwork and try to jump straight to financial impact.
What SaaS Leaders Should Do Next
The discussion points to a practical agenda for SaaS leadership teams and portfolio operators.
1. Separate experimentation from production
Let teams test tools, but define what counts as production use and what governance is required before broader rollout.
2. Prioritize workflow-level value
Look beyond generic chat interfaces. The bigger wins often sit inside billing, support, reporting, pipeline management, collections, and customer operations.
3. Put human checkpoints around consequential decisions
The higher the customer, legal, or financial risk, the stronger the review requirement should be.
4. Build policy before scale creates inconsistency
Do not wait for widespread use before setting rules on data handling, approved tools, and escalation paths.
5. Measure outcomes that matter
Track time saved, error rates, response times, conversion efficiency, retention signals, and employee adoption - not just logins or prompts.
6. Keep customer trust visible
Test whether AI is making interactions easier or simply more automated. Those are not the same thing.
7. Raise leadership fluency
Executive teams should schedule regular reviews of AI use cases, risks, lessons learned, and changing competitive dynamics.
Conclusion: AI Success Will Belong to the Best-Governed Operators
The most valuable idea in this conversation is also the least flashy: AI success is a management challenge before it becomes a technology win.
Yes, AI is moving fast. Yes, the tools are improving. And yes, agentic systems will likely reshape operating models across SaaS, services, and enterprise functions.
But the firms that benefit most will not be the loudest adopters. They will be the ones that combine curiosity with discipline, speed with controls, and automation with accountability.
In other words, the future of AI in business may look less like a technology race and more like an execution test.
For executives, investors, and operators, that is good news. Execution is governable. And governance, done well, is not what slows innovation down. It is what makes innovation durable.
Source: "AI Governance for SaaS Founders: Why Agentic AI Needs Leadership, Not Just Adoption" - heradigital, YouTube, Jul 3, 2026 - https://www.youtube.com/watch?v=W_yX6YtCiE8
